[Skrying]

Originally Posted by Scali

I never said it was. You're the one being unreasonable, and making personal digs to boot.
I've put out some perfectly valid arguments, you haven't. You just throw things like "clueless" and "shit" around. That's not how you have a reasonable discussion.

The numbers show that enabling PhysX with GPU acceleration brings the FPS way down. You've argued that you have GPU to spare but that's hardly the case at all for many people as the most popular current resolution in sales is 1680x1050, not to mention upcoming games will only push this further. Yes, it comes down to a preference of accelerated physics or higher video settings. You've only continued to bring up how PhysX is a library but that doesn't honestly matter much as what everyone cares about (consumer minded) is it's acceleration features. I personally am not set on the idea of using the GPU to handle physics. Frankly I wish games would just offload better onto the multicore processors that are going to be much cheaper to obtain than a new video card. My video card is stressed to the max yet CPU utilization still remains less than stellar. Until features and looks slow down in the image quality I just don't see the point in using the most taxed processing unit in my computer (for gaming) to handle physics, it has shown exactly zero benefit so far.

[SirPauly]

Depends on how you're looking at it; tend to look at this as a huge increase in over-all FPS compared to a CPU.

[Scali]

My remark that PhysX was a complete physics library was to separate it from Havok FX, which can only do GPU-accelerated 'eye-candy' particle effects and such.
Bottom-line is, the developers can decide how much physics they cram into their game, and they can also offer various levels of physics detail.
You don't seem to understand that how much the physics bring the FPS down is not a constant, but a variable. The few games that use PhysX now, are basically showcases for the PPU, none of them were designed with a GPU in mind, nor were they designed to get the framerate up as high as possible. Their main goal was to provide the user with a lot of interaction and physics to demonstrate the selling points of a PPU.

Aside from that, many people on this forum are developers, and we are well aware that the difference between CPU physics performance and GPU PhysX is far too large a gap to be bridged by optimizing our multicore libraries and engines even further.
pjbliverpool already hinted at it, saying "game physics the like of which this makes possible".
GPUs make a level of physics possible that cannot be attained on a CPU, period.

It is highly possible that there will be games which focus less on visuals, but more on gameplay, where this gameplay is focused on interactive physics. Think of Portal, which uses portal rendering as the basis for its gameplay, and doesn't rely on fancy visuals.

Aside from that, there are the obvious facts that you can use multiple GPUs, and GPUs get faster every year (at a much higher rate than CPUs). So even if a single GPU can't really make a difference with physics today, it's obviously going to happen in the future.

I think you just suffer from a lack of vision, to be honest.

[Scali]

Originally Posted by SirPauly

Depends on how you're looking at it; tend to look at this as a huge increase in over-all FPS compared to a CPU.

Exactly, even though the framerate is lower than when you don't have physics, it's still playable. Not even the fastest CPU can make the physics run at playable framerates.

I think people who are focused on FPS miss the bigger picture.
For example, I played both HL2:EP2 and Crysis on my system.
HL2:EP2 got much higher framerates than Crysis, but even so, Crysis was my favourite, because with all its physics, AI and visual effects, it was a more immersive experience, more interesting gameplay. And I certainly didn't bother to turn down effects to get higher framerates. I ran everything at Very High. I just left AA off.

[trinibwoy]

Doesn't the option to dedicate a GPU to physics processing render this a moot point? PhysX now has a large installed base. And those folks who want the full works will spend more money on faster or multi-card setups. Same as always.....

[willardjuice]

PhysX now has a large installed base.

According to the Steam hardware about 19%. Now granted it's a few months old, but I also included the 8600/8500/8400 (and their mobile parts) series in that 19%. So I estimate less than 1/5th the user base can run PhysX on their gpu's.

[Scali]

Originally Posted by willardjuice

According to the Steam hardware about 19%. Now granted it's a few months old, but I also included the 8600/8500/8400 (and their mobile parts) series in that 19%. So I estimate less than 1/5th the user base can run PhysX on their gpu's.

However, the Steam userbase is not the same as the potential PhysX userbase.
A large part of the Steam users play older games like Counterstrike, or casual games like Peggle, which don't require cutting-edge hardware.

Games supporting GPU PhysX are still to be released. So the demographic is more people who play the latest games. They usually have more up-to-date hardware aswell, because these games require it.
I wouldn't be surprised if about 1/5th of the Steam userbase is the type that plays the latest games.

Aside from that, since pretty much all gamers use either nVidia or ATi, and their marketshare usually is about 50-50 at best, but mostly nVidia having more... at least half the gamers will eventually have an nVidia card that supports PhysX. Pretty much anyone who buys an nVidia card from now on, will buy a PhysX-capable card. The first cards that support PhysX are already about 2 years old.

Given all this information, I think people trying to argue against the PhysX installed base are just trying too hard.

[Florin]

Originally Posted by willardjuice

According to the Steam hardware about 19%. Now granted it's a few months old, but I also included the 8600/8500/8400 (and their mobile parts) series in that 19%. So I estimate less than 1/5th the user base can run PhysX on their gpu's.

Still beats DX10.1 nicely

[XMAN26]

Originally Posted by Skrying

The numbers show that enabling PhysX with GPU acceleration brings the FPS way down. You've argued that you have GPU to spare but that's hardly the case at all for many people as the most popular current resolution in sales is 1680x1050, not to mention upcoming games will only push this further. Yes, it comes down to a preference of accelerated physics or higher video settings. You've only continued to bring up how PhysX is a library but that doesn't honestly matter much as what everyone cares about (consumer minded) is it's acceleration features. I personally am not set on the idea of using the GPU to handle physics. Frankly I wish games would just offload better onto the multicore processors that are going to be much cheaper to obtain than a new video card. My video card is stressed to the max yet CPU utilization still remains less than stellar. Until features and looks slow down in the image quality I just don't see the point in using the most taxed processing unit in my computer (for gaming) to handle physics, it has shown exactly zero benefit so far.

Compared to running it on the CