View Full Version : Physics Processor - AGEIA (Cool Demo)
http://www.ageia.com/pdf/wp_2005_3_physics_gameplay.pdf
I have to wonder where Nvidia and ATI were on this one. You have to think that this would be necessary when the games are becoming CPU limited.
http://www.theinquirer.net/?article=21648
Edit: Here is a cool physics demo....
http://www.ageia.com/novodex_downloads.html#
trinibwoy
08-Mar-2005, 04:02
Interesting. I'm amazed that there are titles in the works that take advantage of this technology and this is the first we're hearing about it.
So how would it work - the CPU does the AI, feeds the PPU which in turn feeds the GPU?
digitalwanderer
08-Mar-2005, 04:07
Ooooh, interesting! :D
Thanks for the link, frontpaged and sourced....and I WANT one of these!
trinibwoy
08-Mar-2005, 04:10
I'm assuming the Novodex physics API will have software fallbacks for those without these add-in cards right?
I don't even want to imagine the name of a Gainward card with integrated GPU and PPU :)
DudeMiester
08-Mar-2005, 04:21
I wonder if 3DRealms is going to rewrite half of DNF to take advantage of this? lol!
Still very cool, my only concern is the amount of bandwidth that will be needed to move all the physics data between the CPU/GPU/PPU. With PCIe it shouldn't be a huge problem, but with PCI I just can see how it's feasable.
PhysX Processor Architecture has been designed to enable radical acceleration of:
Rigid body dynamics
Universal collision detection
Finite element analysis
Soft body dynamics
Fluid dynamics
Hair simulation
Clothing simulation
NovodeX Physics SDK
The NovodeX Physics SDK is a stable, high-performance solution for game developers to enable physics-based gameplay and effects in PC and console titles. A powerful API for the PhysX PPU, NovodeX enables game developers to inject both software-only and hardware-accelerated features into their games. The NovodeX Physics SDK is also the first and only asynchronous (multithreaded) physics API capable of unleashing the power of multiprocessor gaming systems.
Key features:
Stable, high-performance solver
First and only multiprocessing physics API
PC and console support
Works alongside other game engines
Supports vehicles, rag dolls, and character controllers
Integrates with any renderer
Full complement of code samples and tutorials
World-class developer support and custom solutions
http://www.ageia.com/technology.html
Now why aren't Nvidia and ATI doing this?
Very interesting but how much.
PC-Engine
08-Mar-2005, 04:55
Now why aren't Nvidia and ATI doing this?
Maybe because they only have so many resources? Maybe they don't have access/experience with physics engines?
IgnorancePersonified
08-Mar-2005, 05:35
Just had a quick look but are other Physics engine's able to take advantage of this like Havok or would they have to be rewritten? Is this proprietary as well - the Chip plus the API? so any other Physics engine would need to license it to be able to use it?
What a great idea :D
Looking forward to this sort of thing being properly utilised.
Coming soon to a DirectX release near you: Direct_Physics :twisted:
I just glossed over the pdf linked at the top. It mentioned that it would provide a better solution than running physics calculations in parallel on a separate CPU core. Given that current CPUs already have vector processing functionality, it is safe to assume that a dedicated physics processor would incorporate something further. What immediately comes to mind is something analogous to the various fixed function pipelines (T&L, texture filtering and compression, etc) in graphics cards.
Does this sound right?
Rigid body dynamics
Universal collision detection
Finite element analysis
Soft body dynamics
Fluid dynamics
Hair simulation
Clothing simulation
Looks like just a math chip with lots and lots of parallel math units.
Cool they have a fairly neat downloadable demo.
http://www.ageia.com/novodex_downloads.html#
Wow the rocket demos are most impressive. Shows how badly we need a wicked geometry processor.
There are 47 demos. Some of the Geometry demos have dominos that you can knock over and hundreds of dominos fall over. There are also exploding buildings made of thousands of bricks.
The rag doll demos are kind of kool as well because they seem like they are made of rubber.
The road kill demo is interesting.
The tread is really cool.
mboeller
08-Mar-2005, 08:31
so not only the XBox2 will have an PPU. Cool.
Simon F
08-Mar-2005, 08:39
Slightly OT, but does anyone else here hear the tune "Maria" from West Side Story playing in their head whenever they read the name "Ageia"? :|
Not until you just mentioned it. :?
"Ageia,
I've run some code from Ageia,
And physics, I've found,
Has a higher upper bound, for me."
Back on topic:
FWIW, here's an earlier thread on the same topic/software. (http://www.beyond3d.com/forum/viewtopic.php?p=330423#330423)
And I'm at work and can't open any of those links :x
Titanio
08-Mar-2005, 10:15
Is Havok going to support this?
And I'm at work and can't open any of those links :x
Maybe you should get some work done instead then? Hang on a minute I'm at work too.. :lol:
Regarding this processor, is it going to be available for the PC Platform too?
so not only the XBox2 will have an PPU. Cool.
How do you know Xbox2 will have a PPU?
Dave B(TotalVR)
08-Mar-2005, 12:24
Is opengl gonna support this?
digitalwanderer
08-Mar-2005, 12:28
so not only the XBox2 will have an PPU. Cool.
How do you know Xbox2 will have a PPU?
No clue, but if he's right it's a very GOOD THING(tm) for the whole PPU thing. :)
digitalwanderer
08-Mar-2005, 12:31
Will the PPU only do physics or would it be possible to use it for something like AI too? :|
PC-Engine
08-Mar-2005, 12:37
Will the PPU only do physics or would it be possible to use it for something like AI too? :|
The AI is probably better left to the cpu since the cpu is better suited for that kinda of stuff and it would be idling anyway. :lol:
Ageia/PhysX card shots and interview at Gamer's Depot
http://www.gamers-depot.com/interviews/agiea/001.htm
digitalwanderer
08-Mar-2005, 12:52
So it does have it's own molex, and what is that lil connector that looks like an SLI connector or something on the 'back' of the card? (Opposite the PCI connector)
More links:
Ageia announces PhysX processing unit
http://www.gamespot.com/news/2005/03/08/news_6119895.html
Q&A: Epic Games' Tim Sweeney on Ageia PhysX
http://www.gamespot.com/news/2005/03/08/news_6119896.html
EDIT:
Reality Engine Integrates NovodeX Physics (from Feb 22nd)
http://www.artificialstudios.com/news.php
Ubisoft And AGEIA Join Forces To Re-Ignite Game Innovation
http://biz.yahoo.com/prnews/050308/sftu106_1.html
AGEIA To Redefine Realism With Hardware-Accelerated Physics
http://biz.yahoo.com/prnews/050308/sftu104_1.html
This thing is likely more general purpose as a GPU and the only thing which gets to use it is a physics engine? What a waste.
mboeller
08-Mar-2005, 16:10
so not only the XBox2 will have an PPU. Cool.
How do you know Xbox2 will have a PPU?
I don't, but if you look at a few of DeanoC's comments about the XBox2 I got the really strong impression that the XBox2 will have something like an PPU, but most likely embedded in the CPU.
This thing is likely more general purpose as a GPU and the only thing which gets to use it is a physics engine? What a waste.
On the contrary I think it is not so bad of idea. Provided that it is widely adapted by industry.
Mulciber
08-Mar-2005, 18:05
So it does have it's own molex, and what is that lil connector that looks like an SLI connector or something on the 'back' of the card? (Opposite the PCI connector)
I believe thats a PCI-E 1x interface. Seems like it has 2 interfaces for some reason :?: :?
edit: also, i would guess it only needs the molex connection if it is installed into a PCI slot, going by the placement
Geeforcer
08-Mar-2005, 18:08
Would it not be more practical to increase general processing power (CPU) instead of adding such narrowly targeted processor?
Would it not be more practical to increase general processing power (CPU) instead of adding such narrowly targeted processor?
Not really considering how specialized hardware can be much more quicker at specialized events. It is unpractical though if you don't have those specialized events. In this case, it will be helpful with games, but not much else for the average consumer.
Inane_Dork
08-Mar-2005, 19:16
You know, if this thing takes off, I wouldn't be too surprised to see animation and vertex shading eventually land on it.
Not until you just mentioned it. :?
"Ageia,
I've run some code from Ageia,
And physics, I've found,
Has a higher upper bound, for me." :lol:
Ragemare
08-Mar-2005, 19:33
so not only the XBox2 will have an PPU. Cool.
How do you know Xbox2 will have a PPU?
I don't, but if you look at a few of DeanoC's comments about the XBox2 I got the really strong impression that the XBox2 will have something like an PPU, but most likely embedded in the CPU.
The Cell's strongest feature is to do the type of calculations that physics processing relys on very quickly. :?
Would it not be more practical to increase general processing power (CPU) instead of adding such narrowly targeted processor?
Not really considering how specialized hardware can be much more quicker at specialized events. It is unpractical though if you don't have those specialized events. In this case, it will be helpful with games, but not much else for the average consumer.
But on the other hand CG industry will go wild because of this. Just think how much physics simulations those guys run nowadays to create stunning effects.
rendezvous
08-Mar-2005, 20:04
Is opengl gonna support this?
Are you serious?
I'll be an early adopter if the price isn't obscene. I think it is cool.
gamepower
08-Mar-2005, 20:09
http://www.ageia.com/pdf/wp_2005_3_physics_gameplay.pdf
I have to wonder where Nvidia and ATI were on this one. You have to think that this would be necessary when the games are becoming CPU limited.
http://www.theinquirer.net/?article=21648
Edit: Here is a cool physics demo....
http://www.ageia.com/novodex_downloads.html#
Physics we have seen on Dreamcast,when you drive in Crazy taxi into
boxes you see the physics when they roll or the ball demo from
Xbox1,look:
http://www.planetxbox.com/games/demos/pingpong/pingpong3sm.jpg
http://www.planetxbox.com/games/demos/pingpong/pingpong1sm.jpg
http://www.planetxbox.com/games/demos/pingpong/pingpong2sm.jpg
I know! This thing is a nVidia plot to destroy ATI! They will implement many fast PS on the GF8800VFF ("Very Fu****g Fast") and use these thingies for VS, since it'll end up being much faster that way and will save costs since two smaller chips are cheaper to make than a big one. And they'll make more money by selling two products. Both SLI-capable, of course. Since they'll have less cooling problems, they'll be able to clock the chip much higher and thus be much faster than ATI with a single-chip solution.
:twisted: :wink:
digitalwanderer
08-Mar-2005, 20:35
Not until you just mentioned it. :?
"Ageia,
I've run some code from Ageia,
And physics, I've found,
Has a higher upper bound, for me."
Now I can't get your ditty out of my head! :lol:
GameCat
08-Mar-2005, 21:27
What makes this a "Physics Processing Unit" as opposed to just a math co-processor in general?
Honestly, as long as you're doing rigid body physics and not soft body or fluid dynamics, the actual physics calculations are cheap. It's the collision detection that tends to kill you. And you can do collision detection on GPUs already, not to mention physics calculations, that's actually pretty simple.
It just seems kind of pointless to have a special unit just for physics when the CPU and GPU can do so much of this stuff already.
digitalwanderer
08-Mar-2005, 21:39
What makes this a "Physics Processing Unit" as opposed to just a math co-processor in general?
'Cause math co-processors went out of vogue with the 486. :roll:
;)
Richard
08-Mar-2005, 21:43
You know, all that Aegis (whatever) PR talk about this being equivalent to 3Dfx's introduction of the VooDoo is not exactly exciting me much. Here're my concerns:
-Like 3Dfx is Aegis going to try and enforce a specific and proprietary API for this (ala GLide?). We need a DirectPhysics and OpenPL I think.
-Is the Novadex (whatever) middleware solution they are "offering" royalty-free? (basically turning PPUs into a commodity and therefor making oodles of cash from their middleware licenses).
-Is there going to be any other companies jumping in and provide competition so the consumers don't have to bend over to Aegis?
-What's the price on the first boards? Because unlike a 3D card which can be used for more than games, a Physics dedicated chip(set) would only be of use for gaming and perhaps civil engineers (talk about niche market).
-Is it passively cooled? I liked that 25W figure but I'm also concerned about noise.
Other than that, looks promising.
Joe DeFuria
08-Mar-2005, 21:53
You know, all that Aegis (whatever) PR talk about this being equivalent to 3Dfx's introduction of the VooDoo is not exactly exciting me much. Here're my concerns:
Doesn't bother me at all. Don't know if it's actually the case...but it doesn't bother me if it is.
-Like 3Dfx is Aegis going to try and enforce a specific and proprietary API for this (ala GLide?). We need a DirectPhysics and OpenPL I think.
Did 3dfx force people to use Glide...or did they have the first mini-gl available and also support Direct3D?
It's expected to have a new API or middleware product concerning they have the first hardware implementation.
Is the Novadex (whatever) middleware solution they are "offering" royalty-free? (basically turning PPUs into a commodity and therefor making oodles of cash from their middleware licenses).
Don't know, but doesn't really matter to me.
-Is there going to be any other companies jumping in and provide competition so the consumers don't have to bend over to Aegis?
If no other companies are taking the risk that Aegis is, then what else should consumers expect?
Richard
08-Mar-2005, 22:48
Did 3dfx force people to use Glide...or did they have the first mini-gl available and also support Direct3D?
What? Is that why I still have to keep my VooDoo 2 whenever I feel like playing EF2000, Diablo 2 with any decent speed, Unreal 1/UT without artefacts, or having to use GLide-wrappers, etc.? Mini-gl is exactly what I don't want to see again: a half-baked api that compromises feature-set to make up for lacking hardware. And the only reason why 3Dfx even did that and provide DX drivers was because of competitors. And then in the event that Aegis goes bust there goes any kind of support for games that rely on them (I can't help but marvel at the irony of Aegis' own comparison with 3Dfx).
It's expected to have a new API or middleware product concerning they have the first hardware implementation.
No argument there. What I'm asking is if they are open (pun intended) to using an open (or at least independent like DirectX) API instead of forcing their own down developers/consumer's throats. Because if that's so and PPUs take off, you can be sure that soon enough there will be new companies with their own hardware and of course, proprietary APIs and then games that support only one of the two APIs.
Joe DeFuria
08-Mar-2005, 22:57
Jesus...3dfx wars redux...
What? Is that why I still have to keep my VooDoo 2 whenever I feel like playing EF2000, Diablo 2 with any decent speed, Unreal 1/UT without artefacts, etc.?
No, it's because you happen to like playing old games. Would you have played any of those games when they came out with the 3D support they had without 3dfx?
Mini-gl is exactly what I don't want to see again: a half-baked api that compromises feature-set to make up for lacking hardware.
No, it was created to get 3D accelerated quake out to the masses of people with 3dfx cards as quickly as possible. Um, Voodoo Graphics was hardly "lacking hardware" for the time. In fact, "revolutionary" usually comes to most people's minds.
And the only reason why 3Dfx even did that and provide DX drivers was because of competitors.
First of all, you're wrong.
The reason why 3dfx did that was because Carmack asked them too. Carmack didn't want to code for any more proprietary APIs after RRedline (Rendition), and he decided to use OpenGL. Creating a full-blown ICD is lots of work, so 3dfx agreed to create "enough" GL support to support Carmack
And, um, what's stopping "competitors" from entering the "Physics Hardware" industry? That's the point. If it becomes a viable industry, there will be more players, and standardization will eventually emerge.
Meanwhile, trying to hammer out "standards" before the first hardware even hits the streed is a bit naieve.
No argument there. What I'm asking is if they are open (pun intended) to using an open (or at least independent like DirectX) API instead of forcing their own down developers/consumer's throats.
And what if said "open" API does not match up with their own hardware?
The reason why 3dfx did that was because Carmack asked them too. Carmack didn't want to code for any more proprietary APIs after RRedline (Rendition), and he decided to use OpenGL. Creating a full-blown ICD is lots of work, so 3dfx agreed to create "enough" GL support to support CarmackWasn't there also something about Carmack/id signing a deal with someone like Matrox, so he had to get 3DFX to do the work rather than doing it himself? I'll see if I can dig up the relevant .plan.
Edit: It seems both Webdog and Blues have taken their .plan systems offline, and it doesn't look like anyone has pre-97 entries from Carmack archived. :(
I'm siding with Joe on this one. I see no Direct Physics, so I envite them to make their own API if they wish. Do you really expect for someone like Microsoft to invest their time and money into coding for something that does not exist???? If and when a standardized API and multiple competitors arrive, then we can start worrying about such currently trivial matters.
Richard
08-Mar-2005, 23:31
No, it's because you happen to like playing old games.
Quake 2, Ultima IX, Half-life, are old games too and I can play them using current hardware/drivers with no problems.
Would you have played any of those games when they came out with the 3D support they had without 3dfx?
Yes because they all had non-3dfx support. It was just never developed enough because we all thought that GLide would be the future.
The reason why 3dfx did that was because Carmack asked them too.
So, it wasn't out of 3Dfx's goodness of their hearts like you imply above? That's my whole point. The only reason they did the little they did non-Glide wise was from external pressure. It seems you're not concerned if there isn't any external pressure on Aegis.
And, um, what's stopping "competitors" from entering the "Physics Hardware" industry? That's the point. If it becomes a viable industry, there will be more players, and standardization will eventually emerge.
And so in 5 years time people won't be able to play UE3 games because they feature a proprietary API that's no longer around?
Meanwhile, trying to hammer out "standards" before the first hardware even hits the streed is a bit naieve.
So the answer to that is resignation in the fact that will take 2 years to see the light and that whatever hardware/games come out in that period will be obsolete before their speed would actually indicate or their gameplay becomes stale?
And what if said "open" API does not match up with their own hardware?
That's probably a good indication whether or not the hardware is good enough (assuming this perfect "open" API of course).
I'm siding with Joe on this one. I see no Direct Physics, so I envite them to make their own API if they wish.
Apparently there's already some work in this. http://www.physicstools.org/ It seems Aegis is contributing to it, which is good news in my book.
Joe DeFuria
08-Mar-2005, 23:43
No, it's because you happen to like playing old games.
Quake 2, Ultima IX, Half-life, are old games too and I can play them using current hardware/drivers with no problems.
And your problem is?
Yes because they all had non-3dfx support.
So then why do you need a 3dfx card then?
It was just never developed enough because we all thought that GLide would be the future.
Who thought Glide would be the future?!
Glide was the easiest and fastest way to port software 3D render engines to support hardware acceleration.
So, it wasn't out of 3Dfx's goodness of their hearts like you imply above? That's my whole point.
Where did I say that?
Do you think it was out of 3dfx "greed" that many developers ported to Glide? Or because the alternatives at the time sucked rocks, and 3dfx had a viable alternative?
The only reason they did the little they did non-Glide wise was from external pressure.
But you need a viable alternative to have external pressure.
If no other competitors step up to the plate, who cares?
It seems you're not concerned if there isn't any external pressure on Aegis.
I'm concerned about external pressure where the market isn't driving it, absolutely.
And so in 5 years time people won't be able to play UE3 games because they feature a proprietary API that's no longer around?
Then play UE5, or keep your old hardware around. Problem?
What if in 1 years time we have no "physics accelerated UE3", because the industry gets bogged down in "standards definition" by companies who don't even have hardware to compete?
So the answer to that is resignation in the fact that will take 2 years to see the light and that whatever hardware/games come out in that period will be obsolete before their speed would actually indicate or their gameplay becomes stale?
My answer is to let the marketplace dictate what the light actually is. Ageia is sticking their neck out and making a very, very, large investment in bringing hardware to market. They deserve to be the market leaders in the software that supports their own hardware.
If the hardware and/or software is CRAP, then it won't get supported.
Dr. Ffreeze
08-Mar-2005, 23:59
All,
This is just SO cool! When I look at the track demo I think of how cool Total Annihilation II would be if all the units had tracks like that! (would ROCK to blow them up!) hahaha I am way more jazzed about the possibility of a good PPU than dual core CPUs, SLI, or any of the next gen video cards. This could have a huge impact of the suspension of disbelief and the funness(tm) of games. =)
This thing is likely more general purpose as a GPU and the only thing which gets to use it is a physics engine? What a waste.
On the contrary I think it is not so bad of idea. Provided that it is widely adapted by industry.
What is? Not exposing the hardware? From some points of view perhaps, from the view of a programmer ... what a waste.
Richard
09-Mar-2005, 01:13
No, it's because you happen to like playing old games.
Quake 2, Ultima IX, Half-life, are old games too and I can play them using current hardware/drivers with no problems.
And your problem is?
It seems we're going in circles. My problem is that Diablo 2 runs like molass in anything except GLide mode, that U1/UT has visual artefacts in anything except Glide mode and that EF2000's non-3Dfx mode is verite's. In essence, my problem is that these (and other games) dedication to a proprietary API means I have to a) run them in software mode b) run them a LOT slower/with visual anomalies c) use a GLide wrapper for more slowdown/artefacts d) keep my VooDoo 2 (which of course only got Beta W2k drivers and I have to use custom made XP drivers). As much as I fancy keeping old hardware for nostalgic reasons (I still have my speccy, NES and Gameboy) I would like to be able to play 5 year old games on my current hardware/software without much fuss like all these other even older games that don't use a proprietary API. That's my problem.
Where did I say that?
You imply that when in reply to my argument that 3Dfx only supported DX and opengl (half) because of competitors. Carmack telling 3Dfx to get their act together was ONLY because there was competitors around.
Do you think it was out of 3dfx "greed" that many developers ported to Glide? Or because the alternatives at the time sucked rocks, and 3dfx had a viable alternative?
You've said it yourself, because 3Dfx had the only viable alternative. I don't dispute that. I contend that reality existed because of 3Dfx's API enforcement and nearly monopolistic tactics. Only with such a behaviour did 3Dfx manage to fend off nVidia (and ATi?) which had superior hardware. They did eventually change (and disappear) but how long did that took?
If no other competitors step up to the plate, who cares?
Do you believe there will be competitors? (not a loaded question).
I'm concerned about external pressure where the market isn't driving it, absolutely.
The consumers belong to that market. If we just stand and take it where's the motivation for Aegis to do the right thing?
Then play UE5, or keep your old hardware around. Problem?
Just because you don't like to play older games (as implied by your lack of concern) and you like having a computer with many, many generations worth of hardware then sure, no problem. But I like playing old games and my rig to be as streamlined and lightwight as possible.
What if in 1 years time we have no "physics accelerated UE3", because the industry gets bogged down in "standards definition" by companies who don't even have hardware to compete?
Aegis (or whoever) doesn't have to wait. That's not what I'm saying. What I'd like is for them to move to those defined standards as quickly as possible and for Tim Sweeny to patch UE3's games that use the current API with the then defined API.
My answer is to let the marketplace dictate what the light actually is. Ageia is sticking their neck out and making a very, very, large investment in bringing hardware to market. They deserve to be the market leaders in the software that supports their own hardware.
If the hardware and/or software is CRAP, then it won't get supported.
No. If Aegis has the monopoly of hardware (which it deserves) AND the middleware (which is apparently not going to) AND the API (which you are a proponent of) then competitors are going to take longer to appear, not make any money if they have to pay royalties/licensing fees to Aegis, have to make potentially inferiour hardware to support a biased API. Delaying the surfacing of potentially better solutions which in turn makes the support of CRAP hardware and/or software a reality.
But anyway, it seems Aegis _IS_ doing the right thing and contributing to this ODF dealie, so some of my questions have been answered.[/u]
I still don't see why this is such a great idea. Physics calculations are all over the gamut, they are not easily parrallized.
Rigid body calculations are the easiest to do in that way, but the second you want to be realistic (eg inelastic collisions) you introduce pathological nonlineararities. I can write a computer program to brute force it, or I can be clever and guess (or copy) the solution for that particular situation. The difference in speed between the two will be many/many orders of magnitude in diffference. Hence there is no general way I can instruct a computer to do this for me and somehow bracket the time frame. Fluid mechanics is notorious for this as well.
So the part seems to me to just end up being general like a CPU, and with probably a lot of software fallbacks where it will be far easier for the developer to just do it themselves.
I mean consider the easiest physical situation in a scene. Lighting! We have graphics cards that have been developed for years by many great minds, but yet not a single one can do it realistically in realtime. We know in principle how to do it, we even know it can be parralelized, but its still prohibitively slow. Every single gfx card on the market uses crude lighting approximations to trick the eye (and the eye is very easily tricked).
Here its not even clear how we can use a set of general approximations to recreate anything that would seem reasonable to our real life situation, they will all have to be specialized per frame by the dev!
IgnorancePersonified
09-Mar-2005, 02:18
as long as it is overclockable I do not really care.
Joe DeFuria
09-Mar-2005, 02:19
It seems we're going in circles. My problem is that Diablo 2 runs like molass in anything except GLide mode, that U1/UT has visual artefacts in anything except Glide mode and that EF2000's non-3Dfx mode is verite's....
So, your problem is with that optional rendering modes?
In essence, my problem is that these (and other games) dedication to a proprietary API means I have to a) run them in software mode b) run them a LOT slower/with visual anomalies c) use a GLide wrapper for more slowdown/artefacts d) keep my VooDoo 2 (which of course only got Beta W2k drivers and I have to use custom made XP drivers).
Contrast this "problem" of yours to the problem of not even having accelerated modes for this software when they were released. Which is worse? You having to hold on to old hardware 5 years later, or deal with occaisional anomolies? UT / UT1 runs under practically every API known to man. ;)
I would like to be able to play 5 year old games on my current hardware/software without much fuss like all these other even older games that don't use a proprietary API. That's my problem.
Then why don't you "fuss" to the game makers to update support for "your API of choice?"
Why don't you tell Epic and the UE3 gamers "hey, forget about accelerated physics in UE3 games...we know it's possible to do, but there is no open standard API...so we'll wait another generation or 2".
A bit selfish?
You've said it yourself, because 3Dfx had the only viable alternative. I don't dispute that. I contend that reality existed because of 3Dfx's API enforcement and nearly monopolistic tactics.
Oh, so um, 3dfx is no longer because of this? The reality existed because there was no viable alternative.
When other hardware equally or comparably capable became available, and as standard APIs matured so they didn't suck as much, the industry moved forward. As it should have.
Only with such a behaviour did 3Dfx manage to fend off nVidia (and ATi?) which had superior hardware.
Huh? Riva128 was not superior 3D hardware. (It was a superior OEM part).
TNT was on par with Voodoo architecture 3D wise. Some plusses, some minues. 3dfx made their contributions to the GL ARB and Direct3D like everyone else driving those standards.
They did eventually change (and disappear) but how long did that took?
As long as it should have. They didn't have the hardware or the OEM presence to compete with nVidia's line-up.
Do you believe there will be competitors? (not a loaded question).
I'm guessing not initially. I'm guessing that potential "competitors" (discrete physics chips) are going to see the waters test with this part. I'm also guessing that the ultimate "competitors" may end up being ATI/Nvidia/Intel/AMD who may be working in similar architectures into future versions of their chips.
The consumers belong to that market. If we just stand and take it where's the motivation for Aegis to do the right thing?
Define "stand and take it." If consumers buy products that Aegis builds or that license Aegis technology...then consumers are voting "yes". If consumers don't buy it, THAT's how they "stand up against it."
Just because you don't like to play older games (as implied by your lack of concern) and you like having a computer with many, many generations worth of hardware then sure, no problem. But I like playing old games and my rig to be as streamlined and lightwight as possible.
Just becayse you don't like to play newer games with the latest technology....
Aegis (or whoever) doesn't have to wait. That's not what I'm saying. What I'd like is for them to move to those defined standards as quickly as possible and for Tim Sweeny to patch UE3's games that use the current API with the then defined API.
I don't understand what you're saying.
Sweeny can patch UE3 with whatever technology he feels fit to do so. If "DirectPhysics" emerges, Sweeney can support that if he wishes. It's not up to Ageia.
[quote]No. If Aegis has the monopoly of hardware (which it deserves) AND the middleware (which is apparently not going to) AND the API (which you are a proponent of) then competitors are going to take longer to appear...[/qupte]
Why?
Did nVidia take longer to "appear" because Glide existed? No.
digitalwanderer
09-Mar-2005, 04:56
Dang, I played thru all the demos with my 5 year old daughter and she was just enchanted with them!
If it can add that to games that smoothly, I don't care if it's a bloody monopoly or not for a while! :lol:
If this catches on there will be others building 'em, guaranteed.
:) I knew this day would come.
http://www.beyond3d.com/forum/viewtopic.php?t=14511&postdays=0&postorder=asc&hig hlight=physics+processing+unit&start=120
Man, I wish there's company that comes up with an idea to create a PPU(Physics Processing Unit). That should leave only the AI to the CPU unless we could make an AIPU.
Mulciber
09-Mar-2005, 05:52
:) I knew this day would come.
http://www.beyond3d.com/forum/viewtopic.php?t=14511&postdays=0&postorder=asc&hig hlight=physics+processing+unit&start=120
Man, I wish there's company that comes up with an idea to create a PPU(Physics Processing Unit). That should leave only the AI to the CPU unless we could make an AIPU.
lol, good job :lol:
ninelven
09-Mar-2005, 06:09
erased
http://www.ageia.com/pdf/wp_2005_3_physics_gameplay.pdf
I have to wonder where Nvidia and ATI were on this one. You have to think that this would be necessary when the games are becoming CPU limited.
http://www.theinquirer.net/?article=21648
Edit: Here is a cool physics demo....
http://www.ageia.com/novodex_downloads.html#
Physics we have seen on Dreamcast,when you drive in Crazy taxi into
boxes you see the physics when they roll or the ball demo from
Xbox1,look:
http://www.planetxbox.com/games/demos/pingpong/pingpong3sm.jpg
http://www.planetxbox.com/games/demos/pingpong/pingpong1sm.jpg
http://www.planetxbox.com/games/demos/pingpong/pingpong2sm.jpg
Neat.
Dang, I played thru all the demos with my 5 year old daughter and she was just enchanted with them!
If it can add that to games that smoothly, I don't care if it's a bloody monopoly or not for a while! :lol:
If this catches on there will be others building 'em, guaranteed.
Did you try the dominos? My system crawls running that one.
What is cool about this technology is that it should also work well with hyper threading and dual cores.
Toms hardware coverage....
.http://www.tomshardware.com/hardnews/20050308_214530.html
Of course there is the question why Nvidia or ATI have not thought about a physics chip so far and if graphics will not be able to handle physics simulation in the not too distant future without the need for an extra chip. Hegde believes that current "graphics chip architecture does not handle physics interaction very well" while Ageia's PhysX chip has "lots of horsepower". He did not provide detailed specs about the calculation capabilities of the chip.
Nvidia was aware of the announcement, is in a wait-and-see position and did not appear to be too concerned about the PhysX chip. "We are approaching half a billion transistors in our GPU. I cannot imagine that we will not be able to do that on our side," said Ned Finkle, vice president of strategic marketing at Nvidia. He considered it "an interesting idea" but likely to be a "niche" solution and not a threat for Nvidia. Finkle also pointed to the hurdles firms such as Ageia face when bringing such a product to the market, including legal issues, marketing and distribution.
Wow, Nvidia should be a bit worried because physics seems to me to be much better investment then SLI. I would imagine however that physics would be much better suited on the video card itself with 512MB memory.
Also interesting is that TSMC is an investor.
Did anyone try creating walls multiple times in a row? It kills performance. When you do it, you'll see why.
You've said it yourself, because 3Dfx had the only viable alternative. I don't dispute that. I contend that reality existed because of 3Dfx's API enforcement and nearly monopolistic tactics. Only with such a behaviour did 3Dfx manage to fend off nVidia (and ATi?) which had superior hardware. They did eventually change (and disappear) but how long did that took?
Nonsense. When Voodoo1 appeared, ATI was making crappy 2D cards and nVidia wasn't even there yet or just started developing nV1 (which sucked a**). Carmack coded GLQuake on 3Dfx because it was the only "real" 3D card back then. Yes, there was Rendition etc. but they were all not even half as fast or feature rich.
Back then, there was no real 3D API available for gaming. That's why glide was so successfull - it was easy to code for and there was a HW which could do a bunch of neat things noone else could.
Now talking about Diablo2 or Unreal, that was all a couple of years later. DirectX was still crap, OGL was only used by id so far and you had a bunch of developers who programmed glide last couple of years. They had the choice to program for glide or to try getting these things work with Directx6 and nVidia Riva128 and ATI's still 2 1/2D RageBlahXXX crap. So logically, they invested their time in coding for the proven, widespread platform which was Glide and haven't wasted their time with programming for those three geeks who whined over not being able to play Unreal on their ATI RageProSuperUltraMaXX0r with 2MB RAM...
Richard
09-Mar-2005, 08:59
Contrast this "problem" of yours to the problem of not even having accelerated modes for this software when they were released. Which is worse?
You continue to think, despite my admonition, that that's the dichotomy I seek.
8)
Huh? Riva128 was not superior 3D hardware. (It was a superior OEM part).
TNT was on par with Voodoo architecture 3D wise. Some plusses, some minues. 3dfx made their contributions to the GL ARB and Direct3D like everyone else driving those standards.
So that's why I sold one of my VooDoo 2s and got a TNT2? It wasn't perhaps because it had 32-bit colour, OpenGL ICD and DirectX support out of the box right? And if what you say is true then why did 3Dfx go under? Wait, you say below "they didn't have the hardware", that's right. What I said.
Define "stand and take it." If consumers buy products that Aegis builds or that license Aegis technology...then consumers are voting "yes". If consumers don't buy it, THAT's how they "stand up against it."
1) You get a lot of like-minded people together and mass mail Aegis with your "manifesto". Hello lobby pressure.
2) In mboards, such as this, instead of praising whatever new comes regardless of consequences, and instead of resignation of what's to come, put your concerns forward now.
3) If all else fails, don't buy something that will ultimately be a bad thing.
But I guess Digi's post exemplifies most people today: hey, that looks cool, I want it, I want it now, me, me, me, now, now, now, what do I care that 5 years down the road I'll be paying $500+ for each part and get shitty availability?
Here's another example: even though I'm very happy with my R9800 Pro a while ago I posted that Valve was doing the wrong thing and setting a bad precedent for not allowing GFFXs to use mixed precision and dropping them to Dx8. "Oh screw GFFX owners, they should have known better", "Valve probably saw their GFFX market share and decided not to bother", "it's their game, they decide". What do we see now? We see games (like Chaos Theory and STALKER) that could run perfectly alright in Dx9 mode in R3xx/R4xx cards but that will have to run in Dx8 mode. I hate to say "I told you so" as I didn't even expect the comeback to arrive so soon. I guess "R4xx buyers should have known better" and "it's their game, they decide".
Just becayse you don't like to play newer games with the latest technology....
How does liking to play older games comes in contradiction to liking to play newer games?
Sweeny can patch UE3 with whatever technology he feels fit to do so. If "DirectPhysics" emerges, Sweeney can support that if he wishes. It's not up to Ageia.
But it is up to Ageia to move quickly in an open direction instead of like 3Dfx which did their best to delay that. Here's a little quote from Carmack July 3 1997:
Minidrivers certainly work fine -- we have functional ones for 3dfx and powerVR, and they have the possibility of providing slightly better performance than fully compliant drivers, but partial implementations are going to cause problems in the future.
Wasn't that the truth?
If they don't, then like _xxx_ said the "logical" (yeah, right) step is to continue to support the available proprietary API.
Did nVidia take longer to "appear" because Glide existed? No.
It did, and then it was having problems getting through to consumers because even in 2000 there were games (like Diablo 2) with GLide mode that performed better, at lower quality settings than <insert competitor here> that had to run in OGL/DX.
Can people return to the actual topic of this thread and not recovering age-old arguments, please? Otherwise, the thread will just get some serious pruning.
DegustatoR
09-Mar-2005, 09:58
Whoa! A Cell APU for the PCI Express :lol:
An interesting idea, but will probably hit the wall as soon as Intel and AMD decide to do some research in this area ;)
Whoa! A Cell APU for the PCI Express :lol:
An interesting idea, but will probably hit the wall as soon as Intel and AMD decide to do some research in this area ;)
Good thing it isn't a cell apu ;)
digitalwanderer
09-Mar-2005, 13:18
Did you try the dominos? My system crawls running that one.
Mine ("Bubbles: Athlon XP-M 2400+ (12x200), NF7-s rev 2, 1Gb PC3200, OEM Sapphire ATi Radeon 256MB X800 pro VIVO (flashed 16 pipes 500/540), Audigy2, Altec-Lansing ADA885, ATi TV-wonder VE, DVDR, 2x10/100NIC, 80GB, 200GB sATA") ran it like butter, the only one that really slowed mine down was the exploding builiding one so far.
My daughter loved whipping the ragdolls around and smashing the block towers to pieces and such. :lol:
Richard
09-Mar-2005, 17:34
Can people return to the actual topic of this thread and not recovering age-old arguments, please? Otherwise, the thread will just get some serious pruning.
Ok, OnT, I'm surprised by the amount of local memory these parts are supposed to have, isn't 128mb (GDR3 at that) a little too excessive? Anyone with hands-on knowledge of physics implementations wants to chime in?
I can understand the concept of caching especially on PCI parts but when I read that interview with more info (I missed the first page) I expected a smaller and faster cache particularly due to PCI-e's full duplex capability.
There's also some who believe a PPU would be of better use combined with video cards. I can certainly see that line of reasoning considering that a PPU is only going to be used for games (at least initially) and that if you're looking to offload some work from the CPU then that probably means you have a capable 3D card that's waiting for the CPU to begin with.
If this is done, is there a potential problem of saturating the PCI-e bus? Not speed wise, but where latency is concerned?
Dr. Ffreeze
09-Mar-2005, 19:09
Cool news,
Hegde envisions the chip to become available in high-end, mid-level and entry-level versions ...
...He hesitated to comment on pricing that should be expected for add-in cards, but mentioned that a range from about $100 to $400 would be realistic....
http://www.tomshardware.com/hardnews/20050308_214530.html
Chalnoth
09-Mar-2005, 19:32
A dedicated physics coprocessor makes no sense at all for the PC market. If it has a use anywhere, it would be in the professional simulations market.
That is to say, there's just no way to smoothly make the transition from software physics to hardware-acclerated physics on the PC, not like there was with 3D graphics.
digitalwanderer
09-Mar-2005, 19:37
But I guess Digi's post exemplifies most people today: hey, that looks cool, I want it, I want it now, me, me, me, now, now, now, what do I care that 5 years down the road I'll be paying $500+ for each part and get shitty availability?
You worry a bit too much, 5 years is a long time and the Digi ain't no fool. :roll:
digitalwanderer
09-Mar-2005, 19:40
That is to say, there's just no way to smoothly make the transition from software physics to hardware-acclerated physics on the PC, not like there was with 3D graphics.
Why not? We've had the option to run graphics in hardware or software mode for a while, why not the same with physics? :|
A dedicated physics coprocessor makes no sense at all for the PC market. If it has a use anywhere, it would be in the professional simulations market.
That is to say, there's just no way to smoothly make the transition from software physics to hardware-acclerated physics on the PC, not like there was with 3D graphics.
You obviously didn't read any of the documentation.
The developers use an API. The API is multi-threaded. You can run on one CPU or more and if you have a PPU it will use that as well as/or instead of the CPU.
I think the physics chip should be on the video card itself however and share memory with the video chip.
Richard
09-Mar-2005, 19:50
Cool news,
...He hesitated to comment on pricing that should be expected for add-in cards, but mentioned that a range from about $100 to $400 would be realistic....
Cool news indeed, that $400 figure just chilled me to the bone.
Try the Building Explode or Big Bang demo and you'll see why you need dedicated chip for physics.
As side note: Big Bang demo totally killed my AthlonXP 3400+ (Tbred)
Building Explode ran smooth for like 5 seconds when there is not much physics used. After that it just goes down.
Cool news,
...He hesitated to comment on pricing that should be expected for add-in cards, but mentioned that a range from about $100 to $400 would be realistic....
Cool news indeed, that $400 figure just chilled me to the bone.
Not everyone will buy the $400 dollar card. I'd settle for the $100-$200 level card. The same works with graphics cards. A X850 XT PE is over priced, but people purchase it. Same goes for the 6800 Ultra or two 6800 Utras in SLI for $1000! It depends on the person. If you look for reasons to not want it you will find them, and that can be said for ANY PRODUCT!
The building explode entrhalled me. I watched it a few times. I'm not crazy though, I wouldn't want to see that stuff in real life.
digitalwanderer
09-Mar-2005, 20:04
If I could pick up a $100-200us priced card that could add super-cool physics to my games I think I'd do it in a heartbeat. :)
There's a cute bug in the Golem ragdoll.
If you grab it by the top of a foot (maybe other places...?) and swing it round a bit, you'll quickly find there's, ahem, perpetual motion...
Right-click to grab.
Jawed
Richard
09-Mar-2005, 20:11
Not everyone will buy the $400 dollar card. I'd settle for the $100-$200 level card. The same works with graphics cards. A X850 XT PE is over priced, but people purchase it. Same goes for the 6800 Ultra or two 6800 Utras in SLI for $1000! It depends on the person. If you look for reasons to not want it you will find them, and that can be said for ANY PRODUCT!
Take it easy, I think you're reading way too much into my one liner. :roll: I was taken aback by this figure. Doesn't mean I'm passing judgement on whether or not it's worth it. For one, I'd like to see which games will make use of it and what performance/feature difference the PPU will provide.
Secondly, the video card analogy is not that good because without one you can't play any games; without a PPU you'll still be able to play them (at least until/if PPUs become a requirement like 3D acceleration today) so I need to know if such an investment is sound by the conditions of the market. That's what I always try to do: don't reject something out of hand nor accept it blindly.
So, there was no innuendo in my previous post so please don't "go look for reasons" there yourself. :wink:
Sorry, no harm meant. 8)
I would want to see reviews for it before I bought it of course, but I'm excited for what it COULD do.
Also try Physics submenu demo (access it by F4 button).
When you're there press "M" key for like 10 seconds (or for as long as your CPU allows).
I also recommend to first create Castle environment (Create Castle) for even better collision presentation.
Check this out :)
http://img185.exs.cx/img185/4639/massivephysics8ft.th.jpg (http://img185.exs.cx/my.php?loc=img185&image=massivephysics8ft.jpg)
http://img185.exs.cx/img185/7757/massivephysics24xb.th.png (http://img185.exs.cx/my.php?loc=img185&image=massivephysics24xb.png)
Dr. Ffreeze
09-Mar-2005, 20:41
PPU Concerns,
Three levels of cards, entry, mid, and high end. I would guess that they are discussing the specs of the high end model so a very from the hip conservative guess might make the cards look like this:
High End: $400 400MHz 125M Transistors 40,000 entities
Mid: $200 200MHz 125M Transistors 20,000 entities
Entry: $100 100MHz 125M Transistors 10,000 entities
CPU: $420 3400MHz 169M Transistors 40 entities
How did I come up with these numbers, well I mostly guessed but here is why I guessed the way I did.
High End:
- $400: they said so at tomshardware
- 125M transistors: they said so
- 40,000 entities: I think they would be doing PR with their best part. and they were using 40,000 entities in their PR blitz.
- 400MHz: Guess. This is between NVIDIA's (400MHz) and ATI's (520MHz) high end GPU speed. As TSMC is an investor it would be in there best interest to get this working well.
Mid:
- $200: Guess. Good midrange price point (for video cards anyway)
- 125M transistors: Guess. I would guess it would make the most since for them to speed bin their parts instead of having different cores. This way they could supply the few fast parts to the high end while the bulk would be for the mid and entry cards.
- 20,000 entities: Guess. Worst case scenario would be ~1/2 the speed of the high end part. While it is much lower than their PR it still kicks a high end CPU's butt (if their PR is to be believed).
- 200MHz: Guess. If their high end parts could hit 400MHz, then the yields would be much better at 1/2 speed, plus it makes the numbers nice. =)
Entry:
- $100: They said so in toms hardware that ~$100 could be the low end range of prices.
- 125M transistors: Guess. I would guess it would make the most since for them to speed bin their parts instead of having different cores. This way they could supply the few fast parts to the high end while the many would quall for the mid and they could salvage the worst performers for the entry level.
- 10,000 entities: Guess. Worst case scenario would be ~1/2 the speed of the mid ranged part. While it is much, much lower than their PR it still kicks a high end CPU's butt (if their PR is to be believed).
- 100MHz: Guess. If their high end parts could hit 400MHz, then the yields would be much better at 1/4 speed, plus it makes the numbers nice. =)
CPU:
- $420: Pentium 4 650 at www.newegg.com
- 169M transistors: so says techreport.com
- 40 entities: so says Ageia's PR. Seems a bit low but I guess it depends on the game. I would guess HL2 would have more than 40 entities you could interact with.
- 3,400MHz: So says Intel
Ok, bases on those seat of my pants numbers I have few questions/thoughts.
Take 1: Lets say the Unreal 3 Warfare engine (which will support the new PPU) will only use the add PPU to increase eye candy. Lets Epic intros a new flame thrower that is plain amazing. The high end card renders it with 40,000 entities, 20,000 for the mid, 10,000 for the entry, and the lowly CPU gets the age old plane texture flame thrower. Well, regardless how your level of physics complexity all four types of players could play online together well. The problem I see here is that it is does not make a very good case to buy the $400 Ultra PPU.
Take 2: Epic decides to make a game that shows off how utterly cool it would be to have 1000 times the entities you can interact with and this actually changes how you play the game, how you solve puzzles. This might allow you to make the assault from the back door, when there was never a door in the back to begin with. You just blow your own door. You could instead make it a side assault or even one from the roof. This would be a sweet reason to buy the PPU Ultra card but how is someone with a lowly 3.4GHz CPU going to be able to play the game?
If we don't see some very compelling reasons to upgrade, more reasons that just eye candy, then I believe Ageia will have a very hard time succeeding. The problem is, I don't see how game programmers will be able to balance the game experience between those that have a PPU and those that do not if they use the PPU power for more than just eye candy.
Thoughts?
Dr. Ffreeze
09-Mar-2005, 20:46
Chalnoth,
A dedicated physics coprocessor makes no sense at all for the PC market. If it has a use anywhere, it would be in the professional simulations market.
That is to say, there's just no way to smoothly make the transition from software physics to hardware-acclerated physics on the PC, not like there was with 3D graphics.
I could would, could you at least read the last bit of my last post? I believe you are talking about what I am also concerned with. Unless it is only eye-candy how will they balance between the PPU haves and have nots?
The implications are pretty clear here. You have either 30fps (or less) without PPU or 100fps (or more) with one.
As games get more complex, here's a way to scale up gaming performance, independently of screen resolution, or how many cores your CPU has. Dual-core, when it comes is eventually going to be an approximate doubling of gaming performance in CPU-hogging physics (well, less, due to other aspects of a game such as AI).
This is a great way to un-tie another dimension of game programming from the PC's CPU-performance brickwall.
Games aren't just graphics. There's also no reason why gaming consoles in 3-5 years' time wouldn't use such a PPU component in their architecture.
Jawed
Richard
09-Mar-2005, 21:09
PPU Concerns,
Three levels of cards, entry, mid, and high end. I would guess that they are discussing the specs of the high end model so a very from the hip conservative guess might make the cards look like this:<snip>
I think your core speeds are too apart. I can certainly see the low-end having only 64mb onboard ram or whatever and have nearly the same core speed as that mid-level part. Also, since Aegis will only supplies the chipsets themselves we could potentially see a lot of different designs from all the AIB partners. Which reminds me, what makes you guess that the low-end parts will have the exact same chip (125m tranny count)? It makes sense for the startup but I think they'll want to ramp out different designs for the different price tiers as soon as they can.
Since they talked about mobo integrated PPUs I'm positive they have different designs already.
Dr. Ffreeze
09-Mar-2005, 21:21
Mordenkainen,
I think your core speeds are too apart.
I agree, I just wanted to go with the "worst case scenario". It could be argued that the MHz doesn't matter really, just the end result in performance. They could fill the entry level boards with a faster part but with just less memory bandwidth and/or amount like you said.
Which reminds me, what makes you guess that the low-end parts will have the exact same chip (125m tranny count)?
The only thing would be the ease of pulling this off at the expense of a more costly entry level chip. They could also have a 50m tranny count chip for the entry level and save money on each chip that went out the door, it would just be more complicated to design 2 different chips for a startup (which they could do, my idea was just a guess).
Since they talked about mobo integrated PPUs I'm positive they have different designs already.
That is a very good point that I had not thought of.
I also find it interesting if they could release a $50 board. If they could do this and it had a very low performance compared to their other offerings but still had a 2 orders of magnitude greater performance than CPUs, they would be able to move many many chips. At $50 it is the same price as a quality new release game. I don't think they would be able to do it, but it would be nice for their install base.
Chalnoth
09-Mar-2005, 22:20
Why not? We've had the option to run graphics in hardware or software mode for a while, why not the same with physics? :|
Because with graphics you could change quality levels to coincide with your hardware. With physics there's just no good way to change quality smoothly in the same way.
Chalnoth
09-Mar-2005, 22:23
You obviously didn't read any of the documentation.
The developers use an API. The API is multi-threaded. You can run on one CPU or more and if you have a PPU it will use that as well as/or instead of the CPU.
I think the physics chip should be on the video card itself however and share memory with the video chip.
Has nothing to do with the API. The API may be a good thing (as it could possibly lead to better usage of current processors, CPU or GPU, for physics), but dedicated hardware just doesn't make any sense for the PC. The reason is simple: physics is either on or it's not (for the most part). There's no good way to have different quality levels. Therefore developers must program for the lowest common denominator, which will always mean PC's without physics accelerating hardware.
The implications are pretty clear here. You have either 30fps (or less) without PPU or 100fps (or more) with one.
Thats assuming physics are a bottle-neck, we find that in a complex scene physics (Havok 2.3) took up 5-10% CPU time, thats everything collision, response, ragdolls and constraints.
So even if the PPU can remove that completely, we will have gone from 30fps to 33fps...
To do more physics, we have a little problem, the problem is that we have to render those individually physic'ed objects, so you also have to speed up rendering. So if you want better physics, first you need to reduce the cost of rendering again (ironic how all the its not all rendering statements end up meaning its all about the rendering ;-) )...
Makes no difference if there system can do 40,000 objects per frame, we could never render them... Also its worth noting most good physics engines can do 100s if not 1000s right now, its the number of interacting (colliding) entities that slow things do (20 interacting ragdolls (body heap) of ~20 bones each is about the normal CPU limit).
Rock meet hard place...
I can see the PC maybe needing a physics chip (if we can't get the GPU taking some of the burden, I suspect they will though...), as your only getting 2 cores near term. Don't see the need on the consoles (except maybe Revolution... god knows what thats doing). Its not like we are short of processing power (using it effectively is a harder problem of course) on them.
And while the publicity has been on Aegis/Novodex and Meqon this GDC, I just want to say how good the new Havok 3 is. We are currently doing a quick intergration with it and continous collisions rock. Sure they won't mind me saying that ;-)
Other interesting thing, is if this PPU is doing discrete or continous collision. I hate discrete collisions system with a passion but most physics engines use them (a lot harder to generate contact points)
Sorry rambling a bit... Just a general musing rather than any well thought out comments.
In Novodex Rocket, with the Building Explode demo, running on A64 3500+ with 9800Pro 128MB, 256-bit (6000-ish 3DMk03, "unoptimised") I get the following frame rates in 1024x768 windowed mode, at about 10 seconds in:
- 12fps with the physics simulation running
- 100fps with the physics simulation paused
- 50fps with the physics simulation paused, 4xAA
Clearly my graphics card is running well below capacity simply because the CPU is struggling with the physics.
This is an extreme example, of course (4200 bricks - number of bricks varies, not sure why...). But the CPU isn't doing much else...
How game developers might use, say, 1-2000 "fragments" of a solid object being blown apart, or 1-200 bodies (say) being thrown around I don't know...
Jawed
Chalnoth
10-Mar-2005, 00:51
It is always possible, of course, that that demo's physics engine isn't as efficient as the one from Havok.
Your kind of missing Deano's point.
Outside of trivial demos there are constraints besides "time it takes to compute physics" in a real game. On a PC in particular drawing a lot of objects is an extremly expensive operation, 40,000 objects means probably >40,000 calls to DrawPrimitive, even with a NULL driver this is extremly expensive.
If your goal is enabling more moving objects, you have to solve a lot more than just the physics cost.
Killer-Kris
10-Mar-2005, 00:58
Why not? We've had the option to run graphics in hardware or software mode for a while, why not the same with physics? :|
Because with graphics you could change quality levels to coincide with your hardware. With physics there's just no good way to change quality smoothly in the same way.
That depends on what you're doing with your physics. If you're doing inverse-kinematics for animations that interact with the environment (stairs, uneven ground, etc), you just use a lower resolution bone structure for the lower performance PPUs, or disable the inverse-kinematics for CPU users.
Same goes for any fluid simulations (bodies of water, clouds, etc...), you use a lower resolution of points on the lower end part, and just have a static fluid surfaces for CPU users.
And I'm sure that there are more creative solutions to scale physics performance than what I've listed. Of course there may not be any sort of good solution to the deformable terrain/buildings to handle the CPU users. So until PPUs become ubiquitous they probably will only serve to make a prettier and more immersive environments. But I imagine that many of these effects will have the same effect on people’s opinions, that bilinear filtering had when it was new. So with any degree of luck the user base will build quickly.
The implications are pretty clear here. You have either 30fps (or less) without PPU or 100fps (or more) with one.
Thats assuming physics are a bottle-neck, we find that in a complex scene physics (Havok 2.3) took up 5-10% CPU time, thats everything collision, response, ragdolls and constraints.
So even if the PPU can remove that completely, we will have gone from 30fps to 33fps...
To do more physics, we have a little problem, the problem is that we have to render those individually physic'ed objects, so you also have to speed up rendering. So if you want better physics, first you need to reduce the cost of rendering again (ironic how all the its not all rendering statements end up meaning its all about the rendering ;-) )...
Makes no difference if there system can do 40,000 objects per frame, we could never render them... Also its worth noting most good physics engines can do 100s if not 1000s right now, its the number of interacting (colliding) entities that slow things do (20 interacting ragdolls (body heap) of ~20 bones each is about the normal CPU limit).
Rock meet hard place...
I can see the PC maybe needing a physics chip (if we can't get the GPU taking some of the burden, I suspect they will though...), as your only getting 2 cores near term. Don't see the need on the consoles (except maybe Revolution... god knows what thats doing). Its not like we are short of processing power (using it effectively is a harder problem of course) on them.
And while the publicity has been on Aegis/Novodex and Meqon this GDC, I just want to say how good the new Havok 3 is. We are currently doing a quick intergration with it and continous collisions rock. Sure they won't mind me saying that ;-)
Other interesting thing, is if this PPU is doing discrete or continous collision. I hate discrete collisions system with a passion but most physics engines use them (a lot harder to generate contact points)
Sorry rambling a bit... Just a general musing rather than any well thought out comments.
That is a good point. Not only that but you must ship 40,000 objects from the physics card to the video card for each frame. At some point the CPU has to manipulate these objects as well.
It is always possible, of course, that that demo's physics engine isn't as efficient as the one from Havok.
Who knows the answer to that?
However this API should rock if you have hyper threading or dual cpus.
Chalnoth
10-Mar-2005, 03:11
However this API should rock if you have hyper threading or dual cpus.
Then again, an optimized physics engine could provide those same things.
Sorry i still dont buy it.
You would have to convince me those 120 m transistors are best spent on a physics ppu, above say, just making the cpu or dual cpu 120m transistors larger and having the dev spend some time on coding. B/c frankly the latter seems to me to be the better solution.
No,it's not the same. CPU can never be as effective as dedicated hardware.
I never coded physics so i don't know how dedicated hardware affects performance,but for example CPUs can never be as fast as 3D graphic card in 3D graphics.
It's simply impossible. The same can be for physics. If the chip is created for a very specific thing with lots of optimisations for right things it can make a big difference in performance.
I think the rendering of 40000 objects is not where it'll be used, rather stuff like deformable terrain where the load on rendering isn't that high.
nutball
10-Mar-2005, 11:13
I wonder if this thing will be programmable "to the metal"? We use a lot of particle-based techniques in astrophysics. One (or more) of these things in each node in a large cluster could make a very interesting vector supercomputer on-the-cheap.
Fudo snapped a piccy
http://www.theinquirer.net/images/articles/ppu.jpg
Your kind of missing Deano's point.
Outside of trivial demos there are constraints besides "time it takes to compute physics" in a real game. On a PC in particular drawing a lot of objects is an extremly expensive operation, 40,000 objects means probably >40,000 calls to DrawPrimitive, even with a NULL driver this is extremly expensive.
If your goal is enabling more moving objects, you have to solve a lot more than just the physics cost.
DeanoC has decided to go off on some extreme tangent talking about 40,000 objects. I'm saying that only 1-2000 objects rendered with real time physics is enough to bring a current CPU to its knees. Those 2000 objects (without physics), will run acceptably on a 9800Pro, now. 4000+ objects, with the physics turned off, runs at 100fps.
The demo shows that the physics kill a PC's ability to render a fairly small number of objects.
Sure it's possible to theorise that this physics engine is optimised for a hardware implementation, and is therefore sub-optimal when computed solely by software. Somebody will have to start benchmarking physics engines :D .
I still think it's something the gaming industry will take seriously. But it might turn out to be another five years before we get there.
We might find that the "spare" CPU power in the next gen consoles is used to model physics with these levels of "accuracy". And it is those levels of intricate physics in console games that finally drives acceptance of a PPU for PCs in the PC gaming market. Simply to keep up with what consoles can do.
Who knows, the next phase of evolution might be a 3- or 6-core PowerPC add-in card for PCs, because Intel CPUs are so useless at core-scaling.
Another way of looking at this is that the CPU power of the next gen consoles is so far ahead of PCs, rendering physics and AI plus graphics much better than a 2005 PC, that game developers lose all interest in the PC platform.
PPU might be the last chance for the PC platform to stay in the race with consoles. In my view, if it is sold primarily as an expensive add-in card ($200+) then it's doomed. It needs to be a $40 option on a motherboard, or a $50 add-in card.
Jawed
digitalwanderer
10-Mar-2005, 12:51
And while the publicity has been on Aegis/Novodex and Meqon this GDC, I just want to say how good the new Havok 3 is. We are currently doing a quick intergration with it and continous collisions rock. Sure they won't mind me saying that ;-)
That's something I've been wondering about, what will this do to Havok? They seem to have the largest share of the independant physics engines by far, will this new technology usurp 'em or incorporate 'em?
Mate Kovacs
10-Mar-2005, 12:58
Who knows, the next phase of evolution might be a 3- or 6-core PowerPC add-in card for PCs, because Intel CPUs are so useless at core-scaling.
Did you mean, x86? :)
I'm saying that only 1-2000 objects rendered with real time physics is enough to bring a current CPU to its knees.
In a real game each of those object have a bunch of other code that will bottleneck long before the physics, think renderer state changes, game code etc.
As I mentioned a PPU would mean a maximum of 10% gain in our game currently... And that was in a delibrate physics test environment with much more real physics objects than most games...
I don't deny it will make things faster just expect 5fps not 50fps, unless this things becomes standard and then you might see it being stressed but until that time this chip will be idling most of the time waiting for the CPU and GPU to do there work...
That's something I've been wondering about, what will this do to Havok? They seem to have the largest share of the independant physics engines by far, will this new technology usurp 'em or incorporate 'em?
Havok have been concentrating on characters and improving the quality of the simulation (things like continous collision and IK systems). How it affects them in the market place I don't know, but IMO its the right choice, who cares about a 1000 bricks if your characters looks craps (I suspose for FPS it might make sense...)
Simon F
10-Mar-2005, 13:25
As I mentioned a PPU would mean a maximum of 10% gain in our game currently... And that was in a delibrate physics test environment with much more real physics objects than most games...
Out of curiousity, how much of that is just checking for collisions?
Simon F
10-Mar-2005, 13:31
Havok have been concentrating on characters and improving the quality of the simulation (things like continous collision and IK systems). How it affects them in the market place I don't know, but IMO its the right choice, who cares about a 1000 bricks if your characters looks craps (I suspose for FPS it might make sense...)
Well, that rather depends, doesn't it, on your game's characters? (http://www.gamespot.com/ps2/action/legostarwars/screenindex.html) :-)
digitalwanderer
10-Mar-2005, 13:37
Thanks Deano. I like the Havok engine, it's gave me a whole lot of gaming happy and I didn't want to see it go anywhere soon. :)
Just one thing...
I supose for FPS it might make sense...
You mean they're STILL making other types of games?!?! :|
I'm saying that only 1-2000 objects rendered with real time physics is enough to bring a current CPU to its knees.
In a real game each of those object have a bunch of other code that will bottleneck long before the physics, think renderer state changes, game code etc.
Really? A destructible environment has a bunch of other code bottlenecking it? Why? What's this code doing?
Are the fragments of the environment all running AI code because they're all planning their escape before destruction? :twisted:
I'm still intrigued by the question of what games developers for next-generation consoles are going to do with all that spare CPU power.
SC:CT, which we can be fairly sure is an Xbox 360 release title, obviously needs a serious amount of graphics rendering power - but what are the CPUs going to do?
Jawed
SC:CT, which we can be fairly sure is an Xbox 360 release title, obviously needs a serious amount of graphics rendering power - but what are the CPUs going to do?
Jawed
Fix the totally borked controls and camera movement? :twisted:
I don't think this PPU will be successful. It's a nice idea, but as long as it's not in every box, we'll hardly see any games making full use of it. If they had made it into a next-gen console... but as it stands, this is rather a nice add-on like EAX HD, and as such most gamers won't pay much more for it than for a capable but cheap sound card.
Dr. Ffreeze
10-Mar-2005, 15:51
Crazy Thought,
With PCIe think about if you had a Semperon 2100+ ($50 CPU) PCIe card. What could you do with the extra CPU power? In the dawn of massive multi CPU cores, we should see massive multi threads. I understand that a Semperon 2100+ on PCIe won't touch the speed of a Semperon 2100+ on in a CPU socket but still....
Might be interesting
"I never coded physics so i don't know how dedicated hardware affects performance,but for example CPUs can never be as fast as 3D graphic card in 3D graphics. "
3d graphics cards are a good idea, b/c for instance the transform stage involves a lot of linear algebra. This is very easily parrelized, and thus a different architecture than a cpu is desirable and many orders of magnitude faster, even if it means you have to deal with using an external bus.
Physics is not like that at all, outside certain trivial operations like rigid body mechanics. In the general situation, a parralel architecture might actually bottleneck you where there was no bottleneck before.
And if this is indeed a general solution, it will just begin to look more and more like a cpu architecture, in which case it begs the question (why have it external)
So again, 120m transistors in a gaming situation... Please convince me that those are not best and most efficiently used elsewhere..
I'm saying that only 1-2000 objects rendered with real time physics is enough to bring a current CPU to its knees.
In a real game each of those object have a bunch of other code that will bottleneck long before the physics, think renderer state changes, game code etc.
Really? A destructible environment has a bunch of other code bottlenecking it? Why? What's this code doing?
Are the fragments of the environment all running AI code because they're all planning their escape before destruction? :twisted:
I'm still intrigued by the question of what games developers for next-generation consoles are going to do with all that spare CPU power.
SC:CT, which we can be fairly sure is an Xbox 360 release title, obviously needs a serious amount of graphics rendering power - but what are the CPUs going to do?
Jawed
Most game code like most other application code for that matter just shuffles around a lot of memory. It's all just book keeping.
A lot of PC games do a LOT of work to minimise what they send to the graphics card, this is generally less true of the better performing console titles.
I think that even if a physics processor is succesful in theshort term, all it would really accomplish is having AMD and INTEL focus on better math performance from their processors.
Personally I'm not too worried about lots of falling bricks, but I'd like a lot of horsepower to focus on animation posing, which is a combination of hard constraints, Inverse Kinematics and what I term "soft constraints". I don't know of a physics API that supports the latter. One of the problems with a hardware solution, is that if it doesn't provide what you want your at the mercy of the vendor.
Unlike graphics, physics is something the CPU could easilly usurp, if it was a priority for the vendors.
As I mentioned a PPU would mean a maximum of 10% gain in our game currently... And that was in a delibrate physics test environment with much more real physics objects than most games...
Out of curiousity, how much of that is just checking for collisions?
I've never benchmarked Havok 3, but Havok is pretty much built around the collision system. It's what Havok does well, so if you have a reasonably large number of constraints or interactions the collision time is usually minimal.
Really? A destructible environment has a bunch of other code bottlenecking it? Why? What's this code doing?
Are the fragments of the environment all running AI code because they're all planning their escape before destruction? :twisted:
No but the entity system has to know about them, for each simple dynamics object a few non physics things include component updates (most entities will have some non-physics properties associated with them), frustum cull, shadow frustum cull, proximity and visibility hierachy updates.
If these dynamics things affect the game, they obviously slow down lots of other systems as well or do you want a wall collapsing on a player just to bounce off without the player reacting?
No but the entity system has to know about them, for each simple dynamics object a few non physics things include component updates (most entities will have some non-physics properties associated with them), frustum cull, shadow frustum cull, proximity and visibility hierachy updates.
OK, let me try a different tack. Are you saying that the number of objects in a typical scene is already so high, that adding the graphical load of 1000 or 2000 fragments in an explosion (forgetting the physics completely for a second) would bring the game to a grinding halt? And so, the overhead of the fragments is already so high that the physics workload will make a negligible difference.
In Novodex, Big Bang, with 16000 bricks, runs at 75fps (physics off). That's against 4000 bricks running at 100fps in Building Explode. Physics on, it's about 4fps, versus 12fps for Building Explode.
Jawed
OK, let me try a different tack. Are you saying that the number of objects in a typical scene is already so high, that adding the graphical load of 1000 or 2000 fragments in an explosion (forgetting the physics completely for a second) would bring the game to a grinding halt? And so, the overhead of the fragments is already so high that the physics workload will make a negligible difference.
If by fragment you mean 'real' objects not particles or instanced geometry then yes rendering 2000 will kill performance dead.
What I'm saying is that for real objects, that interact with the player, that have non-optimal rendering characestics, a few thousands won't be just a problem for the physics.
If you just want a fancy particle system just use the GPU, it can render thousands of the same mesh that collide, bounce and react realistically...
Maybe I'm thinking that by physics you mean stuff that actually affects the game rather than just look pretty?
Joe DeFuria
10-Mar-2005, 22:26
Maybe I'm thinking that by physics you mean stuff that actually affects the game rather than just look pretty?
I think this is basically the main issue wrt having a physics co-processor as a viable on PCs.
I think there is some market for a $50-$100 add-in card that in reality, just makes it look "more pretty". Smoother animation, "better explosions", etc. heck, I'd throw down that amount of money for that.
I see it as analogous to sound cards. At worst, you might increase your performance...at best, some better eye candy.
No one "needs" anything better than your typical on-board codec for gaming sound. But add-in cards make it more "pretty" and/or lowers CPU utilization for higher frame rates.....and there is obviously a market for it.
Where I don't see a physics-add in card making any inroads to the PC space, is any claims that it would "change gameplay." It's just not going to happen...gameplay is always coded around the lowest common denominator.
No one makes a game "requiring" 3D positional audio, or EAX, or with sound such that CPU utilization slows the game to a crawl.
Chalnoth
11-Mar-2005, 01:25
Exactly, Joe, and I just don't see many places where dedicated hardware for physics will make a significant difference in effects that can be turned on with it.
After all, you really need to have a, "Wow, I have to have that!" application for your dedicated hardware if you want to get people to even start buying it. This was there back in the early days of 3D graphics. There was simply a night and day difference between GLQuake, for instance, and normal Quake. I just don't see any similar "Wow!" for dedicated physics hardware to start convincing people that it's a worthwhile contribution.
Joe DeFuria
11-Mar-2005, 01:32
Well, that's not exactly what I said. ;)
Even if there isn't a "huge wow" factor...as long as there is some tangible difference, I think there's a market for it as long as it's sub $100.
Again, there isn't a really huge "wow" factor between on board audio and a Audigy2 zs...but there is enough of a difference that $50-$100 retail cards are sold in a viable market.
The challenges for Agia as I see it:
1) The "PPU" seems large and relatively expensive...and also requires memory. Is the product economically viable for about $100 retail?
2) Marekting: it's a lot harder to market "physics" than it is graphics. Graphics you can do with a screen shot....physics is a lot harder. Again, its similar to sound....
However, if Ageia gets enough developer support / endorsement..."You really should play this game with the hardware...", then I can see a modest market being carved out for this stuff.
Again though, I don't ever see it being a "Requirement" for any PC game.
Well i think they should put out a powerfull version for 100$ and then mabye a cheaper one for 50$ and a better one at 200$
I don't think anyone wil lbe willing to spend more than 200$ on this and i think at the start even 100$ is going to be a hard sell untill a ton of games support it and show a diffrence because of that support
Chalnoth
11-Mar-2005, 06:13
Again, there isn't a really huge "wow" factor between on board audio and a Audigy2 zs...but there is enough of a difference that $50-$100 retail cards are sold in a viable market.
Here you have an entirely different scenario, however, as people have been buying sound cards for a very long time, so there is no need to create a new audience: it's already there.
Chalnoth
11-Mar-2005, 06:15
Well i think they should put out a powerfull version for 100$ and then mabye a cheaper one for 50$ and a better one at 200$
Except nobody's going to be able to make one for as little as $100 until a mass market is available, let alone $50.
digitalwanderer
11-Mar-2005, 06:21
If it adds enough whiz-bang to gaming I'd be ok with paying $200 for a mid-range one, but if it does NOT add much whiz-bang to gaming and I paid $200 for it I'd be pissed off.
I imagine a whole bunch of other people would be too, and it would bomb and fail.
That's why I'm going optimistic about it. I'll give it a shot and see if it can prove itself or not.
Chalnoth
11-Mar-2005, 07:28
I don't see how it's possible.
digitalwanderer
11-Mar-2005, 13:49
I don't see how it's possible.
Me neither, but I'd love it if they can pull it off. :)
Joe DeFuria
11-Mar-2005, 13:53
Again, there isn't a really huge "wow" factor between on board audio and a Audigy2 zs...but there is enough of a difference that $50-$100 retail cards are sold in a viable market.
Here you have an entirely different scenario, however, as people have been buying sound cards for a very long time, so there is no need to create a new audience: it's already there.
I don't agree with that logic.
Again, there isn't a really huge "wow" factor between on board audio and a Audigy2 zs...but there is enough of a difference that $50-$100 retail cards are sold in a viable market.
Here you have an entirely different scenario, however, as people have been buying sound cards for a very long time, so there is no need to create a new audience: it's already there.
I don't agree with that logic.
Sound cards first became available when the only other choice was cpu driven sound and I think youll agree that there was a big difference then. On board audio is getting a lot better and less people go for the addin cards now.
I like the idea of this physics coprocessor, but I think it would be better suited as a a coprocessor like the old floating point ones, situated on the mainboard. Might be a tad expensive for that, but I dont think its got much chance as a discreet addin card apart from a niche market.
Now if they could get a design win with a large OEM...
Chalnoth
11-Mar-2005, 16:59
The only possibility for a physics coprocessor to make it would be as integrated into current CPU's or GPU's.
Dave B(TotalVR)
11-Mar-2005, 17:21
I think the biggest and easily implimentable benifit of this PPU would be collision detectoin. a lot of games use bounding boxes and the like to define edges of non flat srufaces.
Look in counterstriek source. Good physics but low detailed (so to speak).
A filing cabinet might fall onto something and when you look at it, it isn't in full contact because the bounding box isn't detailed enough.
With this PPU you could probably define the bounding boxes for collision directly from the geometry. On top of that. Yes, you could define your sofa as 'squashy' so the filing cabinet woudl fall onto it, squashit down then spring back up a little causing it to roll off of it or something. As it is in CSS it will fall on the couch and stay ther like its made of steel.
Chalnoth
11-Mar-2005, 17:29
No, you wouldn't, DaveB. You'd simply do what all games that currently use polygonal collision detection do: use your normal bounding box to check and see if you should do more work with a specific object.
Anyway, lots of games in the past have used polygonal hit detection. I don't see why a PPU would really help.
skilzygw
11-Mar-2005, 17:56
My question:
Is this thing even necessary with Dual Core processors just around the corner? I mean cant you theoretically use an entire core just for physics.
(as you can tell I am a wondering tech Noob)
sonix666
11-Mar-2005, 17:56
I think the biggest and easily implimentable benifit of this PPU would be collision detectoin. a lot of games use bounding boxes and the like to define edges of non flat srufaces.
Look in counterstriek source. Good physics but low detailed (so to speak).
A filing cabinet might fall onto something and when you look at it, it isn't in full contact because the bounding box isn't detailed enough.
With this PPU you could probably define the bounding boxes for collision directly from the geometry. On top of that. Yes, you could define your sofa as 'squashy' so the filing cabinet woudl fall onto it, squashit down then spring back up a little causing it to roll off of it or something. As it is in CSS it will fall on the couch and stay ther like its made of steel.
And still they will use a big correction to make it possible to hit objects fairly easily at high distances. So much for per poly accuracy. ;)
Chalnoth
11-Mar-2005, 19:30
My question:
Is this thing even necessary with Dual Core processors just around the corner? I mean cant you theoretically use an entire core just for physics.
(as you can tell I am a wondering tech Noob)
Well, it's not as simple as that, but I think that's been a major argument in this thread: normal CPU's are typically powerful enough for physics calculations.
For the money Id rather buy a co-processor card with a Cell if it were available.
Dr. Ffreeze
11-Mar-2005, 20:18
Chalnoth,
Question for you, as you feel this PPU chip more than likely won't pan out. Would it be possible for an engine to get full use out of a 40,000 entity PPU if it only added flowing hair, cool water (liquids), flowing cloth, some extra entities when things blew up, and some deformable objects? If this were so, then a PPU would be able to increase the immersion but yet not have some of the issues you are describing (or would they?).
I don't know, I am just trying to think of a way that a PPU can add enough value to a game so that gamers wish to buy one without requiring a PPU to play the game. The only problem with my above ideas are if, as you say, they would still be too taxing on the system for the non physics code that would still need to be run in order to keep track of them and to render them properly.
Just a thought,
Dr. Ffreeze
11-Mar-2005, 20:24
All,
If a PPU like the one talked about were to be in many gamer's systems what types of other things could it be programmed to do? I don't have a clue what types of math functions a PPU is likely do in hardware.
What uses similar math?
3D Audio?
MPEG Encoding?
RC5 Cracking?
Gene Folding?
MP3 Encoding?
Other Media Encoding/Decoding?
Some cool fluid dynamics simulation (what is the airflow in my system REALLY doing?) =)
Other?
Thoughts,
trinibwoy
11-Mar-2005, 20:25
The only problem with my above ideas are if, as you say, they would still be too taxing on the system for the non physics code that would still need to be run in order to keep track of them and to render them properly.
What would this non-physics code be doing? For the scenarios you describe I don't think any AI routines would be affected by the authenticity of water, cloth or hair etc.
Dr. Ffreeze
11-Mar-2005, 21:05
trinibwoy,
What would this non-physics code be doing?
I really have no clue, that is why I asked Chalnoth as he seems to be a knowledgeable person and also a detractor of the PPU (not a bad thing). Personally I would love to see Better Games(tm).
Right now Chalnoth seems to believe that a PPU will not be able to deliver. I am just wondering if there can be any areas where he (or other detractors) would feel that the PPU might work. I tried to think of some things that a PPU might be able to be put to work that would not have the issues he mentioned. I don't know if I came up with any, so I asked him. =)
Did a little page dedicated to AGEIA PhysX PPU info:
http://personal.inet.fi/atk/kjh2348fs/ageia_physx.html
Dr. Ffreeze
11-Mar-2005, 23:07
SlmDnk,
Thanks man, I will take some time over the weekend and read you links.
Chalnoth
12-Mar-2005, 09:29
Question for you, as you feel this PPU chip more than likely won't pan out. Would it be possible for an engine to get full use out of a 40,000 entity PPU if it only added flowing hair, cool water (liquids), flowing cloth, some extra entities when things blew up, and some deformable objects? If this were so, then a PPU would be able to increase the immersion but yet not have some of the issues you are describing (or would they?).
See, here's the short of it:
In order to get people to pay hundreds of dollars for an add-in card (and, mark my words, it would have to cost hundreds of dollars at first because it would necessarily be a low-volume product at first), you need to really show them something incredible, something that makes people think, "Wow, I just have to have that."
But with these physics calculations, if you can get fairly close with just a CPU doing the physics, there never will be that wow factor (after all, people could probably get the same effect by instead upgrading their CPU). And without it, a physics processor just will never reach the economies of scale to ever become a wide-market product.
digitalwanderer
12-Mar-2005, 13:39
See, here's the short of it:
In order to get people to pay hundreds of dollars for an add-in card (and, mark my words, it would have to cost hundreds of dollars at first because it would necessarily be a low-volume product at first), you need to really show them something incredible, something that makes people think, "Wow, I just have to have that."
Ok, so far I am in total agreement with you. I think the first ones will be hundreds of dollars and they're going to have to have some pretty spectacular gaming improvements that are easily visible to sell 'em. http://mastabeta.com/forum/images/smilies/yep.gif
But with these physics calculations, if you can get fairly close with just a CPU doing the physics, there never will be that wow factor (after all, people could probably get the same effect by instead upgrading their CPU). And without it, a physics processor just will never reach the economies of scale to ever become a wide-market product.
Ok, this bit I agree with/disagree with; here's why:
I guess I'm just going under my natural assumption that these people might know more about it than I do since they're building/developing 'em and I'm just sitting here reading about them. I have a feeling that at some point one of their brighter puppies HAD to have had the same thought occur to him as it has to you, so why are they still thinking of trying to sell it?
Because maybe, just maybe, it DOES add a lot to games and they think it adds enough that people will pay for it.
I know the CPU could probably do the same things, but if it doesn't have to that frees it up for AI and all the other fun crap that brings my PC to her knees in some modern games.
I guess I'm just saying I don't see offloading more off the CPU as a bad thing if it actually works and if it doesn't work then we get to all have a good time ripping on them and pointing out all their flaws/BS.
It's kind of a win/win situation in my book. ;)
Chalnoth
12-Mar-2005, 16:18
Not every product put together by smart people has proven to actually be good for games. The NV1 is a good example of this.
Joe DeFuria
12-Mar-2005, 16:21
Not every product put together by smart people has proven to actually be good for games. The NV1 is a good example of this.
This is assuming of course, NV1 was put together by smart people. ;)
Chalnoth
12-Mar-2005, 16:27
Well, of course it was. It worked, didn't it? The problem with the NV1 was a problem in strategy, and such things happen in business all the time.
Joe DeFuria
12-Mar-2005, 16:30
Well, of course it was. It worked, didn't it? The problem with the NV1 was a problem in strategy, and such things happen in business all the time.
Right...they weren't that "smart."
See, here's the short of it:
In order to get people to pay hundreds of dollars for an add-in card (and, mark my words, it would have to cost hundreds of dollars at first because it would necessarily be a low-volume product at first), you need to really show them something incredible, something that makes people think, "Wow, I just have to have that."
Ok, so far I am in total agreement with you. I think the first ones will be hundreds of dollars and they're going to have to have some pretty spectacular gaming improvements that are easily visible to sell 'em. http://mastabeta.com/forum/images/smilies/yep.gif
But with these physics calculations, if you can get fairly close with just a CPU doing the physics, there never will be that wow factor (after all, people could probably get the same effect by instead upgrading their CPU). And without it, a physics processor just will never reach the economies of scale to ever become a wide-market product.
Ok, this bit I agree with/disagree with; here's why:
I guess I'm just going under my natural assumption that these people might know more about it than I do since they're building/developing 'em and I'm just sitting here reading about them. I have a feeling that at some point one of their brighter puppies HAD to have had the same thought occur to him as it has to you, so why are they still thinking of trying to sell it?
Because maybe, just maybe, it DOES add a lot to games and they think it adds enough that people will pay for it.
I know the CPU could probably do the same things, but if it doesn't have to that frees it up for AI and all the other fun crap that brings my PC to her knees in some modern games.
I guess I'm just saying I don't see offloading more off the CPU as a bad thing if it actually works and if it doesn't work then we get to all have a good time ripping on them and pointing out all their flaws/BS.
It's kind of a win/win situation in my book. ;)
I think it is more like this:
People have compe up with all kinds of neat things you could do with a fast, SIMD, superpipelined math engine in the last few decennia. Like the things people run on a GPU, or on a box filled with stream processors. Intel and AMD even have put some of them inside their CPUs.
But, when you have such a neat chip, how are you going to market it? You need to have some kind of volume market, all the individual highly specialized needs won't cut it. So they thought of the mainstream, and looked at what people want to spend quite a bit of money on: games.
digitalwanderer
12-Mar-2005, 17:48
And if they can add enough to games I'll buy their product. :)
I think it's too early to either dismiss or embrace this technology, we'll all have plenty of time and hopefully a lot more info before we even have a chance to buy one.
Chalnoth
12-Mar-2005, 18:26
Right...they weren't that "smart."
You know, that's a really stupid comment. A mistake doesn't make somebody unintelligent. There's not a scientist alive (or dead) who's not made a very significant blunder. It's always easy to look back with hindsight and say that something was stupid, but there are always good reasons for doing certain things.
With the NV1, for instance, nVidia felt that triangle-based rendering techniques were too limited, that they could make a splash in the market place by offering something categorically better. Unfortunately, they underestimated just how challenging it was to make games run on their hardware.
Joe DeFuria
12-Mar-2005, 18:28
Methinks you're taking this a bit too personally, Chal. You on the NV1 design team or something?
(Hint: Notice the " " I used when I wrote "smart".)
Chalnoth
12-Mar-2005, 18:40
No, I take offense at people labeling stupidity by mistakes.
DemoCoder
12-Mar-2005, 18:44
It is kind of a cheap shot. In the early days of 3D, people didn't know what approaches were going to work best. A failure doesn't indicate a lack of intelligence, just a misprediction. Everyone was experimenting then with different approaches. Invention and innovation themselves aren't a science. Finding a design that simultaneously does whats expected (neets requirements) and simultaneously satisifies the needs and desires of all others as well as unforseen scenarios is a hit or miss search in a massive search space. Atleast some part of having a successful product comes down to luck.
Is Carmack a smart guy? Did his curved surfaces ideas in the Quake3 era pan out and hit gold? Is Doom3 "the future" of game engines? Carmack mis-predicted and built something very nice that satisified his ideas and desires, but does not neccessary satisify the needs of other developers and the future market as a whole. This does not him "not smart". Sometimes someone hits upon a great idea, it pans out, and they are lauded. Sometimes they hit upon an idea which doesn't pan out and they are not. And sometimes, they have the right idea at the wrong time. It is possible for all three people to be equally smart.
During the dot.com era, millions of smart people were exploring a large search space of ideas and none of them really knew "what would work". Some hit the lottery, some didn't. Rationality can't always be used to pick a winner.
Joe DeFuria
12-Mar-2005, 19:18
No, I take offense at people labeling stupidity by mistakes.
Let me try again:
"smart"
Chalnoth
12-Mar-2005, 21:37
Let me try again:
"I take offense"
Joe DeFuria
12-Mar-2005, 21:42
Let me try again:
"I take offense"
And again...
"Methinks you're taking this a bit too personally."
ondaedg
12-Mar-2005, 22:02
Methinks you're taking this a bit too personally, Chal. You on the NV1 design team or something?
(Hint: Notice the " " I used when I wrote "smart".)
Why call someone stupid because of a design that didn't succeed? My guess is it had NV in the name. Better yet, I would call it an "educated guess". :roll:
I don't know about you, but I wouldn't call them stupid for creating some very advanced technologies at the time. Perhaps you should read up first:
[/url]http://www.firingsquad.com/features/nvidiahistory/page2.asp
Joe DeFuria
12-Mar-2005, 22:26
For crying out loud:
"Smart" = SMART IN QUOTES is not a literal definition of smart. By putting it in quotes I mean that in effect it didn't turn out to be the right decision.
Are we all clear on this now? Or should I make an "educated guess" that some people still won't find a clue even if it's spelled out for them?
And the link...thanks but no thanks. I was well into researching 3D gaming cards back in the NV1 (Diammond Edge) and 3DLabs' GamingGlint (Creative Labs' 3D Blaster) days. I know the history of NV1, and can even clearly remember the "tech ads" in ComputerShopper print magazine trying to educate the public: "See, we can draw this sphere here with 4 "supoer duper curved" polygons, vs. a hundred traditional triangles...and ours looks better to boot!"
Unfortunately for Diammond / NVidia, I was the type of consumer that was interested in what games would support it, aside from Pranzer Dragoon...and when it became obvious that proper Direct3D support wouldn't materialize, my interest dropped.
digitalwanderer
12-Mar-2005, 22:38
Or should I make an "educated guess" that some people still won't find a clue even if it's spelled out for them?
In my case probably "yes", but you probably already knew that too. ;)
Chalnoth
12-Mar-2005, 22:38
I guess my point is, avoid being sarcastic on forums.
Joe DeFuria
12-Mar-2005, 22:49
My point (not really directed at you) is...avoid being an ass on forums.
digitalwanderer
12-Mar-2005, 22:59
I guess my point is, avoid being sarcastic on forums.
My point (not really directed at you) is...avoid being an ass on forums.
So neither of you wants me to post anymore? :?
Fine, be that way. :cry:
"Smart" = SMART IN QUOTES is not a literal definition of smart.
I dont see anyone taking it like that, they understood it was sarcasm ... they just think you overestimate how much being smart has to do with being successful.
Briareus
13-Mar-2005, 21:32
Can someone explain why Sony's Cell architecture would be useful but a PPU would be a waste of resources?
Chalnoth
13-Mar-2005, 22:02
Can someone explain why Sony's Cell architecture would be useful but a PPU would be a waste of resources?
I see zero connection between these two things.
Can someone explain why Sony's Cell architecture would be useful but a PPU would be a waste of resources?
From what it looks to me the Cell is tonnes of integer units while the PPU would be floating point (or at least I think). Most game code would be integer while 3D math would be floating point.
Chalnoth
14-Mar-2005, 03:36
I don't think so, rwolf. Integer processing isn't good for geometry.
pegisys
14-Mar-2005, 03:42
I dont know if any one cares but I emailed AGEIA and asked how much a ppu would cost and I was told between 100-400 dollars for an add in card
digitalwanderer
14-Mar-2005, 03:50
I dont know if any one cares but I emailed AGEIA and asked how much a ppu would cost and I was told between 100-400 dollars for an add in card
Thank you, that is useful info.
So it sounds like there will be a broad spectrum of varies level cards to choose from, cool. :)
Now I just gotta know which is the best value/performance.... ;)
400$ no way will that sell in the first few years .
THey have to first have reasons for it. Put it out for a 100-200 and have games that make heavy use of it and show its strengths .
skilzygw
14-Mar-2005, 04:38
I seriously cant see this thing seeling if its not @ soundcard prices. Sub 100. Im a big time gamer, but im sure as hell not gonna spend 300 on a videocard to then spend 300 on this very narrow piece of hardware.
I can only see this being popular if it is included on a video card or on a sound card etc...
On its own @ 300+ no chance in hell of it succeeding.
I mean if PC games get to that point, i thrown in the towel and stick to consoles.
Id be interested to know if anyone here would actually be willing to but it at that price? Asside from all of us being fans of cool technology and i admit this is pretty damn interesting. I wouldnt in my right mind spend that kind of money(premium video card money) on such a single function card.
!Now if it sped up microsoft word that be awesome!
Can someone explain why Sony's Cell architecture would be useful but a PPU would be a waste of resources?
Cause you could use Cell for more than just physics (or games for that matter). No doubt this thing is also programmable, the programmability just wont be exposed from the looks of things ... what a waste.
Mulciber
14-Mar-2005, 07:02
people bought the Voodoo at $200 just to play less than a handful of patch-work 3d games.
if there are 2 or 3 really impressive games that require a $200 physics card, I'd buy one.
when you add in board cost, 128mb of gddr3 memory, and a 125M transistor chip, I dont see how they could possibly sell it for less than $150 and still make a decent profit. initially their production runs wont be 1/100th the size of ATI or nVidias, so they will be working on a much different economy of scale.
Chalnoth
14-Mar-2005, 07:32
people bought the Voodoo at $200 just to play less than a handful of patch-work 3d games.
if there are 2 or 3 really impressive games that require a $200 physics card, I'd buy one.
See, this is the problem. It can't be cheap because there won't be a large market for it. And I don't believe there will be a huge wow factor like there was at the inception of 3D graphics to sell the cards.
I feel that the only way we will see hardware dedicated to physics processing will be if it's integrated into either the CPU or GPU.
Unknown Soldier
14-Mar-2005, 07:41
There's gonna be more than one PPU?? I don't understand this. How will the cards differ?
I would've thought the PPU would be one card only ... since it is doing physics. How can it do less physic's detection?
*shakes head*
I wouldn't mind buying a card .. but i'd want the main card .. but at a good price.
US
Chalnoth
14-Mar-2005, 08:06
I would've thought the PPU would be one card only ... since it is doing physics. How can it do less physic's detection?
Not sure what the issue here is. It's still a processor. You can always have a faster processor.
That said, if this ever makes it to market, the first models will probably only be available in one flavor, since there just won't be enough volume to bother with other flavors.
There's gonna be more than one PPU?? I don't understand this. How will the cards differ?
More/less memory, different clock speeds...
Unknown Soldier
14-Mar-2005, 12:30
hmm.. that sucks.
US
Can't the PPU be used for things like making trees and grass react realistically to wind and movement? Can that be done with a PPU, and can it be done with just a dual core CPU? How about modelling friction properly? Deformable surfaces? A working trampoline, for example, or a springboard?
How long is it before we have AI working out how to use an NPC's muscles to make him run, stand still or get up off the floor, and then full phsical modelling of those processes? Like Honda's robots, but in an artificial gameworld (and hence with none of the engineering problems)
Tom
trinibwoy
14-Mar-2005, 16:11
Can't the PPU be used for things like making trees and grass react realistically to wind and movement? Can that be done with a PPU, and can it be done with just a dual core CPU? How about modelling friction properly? Deformable surfaces? A working trampoline, for example, or a springboard?
There are a couple demos on the website demonstrating the Novodex engine's handling of grass and friction.
digitalwanderer
14-Mar-2005, 19:52
400$ no way will that sell in the first few years .
THey have to first have reasons for it. Put it out for a 100-200 and have games that make heavy use of it and show its strengths .
Yes they will, people were buying SLI rigs before they were proven to work very well. ;)
The $400 will be the "reviewer edition card" methinks, and the $200 one will be their bread-n-butter.
Everyone here who plans to buy the bottom of the barrel card raise your hand. (The Dig looks around at the lack of hands) See? It's gonna be the mid-tier card that is gonna make/break 'em, so I expect it to be a pretty good value.
digitalwanderer
14-Mar-2005, 19:54
I don't believe there will be a huge wow factor like there was at the inception of 3D graphics to sell the cards.
You really don't, do you? Can I ask why? The idea of sooo much more physics being added to a game flat out excites me just thinking about it, why do you believe it'll be a big load of nothing? :|
(And I'm not trying to flame/fight you here Chal, I'm really curious.)
i wonder how this is going to work for mulitplayer games?
i can see it working with non-player interactive items.. but surely nobody expects it to have 10,000 or so items flying about ?
especially if not everyone has the add in cards . . . . .
Chalnoth
14-Mar-2005, 20:10
You really don't, do you? Can I ask why? The idea of sooo much more physics being added to a game flat out excites me just thinking about it, why do you believe it'll be a big load of nothing? :|
1. You can do quite a lot of physics just with a normal CPU. Current physics engines are quite efficient. Things will get even better when multicore comes out.
2. What, then, can you add that will leverage the power of a PPU, that couldn't also be done on just a faster CPU?
And for my answer to two I will just point out how new features in video cards are first implemented. Consider the high dynamic range lighting in FarCry as an example (I think it's a fairly representative example, personally). In this game you have a system that looks awesome in some places, but just plain bad in others. This really has always been the hallmark of tacked-on features.
See, 3D graphics was able to avoid this problem because all you have to do is design the game to run accelerated on the graphics card, and you get higher resolution and bilinear filtering for free. There was this immediate massive multiplication in visual quality that really made a good number of people willing to shell out a couple hundred dollars.
With physics, however, there's nothing that easy. You can't just "turn it on," and so it will require much more developer effort. But developers aren't going to be willing to spend that much time on a product that few people have, and most won't buy. Most will only be willing to say something to the tune of, "Okay, now that Havok supports AGEIA, I'll go ahead and give users the option to make the cloth and water simulation in my game higher-resolution."
Most users won't even be able to tell the difference. And many of those enthusiasts that might care would probably already have a fast enough CPU to turn these things up anyway.
Keep in mind that I'm not saying that we shouldn't have better physics simulation. Of course we should. I just feel that it's going along fine, and that we have all the processing power we need in the CPU and the GPU to keep physics simulations moving forward well enough.
Colourless
14-Mar-2005, 20:10
I imagine the difference would be something like this. With the card you would get much richer environments with more 'incidental' items that don't really do anything other than decorate the environment. They would all be able to be interacted with, even if they do nothing. Without it, the items would be fixed, if they are even there at all.
Chalnoth
14-Mar-2005, 20:13
I imagine the difference would be something like this. With the card you would get much richer environments with more 'incidental' items that don't really do anything other than decorate the environment. They would all be able to be interacted with, even if they do nothing. Without it, the items would be fixed, if they are even there at all.
And my argument against that sort of implemention is:
1. We have games that allow interaction with rather large numbers of objects today. How is AGEIA going to help us beyond what faster CPU's will do naturally?
2. How many game developers will be willing to spend the time to add in these extra objects for only a very few select users?
Colourless
14-Mar-2005, 20:22
Yeah, I don't really know if it is a 'good' idea. Replying to your points:
1) It all depends how much faster it actually is, and how much it costs. There is the posibility that it would allow somone with a slower processor to install a card and make their 'games' run better. Assuming that it costs less to buy a card then it would to get a new CPU
2) Probably not very many, unless there was some marketting agreement.
2. How many game developers will be willing to spend the time to add in these extra objects for only a very few select users?
Simple, the thing just needs its own killer application, like GLQuake was for initial 3D acceleration. You are probably looking in the wrong direction though, if you are expecting this killer app to be a next upcoming triple-A title from usual suspects.
Perhaps here ... (http://www.gamasutra.com/features/20050207/carless_01.shtml), perhaps something new entirely.
flick556
15-Mar-2005, 01:14
I think Unreal Engine 3 support alone will give them a good bit of sales. The question in my mind is how the physics simulation will scale between those that have the ppu and those that don't. If developers are forced to choice to either add tons of geomatry with physics properties and only people with a ppu will get good performance or scale back for the cpu only people I think they will almost always choose the later. If UE3's caned effect systems scale well to take advantage of the ppu and scale down for people without it than I think many people will want one.
Going forward this technology will likely become standard either integrated on the graphics card or the cpu
The $400 will be the "reviewer edition card" methinks, and the $200 one will be their bread-n-butter.
Everyone here who plans to buy the bottom of the barrel card raise your hand. (The Dig looks around at the lack of hands) See? It's gonna be the mid-tier card that is gonna make/break 'em, so I expect it to be a pretty good value.
Don't forget that this is computer hardware we are talking about. Within six months they will cost $20. :lol:
Chalnoth
15-Mar-2005, 02:42
Don't forget that this is computer hardware we are talking about. Within six months they will cost $20. :lol:
You're making a statement without any consideration to why costs lower for most new products in the hardware market.
Prices come down because of economies of scale. As more and more products of a certain type are produced, the marginal cost for producing one more becomes less and less (due largely to refinements in manufacturing and R&D costs that have been recouperated).
If these things don't sell in volume, then the price cannot come down. Without really convincing lots of people that they really have to have the product, they just won't sell in volume.
digitalwanderer
15-Mar-2005, 02:53
Without really convincing lots of people that they really have to have the product, they just won't sell in volume.
And there in lies the entire difference in our opinion. :)
I of the mind that the people making these cards know something we don't and this technology will have some serious whiz-bang/must have stuff come out with it, or else they wouldn't be doing it.
You seem to be of the opinion that they're making these things and hoping they catch on in the industry.
It really is a case of we won't know 'til we know. :)
You're making a statement without any consideration to why costs lower for most new products in the hardware market.
Prices come down because of economies of scale. As more and more products of a certain type are produced, the marginal cost for producing one more becomes less and less (due largely to refinements in manufacturing and R&D costs that have been recouperated).
If these things don't sell in volume, then the price cannot come down. Without really convincing lots of people that they really have to have the product, they just won't sell in volume.
Of course, but since the company has already given an estimate on pricing and has working chips without any further R&D they could produce them very cheap if they could get the requisite volume. I was thinking of this the other day, what if they gave MS a sweetheart deal to use them in the Xb2, an IP licence to cover development cost, they would then be able to produce them relatively cheap for other markets.
colinisation
15-Mar-2005, 03:09
I know im in late but bear with me.
Was it not one of the purposes of vertex shaders to be able to off load some of the physics calculations from the CPU to the GPU ?
or is all the excitement about this new PPU down to the fact that it significantly faster than the vertex units found on modern day GPUs ?
I feel that the only way we will see hardware dedicated to physics processing will be if it's integrated into either the CPU or GPU.
I agree with Chalnoth. There. I said it.
I'll go one step further and predict it ends up in the GPU, that it gets incorporated into DXxx, and that one of the big boys buys up these guys or their assets a little down the road.
Uhhh, anyone notice way back there on page 2 or 3 that NV's official spokesman cleared their throat about "legal issues"?? Wuzzup wit dat? That sounds ominously like lawyer-boy thinking he's sitting on patents to clobber the upstarts with once it gets financially worthwhile to do the clobbering. . .
Chalnoth
15-Mar-2005, 04:09
Are you talking about the quote posted here:
http://www.beyond3d.com/forum/viewtopic.php?t=21017&postdays=0&postorder=asc&sta rt=63
If so, it sounded more to me like a statement of the way things are in the industry, rather than any sort of threat.
digitalwanderer
15-Mar-2005, 04:13
If so, it sounded more to me like a statement of the way things are in the industry, rather than any sort of threat.
I agree, it actually sounded like a fairly tepid reply from nVidia...I don't think they see this thing as catching on or being a threat right now.
Are you talking about the quote posted here:
http://www.beyond3d.com/forum/viewtopic.php?t=21017&postdays=0&postorder=asc&sta rt=63
If so, it sounded more to me like a statement of the way things are in the industry, rather than any sort of threat.
That's the one. Mebbee it was generic. Seemed an odd thing to toss out there to me. I still think he had IP issues in mind in more than a "like everybody" sense, whether or not it was NV's portfolio he specifically was thinking about as the hurdle.
Here's a wild shot in the dark (hey, at least I label mine as such! :) ) --Sage? Mebbee in the way it is a discrete chip communicating with the gpu?
[Note to self: Mention "Sage" and thread dies as natives flee in terror.]
I imagine the difference would be something like this. With the card you would get much richer environments with more 'incidental' items that don't really do anything other than decorate the environment. They would all be able to be interacted with, even if they do nothing. Without it, the items would be fixed, if they are even there at all.
And my argument against that sort of implemention is:
1. We have games that allow interaction with rather large numbers of objects today. How is AGEIA going to help us beyond what faster CPU's will do naturally?
2. How many game developers will be willing to spend the time to add in these extra objects for only a very few select users?
Imagine if all the characters have hair that flows in the wind or when you take out the chain gun you can blow holes in the wall. Imagine buildings that fall down when you blow up a weak support. Think about shooting a rocket launcher at a mountain and having boulders rain down on your opponents.
You are thinking in the context of todays games not what could be.
Chalnoth
16-Mar-2005, 07:43
Imagine if all the characters have hair that flows in the wind or when you take out the chain gun you can blow holes in the wall. Imagine buildings that fall down when you blow up a weak support. Think about shooting a rocket launcher at a mountain and having boulders rain down on your opponents.
You are thinking in the context of todays games not what could be.
And how would these games play on games without the physics processor? Surely you can't believe that there would be any games built around such a low-volume piece of hardware.
See, even assuming that this hardware could allow dramatically better physics interaction than we're going to get anyway in the next couple of years, it'd just take too much developer work to make something truly amazing for it to ever happen before these products reach market saturation.
BlackAngus
16-Mar-2005, 13:11
Imagine if all the characters have hair that flows in the wind or when you take out the chain gun you can blow holes in the wall. Imagine buildings that fall down when you blow up a weak support. Think about shooting a rocket launcher at a mountain and having boulders rain down on your opponents.
You are thinking in the context of todays games not what could be.
And how would these games play on games without the physics processor? Surely you can't believe that there would be any games built around such a low-volume piece of hardware.
See, even assuming that this hardware could allow dramatically better physics interaction than we're going to get anyway in the next couple of years, it'd just take too much developer work to make something truly amazing for it to ever happen before these products reach market saturation.
Well really as long as the physics are not involved directly in game play, and they are just used to enhance emersion it could be done.
I would like to see more life like hair, trees with leaves that all move in the wind, real dust storms not texture walls. All of those could use lots of physics power, and (hopefully) should not be too hard for devs to work with in accelerated vs non accelerated physics modes. I dont know for sure it how difficult it would be, but it does seem to me that if you keep accelerated physics doing emersion tasks, those could be written for in 2 modes rather easily, or at least far more so than if you intertwined game play and accel physics.
Just a thought.
Edited for clarity
Chalnoth
16-Mar-2005, 18:26
All of these dynamic objects will require a significant amount of work to make look right when done fully-dynamically. Anything that requires a lot of work just won't be done, or, at least, won't be done right when done for a small market.
Imagine if all the characters have hair that flows in the wind or when you take out the chain gun you can blow holes in the wall. Imagine buildings that fall down when you blow up a weak support. Think about shooting a rocket launcher at a mountain and having boulders rain down on your opponents.
You are thinking in the context of todays games not what could be.
And how would these games play on games without the physics processor? Surely you can't believe that there would be any games built around such a low-volume piece of hardware.
See, even assuming that this hardware could allow dramatically better physics interaction than we're going to get anyway in the next couple of years, it'd just take too much developer work to make something truly amazing for it to ever happen before these products reach market saturation.
Same way they do things now. You can turn on/off shadows in some games can't you. What is the difference? Turn off hair and get polygon texture head.
scificube
17-Mar-2005, 01:21
How about this scenario:
Given:
a. Devs can scale physics
b. Dual-Core processors are due this year and mult-core is the future
c. Unreal Engine3.0 supports multi-core CPUs
d. The Novodex Physics engine supports multi-core CPUs and can be accelerated by the PhysX chip
Assumptions:
a. The Unreal3.0 engine will not be alone in supporting multi-cores
b. Other physics middleware will not be alone in supporting multi-cores
Puzzle piece 1: Dual and multi-cores provide a good fall back to "software" handled physics interactions.
Puzzle piece 2: Puzzle piece 1 is true as long as devs ensure PPU accelerated physics interactions related to game play always have fallbacks that can be handled in software mode.
Puzzle piece 3: Puzzle piece 2 remains true until such a time that PPUs are commonplace as integrated mobo parts, into GPU solutions or part of CPU architectures probably starting from most to least likely. At such a time physics interactions no longer need a fall back and thus hardware-accelerated physics can enter into game play.
Puzzle piece 4: In the interim PPUs are used to accelerate non-game play interactions such as fluid dynamics simulating water, cloth on players and NPCs, fragment and particles during explosions, etc. and anything that would normally be handled in software by the CPU in the absence of a PPU to further remove the load the CPU has.
Note: It may not be necessary to wait for or depend on multi-core CPUs as a single core solution may be up to the task at least for a little while. Multi-core seemed more logical given Intel's recent demonstration of physics using the Unreal3.0 Engine on a dual-core processor and their relatively soon arrival to market as well as AMDs offering and IBM as well if they so chose.
I think this scenario is plausible. Support for the PPU would sort of be inherent if you take advantage of multi-cores like Intel displayed with RoboWars...it would not be a difficult task to have the engine put the PPU to the task in place of the CPU given I think this is precisely the case with the already existing Novodex engine.
Devs therefore need not labor hard or perhaps at all to support and take advantage of a PPU. In this sense the PPU could actually be looked at as a widely supported part in spite of.
The PPU would have the support it needs and then the only factor would be user adoption. If accepted then others would perhaps join the fray in particular ATI and Nvidia as to wedge out another contender to the add-in market. At this point some level of standardization would have already been or will be coming to a head. Most likely MS but an Open group isn't so out there either.
I think the PPU could have a future if events go something like this and devs act accordingly to what I would think is common sense.
I apologize for being so verbose :D
Joe DeFuria
17-Mar-2005, 01:36
I think the PPU could have a future if events go something like this and devs act accordingly to what I would think is common sense.
I apologize for being so verbose :D
No apologies needed...and your post sums up my feelings almost exactly to boot! :)
Chalnoth
17-Mar-2005, 01:49
Same way they do things now. You can turn on/off shadows in some games can't you. What is the difference? Turn off hair and get polygon texture head.
You're not understanding what I'm trying to say. What I'm trying to say is that it's too much work to get these things to look good for game developers to bother for such a miniscule market.
Chalnoth
17-Mar-2005, 01:51
How about this scenario:
Given:
a. Devs can scale physics
b. Dual-Core processors are due this year and mult-core is the future
c. Unreal Engine3.0 supports multi-core CPUs
d. The Novodex Physics engine supports multi-core CPUs and can be accelerated by the PhysX chip
Problem here:
You don't need an API like Novodex to make use of multi-core CPU's. These will, in fact, reduce the need for any physics processor.
Problem here:
You don't need an API like Novodex to make use of multi-core CPU's. These will, in fact, reduce the need for any physics processor.
The benefit is that it does not preclude you from utilizing additional CPUs it just provides an API that will do the best with what hardware is available, including a dedicated physics chip.
Chalnoth
17-Mar-2005, 03:03
But what does that buy you that a physics engine like Havok does not?
Same way they do things now. You can turn on/off shadows in some games can't you. What is the difference? Turn off hair and get polygon texture head.
You're not understanding what I'm trying to say. What I'm trying to say is that it's too much work to get these things to look good for game developers to bother for such a miniscule market.
True enough. Considering the cost of game development that is a big risk for developers. But if you get several kick butt titles that make it worth while it could be quite successful.
But what does that buy you that a physics engine like Havok does not?
I don't know anything about Havok, but Novodex will scale across processors because it is multi-threaded. So it should technically run fast on HT, multi-processor, or dual core boxes. You can buy the hardware for it and it might just scale across that and the CPUS.
Can Havok do that? I don't know. Novodex is saying they are the ONLY threaded API out there. I would say that is a big feature.
Chalnoth
17-Mar-2005, 06:41
I don't know anything about Havok, but Novodex will scale across processors because it is multi-threaded.
And what I'm suggesting is that there's nothing preventing a physics engine like Havoc from also being made multi-threaded. I know that currently Epic is working on making UE3 as multi-processor friendly as possible.
See, the problem with making use of a physics API before the hardware is available is that it probably won't be as optimized as a full-blown physics engine that is designed to run in software. And once multicore CPU's are available, you can bet that every physics engine worth its salt will be designed to make use of them.
After all, though it isn't a 100% applicable analogy, have you ever seen a game run its software rendering code as a software implementation of Direct3D or OpenGL? There's a good reason for that.
Unknown Soldier
17-Mar-2005, 09:13
Since MS has bought rights to use Novodex's PPU in the X-Box2 (so i've been reading), isn't it possible that MS might add support for the PPU in it's next release of DirectX namely WGF 1.0 and 2.0?
US
I don't know anything about Havok, but Novodex will scale across processors because it is multi-threaded.
And what I'm suggesting is that there's nothing preventing a physics engine like Havoc from also being made multi-threaded. I know that currently Epic is working on making UE3 as multi-processor friendly as possible.
See, the problem with making use of a physics API before the hardware is available is that it probably won't be as optimized as a full-blown physics engine that is designed to run in software. And once multicore CPU's are available, you can bet that every physics engine worth its salt will be designed to make use of them.
After all, though it isn't a 100% applicable analogy, have you ever seen a game run its software rendering code as a software implementation of Direct3D or OpenGL? There's a good reason for that.
That is a lame arguement.
1. Havok is not multi-threaded and would have to be re-written. Novodex is available for developers now.
2. How on earth do you come to the conclusion that a software api isn't going to be optimized because the hardware may not be available? Ever think they may use alternate code paths?
3. Software renderers are slow because a GPU has multiple parallel processing units and dedicated hardware for processing triangles, matrix math calculations, memory bandwidth, pixel color blending and all other things a gpu does. The PPU for example has dedicated hardware designed to blast through the types of calculations that the CPU is slow at.
Chalnoth takes a moment to reflect ....
http://images.anandtech.com/reviews/cpu/multicore/Unreal3_bezerker_small.jpg
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2377
Since MS has bought rights to use Novodex's PPU in the X-Box2 (so i've been reading)
Now THAT's real news. I hope it's true.
Jawed
Don't know if anyone posted this yet....
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2376
Of course more parallel processing power will help no matter how it comes. But the deeper issue is data movement. The example AGEIA gave us to think about was a huge pile of bricks; when you push down on one brick near the top, forces are transferred to all other bricks in the stack. This may not be difficult on the scale to hundreds, but how about 30000 objects in a stack? How about keeping track of that while handling deformable (soft body) objects, fluids, and all collision detection in the scene?
The was all this needs to be handled is not simply with lots of parallel independent floating point power, but with lots of parallel floating point power connected by huge bandwidth. The fact that some initial Intel dual core chips will have to go off chip and back on to communicate, not as much performance is gained as possible. Certainly more parallelism is better no matter what, but it’s the high bandwidth that clenches the deal.
Chalnoth perhaps this is the answer you are looking for as to why we need physics hardware and why the CPU won't satisfy our needs for the long term.
Unknown Soldier
17-Mar-2005, 10:38
Since MS has bought rights to use Novodex's PPU in the X-Box2 (so i've been reading)
Now THAT's real news. I hope it's true.
Jawed
Hmm .. ok semms I was a bit off. MS Hasn't bought rights but NovodeX's PPU is planned to support the X-Box2.
--------------------------------------
AGEIA NovodeX Physics SDK to Support Next Generation Xbox Platform
Multithreaded NovodeX Physics SDK enables game developers to exploit powerful multicore architecture of next generation Xbox platform
MOUNTAIN VIEW, Calif. - March 9, 2005 - AGEIATM Technologies, Inc., a company dedicated to delivering pervasive interactive reality to next generation games, announced today that its NovodeXTM Physics SDK will enable unmatched performance for game developers targeting the next generation Xbox® platform. As a result, users of the Xbox video game system from Microsoft will be able to enjoy some of the most exciting, lifelike game experiences available on any platform.
As the games industry's first multithreaded physics SDK, NovodeX is uniquely equipped to exploit the multicore architecture of the Xbox, allowing game developers to deliver incredibly lifelike worlds and characters for Xbox users.
"AGEIA's NovodeX Physics SDK is a powerful middleware engine for game developers who want to make the most of the multicore architecture of the next generation Xbox platform," said Tracey Frankcom, program manager, Xbox Tools and Middleware. "The real-time physics capabilities of NovodeX, combined with the high-definition graphics and audio of the Xbox, enable the creation of totally immersive environments and unbelievably lifelike character interactions."
"The next generation Xbox platform is the first available game console system to unleash the multithreading capabilities of the NovodeX SDK," said Manju Hegde, CEO and co-founder of AGEIA. "Today marks the beginning of a new era of hardware-accelerated physics for developers and users alike, in which pervasive interactive realism enters the world of gaming."
About AGEIA
AGEIATM Technologies, Inc. is a fabless semiconductor company dedicated to delivering pervasive interactive reality to next-generation games. Its flagship product, the PhysXTM chip, is the world's first dedicated Physics Processing Unit (PPU) - a completely new hardware category that bridges the gap between static virtual worlds and responsive physical reality. The PhysX chip allows developers to use active physics-based environments for a more life-like entertainment experience. For more information about AGEIA visit www.ageia.com.
About Xbox
Xbox ( http://www.xbox.com ) is the video game system from Microsoft that brings people together for the most exhilarating game and entertainment experiences. Xbox delivers an expansive collection of breakthrough games, powerful hardware and the unified Xbox Live online service. The new tagline, "it's good to play together," captures the spirit of Xbox as the social hub of the new digital entertainment lifestyle. Xbox is now available in North America, Asia, Europe and Australia.
Xbox is a registered trademark of Microsoft Corporation in the United States and/or other countries.
Unknown Soldier
17-Mar-2005, 10:39
Meqon Gears Up For Next Generation Xbox Platform
Game Developer Conference, San Francisco, CA. March 9th. Meqon Research AB, creators of advanced physics middleware systems for the videogames market, is gearing up for the future, with new funding, new staff and a new license to develop their Meqon Game Dynamics system for the next generation Xbox® platform.
The company today announced it has completed a new round of funding, which has been raised specifically to allow Meqon to develop its game dynamics system for future videogames consoles.
The company has also strengthened its management team, with the appointment of a new Vice President of Sales & Marketing. Thomas Ahlström is a veteran of the videogames industry, with over 10 years experience in game design, production and management.
Meqon is already hard at work creating a new version of Meqon Game Dynamics for the next generation Xbox platform. The company has been awarded a license to create a dedicated version of the system to allow game developers creating new titles, to incorporate the most advanced game dynamics possible, including complete support for character physics, vehicle handling, fluid dynamics and full object interaction.
Meqon has already signed a number of high profile companies, including 3D Realms, creators of the much anticipated Duke Nukem Forever title. The company will also be announcing a number of new clients in the very near future.
Jonas Lindqvist, the Chief Executive Officer of Meqon, said, "We are delighted to be announcing a number of very significant new steps for our company at GDC 2005. The videogames industry is at a very exciting point. Games are reaching more people than ever before and the audience is becoming increasingly demanding in its expectations. The next generation of games consoles promises to blur the boundaries between gaming and movies. However to create truly interactive environments which act and 'feel' right, promises to be a major challenge. Meqon intends to be at the forefront of the new generation helping game developers create new experiences within games which act, feel and play better than ever before."
Tracey Frankcom, Program Manager, Xbox Tools & Middleware, agrees, "As the power of games hardware increases, the complexity of games must keep pace. Providing developers with tools that take advantage of the multi-core architecture of the next-generation Xbox platform to facilitate the creation of ever more sophisticated environments and interactions will breathe life into the high-definition gaming experience. This will be crucial to the next generation of games titles. We are pleased to have Meqon's Game Dynamics software in development for the next generation Xbox platform."
ENDS
About Meqon
Meqon was founded in 2002 and has its headquarters in Norrköping, Sweden. The company develops physics middleware for the computer game industry and has in its short existence enhanced both games and industrial products with its simulation products. More information available at http://www.meqon.com
About Xbox®
Xbox® (http://www.xbox.com) is the video game system from Microsoft that brings people together for the most exhilarating game and entertainment experiences. Xbox delivers an expansive collection of breakthrough games, powerful hardware and the unified Xbox Live online service. The new tagline, "it's good to play together," captures the spirit of Xbox as the social hub of the new digital entertainment lifestyle. Xbox is now available in North America, Asia, Europe and Australia.
Xbox is a registered trademark of Microsoft Corporation in the United States and/or other countries.
scificube
17-Mar-2005, 14:35
I guess I didn't cover enough bases in my earlier post to satisfy everyone.
Sorry about that. I've never claimed to be sharpest knife in the kitchen. :lol:
I'll try to fill in some of the gaps I've left open.
a. physic interactions on the CPU do not compare with what a dedicated part can do to the task. The obvious example is the GPU.
b. Right now there is a single API that will allow you to use a PPU. Although not exactly the same it bears some resemblance to "GLIDE" in essence when 3Dfx shook up things in the graphics world back when. It's not so scary for the moment and I tried to allude to others coming in and making a bid as well. I would imagine the minds behind Havok, Maqon etc. woud not sit idlly by and get out gunned by the Novodex crew. It is also a logical asumption that they will support multi-core regardless of Novodex because they would do so for mult-cores from Intel, AMD, IBM, Transmeta...whoever in spite of. I also suggested that at some point standardization would enter into the picture as is only logical for the benefit of devs and unlike the sound card market where Creative's EAX is all that is used because Creative owns most all the market...I think others will join the fray and compete with Aegia. (is that a sentence?)
Whoops Gotta Go!
Here is a link to a the Novodex Beta:
edit: clearly the fist page of this thread slipped my mind...but no harm's done right?
Building Exlode and Big Bang will make it quite clear CPUs are not up to the task for this level of physics interaction.
http://www.ageia.com/novodex_downloads.html#
Now simply exchange boxes for fluid dynamics etc for the interim period and put the boxes back into play when PPUs are standard parts.
Sorry gotta go now. I'll be back to see if I've fumbled or not.
Just a wild idea, but what if the middleware is ported to Cell too. So games could use Cell SPU's for physics and could be with reasonable effort be ported to PC's, leveraging the AEGEIA PPU to get comparable performance.
Otherwise games that make good use of Cell SPU's can't be ported to PC's. There just isn't that kind of number crunching power available on the PC.
Chalnoth
17-Mar-2005, 17:17
1. Havok is not multi-threaded and would have to be re-written. Novodex is available for developers now.
Updated. Not rewritten, updated.
2. How on earth do you come to the conclusion that a software api isn't going to be optimized because the hardware may not be available? Ever think they may use alternate code paths?
Not because it isn't available, but because it's designed for hardware acceleration, not a software implementation.
3. Software renderers are slow because a GPU has multiple parallel processing units and dedicated hardware for processing triangles, matrix math calculations, memory bandwidth, pixel color blending and all other things a gpu does. The PPU for example has dedicated hardware designed to blast through the types of calculations that the CPU is slow at.
My argument didn't concern software renderers vs. hardware renderers, but rather why no game (that I know of) has ever made use of a hardware 3D rendering API for its software rendering mode.
Chalnoth
17-Mar-2005, 17:21
Otherwise games that make good use of Cell SPU's can't be ported to PC's. There just isn't that kind of number crunching power available on the PC.
Except by the time the PS3 is out, we'll have multi-core CPU's. Not only that, but the PS3 will do all of its vertex processing on the CPU, and thus PC's will be able to offload most of the PS3 CPU's calculations to the GPU already. So no, I don't think the PC will be at a huge disadvantage to the PS3, even at release.
scificube
17-Mar-2005, 18:02
The Cell being a physics monster is interesting I guess should clarify it whether it will be as compared to a PPU or even mult-cores.
From the point of view that the Cell or the Xbox 360 simply outperforms PC mult-core CPUs then it would put pressure on the PC sector to respond.
As far as porting Cell code to PCs...I don't think this would be the case.
As I understand Cell requires one to go down and mess with the HW directly allot where as on the PC abstraction from HW is pretty much the name of the game. Unless I am mistaken it in order a code re-write or at least some significant re-translation would be necessary in any case.
This is assuming Sony or whoever does not introduce some abstraction layer.
Chalnoth
17-Mar-2005, 18:19
Yeah, I was just reading this article over at Anandtech about the Cell processor. This really is the direction that I've said in the past that all processors will need to go for eventually. But, yeah, this processor really is too specialized for anything to be ported close to directly. That said, there's little reason to believe that ports can't be made. After all, a large part of porting a program from the Cell architecture would be removing instructions (related to moving data in and out of each SPU's local memory), and moving other instructions from the CPU to the GPU.
Anyway, similar to the PS2, I really expect it to be a while before developers really get the hang of making full use of the Cell architecture. But, since all processor cores are going to be taking a route similar to Cell (few powerful single-threaded cores coupled with many simple cores), everybody else is also going to have challenges moving forward, too. What we really need, then, is a revolution in programming languages that really makes it much easier for programmers to handle multithreaded applications.
scificube
17-Mar-2005, 18:25
Otherwise games that make good use of Cell SPU's can't be ported to PC's. There just isn't that kind of number crunching power available on the PC.
Except by the time the PS3 is out, we'll have multi-core CPU's. Not only that, but the PS3 will do all of its vertex processing on the CPU, and thus PC's will be able to offload most of the PS3 CPU's calculations to the GPU already. So no, I don't think the PC will be at a huge disadvantage to the PS3, even at release.
Where'd you hear all that? I thought this was still in the realm of conjecture.
I understand the rendering deal if this is true but how would this be true of physics interactions. If the PS3's Cell is rendering most of the time how would it still be able to handle the level of physics interactions a PPU could simultaneously?
Intel has already showcased a dual-core running the physics alone in a separate thread on a separate core using the Unreal3.0 engine. The number of objects on-screen wasn't particularly all the impressive and they spoke plainly about having to tone stuff down. You could also look at it like this. A single core CPU can handle 40 to 100 interactable physical objects given the types of interactions you elect they should have. So at best Intel had 100 objects fully interactable on screen. If both cores were put to task 200 3 etc with a linear growth relation to ability to perform physics interactions barring multi-cores do not forgo being general-purpose processors.
A PPU is noted to be able to handle 30,000 to 50,000 fully interact able objects. It is apparent even multi-core CPUs are and will be at a disadvantage for any reasonable amount of time or given some logical restrictions about the number of cores you could ever put on one chip and sell mass market.
In this respect it's probably even the Cell chip is at a significant disadvantage given what we think we know.
Correction what I think I know :D
scificube
17-Mar-2005, 18:33
Yeah, I was just reading this article over at Anandtech about the Cell processor. This really is the direction that I've said in the past that all processors will need to go for eventually. But, yeah, this processor really is too specialized for anything to be ported close to directly. That said, there's little reason to believe that ports can't be made. After all, a large part of porting a program from the Cell architecture would be removing instructions (related to moving data in and out of each SPU's local memory), and moving other instructions from the CPU to the GPU.
Anyway, similar to the PS2, I really expect it to be a while before developers really get the hang of making full use of the Cell architecture. But, since all processor cores are going to be taking a route similar to Cell (few powerful single-threaded cores coupled with many simple cores), everybody else is also going to have challenges moving forward, too. What we really need, then, is a revolution in programming languages that really makes it much easier for programmers to handle multithreaded applications.
I would certainly agree that it's going to be a significant challenge to thinking with respect to programming with the Cell and if Cell is the model for how mult-cores will go in the future. For that matter mult-threading period is a going to be a challenge.
I suppose you could put more "simpler" cores on a chip then more robust ones but then you may affect the nature of CPUs being genral-purpose parts adversely. It remains to be seen if the need for more raw processing power would be worth the sacrifice.
I also didn't mean to imply PS3 ports would be impossible. I meant to imply it's not going to be a simply task thus not an everyday occurrence with respect to PS3 ports.
...but I'm just shootin in the dark.
scificube
17-Mar-2005, 18:39
I apologize.
Here's a link to the Intel dual-core demonstration I was referring to.
http://hardware.gamespot.com/Story-ST-x-1570-x-x-x
It's the RoboHordes quicktime about a quarter from the bottom of the page on the right.
Joe DeFuria
17-Mar-2005, 21:39
What we really need, then, is a revolution in programming languages that really makes it much easier for programmers to handle multithreaded applications.
Tim Sweeney has been evangelizing the very same sentiment.
What we really need, then, is a revolution in programming languages that really makes it much easier for programmers to handle multithreaded applications.
Tim Sweeney has been evangelizing the very same sentiment.
Is this a conceptual breakthrough required (y'know, UI, organization, etc) or a grunt-work breakthrough, or both?
Joe DeFuria
17-Mar-2005, 22:33
Is this a conceptual breakthrough required (y'know, UI, organization, etc) or a grunt-work breakthrough, or both?
Both, I'd say.
Kinda like going from standard procedural code / manner of thinking to object oriented.
Both, I'd say.
Kinda like going from standard procedural code / manner of thinking to object oriented.
Calling all grad students. . .waiting for genius can be a bitch.
Hehe, since when is DirectX a standard? I think it would be pretty funny if MS decided to start licensing its use one day. But I guess some people think an MS' API is different than a manufacturer's API.
Neat stuff though, I'd definately pick one up if the price was right. This is where real game mechanics innovation is headed imo (physics I mean, not necessarily specialized hardware).
scificube
17-Mar-2005, 23:07
I only meant DX is a standard in the sense that it is something IHVs adhere to complying to after giving their input to MS about what it should be.
Something devs can rely upon to be the common denominator across the board with respect to rendering.
I doubt MS would be foolish enough to make anyone directly pay to use it...then again last I heard MS takes 400% profit off it's windows sales so...
No I don't hate MS.
I didn't mean standard in the sense of a level of excellence to be met but rather the level of compliance to the spec.
Also...
I liken this to the recent cross over of graphic techniques into factors in gameplay such as in Splinter Cell, Doom etc.
Physics isn't just fluff now such as in HL2 and most likely many other titles as well.
Chalnoth
17-Mar-2005, 23:17
Where'd you hear all that? I thought this was still in the realm of conjecture.
I understand the rendering deal if this is true but how would this be true of physics interactions. If the PS3's Cell is rendering most of the time how would it still be able to handle the level of physics interactions a PPU could simultaneously?
Um, who ever said that the PS3 would be acapable of the level of physics interactions that a PPU could perform? It's not like that's certain, either.
I didn't mean standard in the sense of a level of excellence to be met but rather the level of compliance to the spec.
Right, a spec controlled by a a player in the marketplace. They have made decisions that benefit one manufacturer over another, and have been partners with them in different context. If MS takes Xbox further and actually starts making graphics parts, or buys somebody that does, is DirectX still a standard? If they start charging for it, is it still a standard? Who's to say AGEIA won't support other company's physics processors down the line.
Basically, I'm just nitpicking. :P
Its not like there aren't any parallel programming languages out there - Linda, Erlang, and Mozart come to mind...
I'm not an expert on any of these languages, but it seems to me that the problem is not lack of innovation in the language field (parallel or otherwise, check out Beta for a really cool OO language), but the fact that there is no widely accepted "parallel" language XXX out there that the vast majority of coders familiar with procedural/OO languages (i.e. myself) can pick up, knowing beforehand that it will be well supported and at least fairly widely adopted in a particular programming field. Adoption of a new language is also daunting because most existing libraries and APIs (DirectX, OpenGL, middleware such as physics SDKs, game engines, in house code) are written in some "old fashioned language", which means you have to do a lot from scratch, or deal with the pain of interfacing between XXX and some other language...
Chalnoth
18-Mar-2005, 02:17
Another significant problem from a programming point of view is that many of those languages that are "more advanced" just aren't as efficient as C++. Since performance is the primary reason for going parallel, you'd really have to convince people that the performance is there.
I mean, there are a lot of languages that are really neat from a programming perspective (Smalltalk, for instance), but just don't perform very well due to a number of choices made in the design of such langauges.
That said, though I know many programmers have stated that it is very challenging to write good, parallel code....I'm just wondering why it's really that difficult. I mean, I've done some limited parallel programming. Granted, it's been on trivially-parallelizable code, but I guess I just don't see why developers can't easily avoid some of the problems related in that Anandtech interview, such as two threads each waiting for the other to complete (why not just have one master controller thread that handles all of this?).
Here's a couple reasons:
- it's easy to shoot yourself in the foot because (at least in C/C++) safe usage of concurrency primitives is up to the programmer (as opposed to being determined or at least partially checked by the language)
- race conditions and deadlock are problems that can be very hard to track down
- whether/when a problem occurs can depend very much on the OS and the hardware (e.g. can appear/dissappear by changing the underlying number of CPUs)
Chalnoth
18-Mar-2005, 07:04
Here's a couple reasons:
- it's easy to shoot yourself in the foot because (at least in C/C++) safe usage of concurrency primitives is up to the programmer (as opposed to being determined or at least partially checked by the language)
First of all, what are concurrency primitives?
That asked, there are lots of ways to shoot yourself in the foot with C/C++ with just single-threaded programming. This is typically solved by having good programming habits. How is this any different?
- race conditions and deadlock are problems that can be very hard to track down
How is this a problem if you use a single master controller thread? Besides, there is debugging software available for multithreaded applications.
- whether/when a problem occurs can depend very much on the OS and the hardware (e.g. can appear/dissappear by changing the underlying number of CPUs)
Makes sense, which would make it harder to detect when there's a problem. But, like with so many things, this would all be made very simple with strict programming practices.
Simon F
18-Mar-2005, 09:16
What we really need, then, is a revolution in programming languages that really makes it much easier for programmers to handle multithreaded applications.
Tim Sweeney has been evangelizing the very same sentiment.
Occam with data structures?
vBulletin® v3.8.6, Copyright ©2000-2013, Jelsoft Enterprises Ltd.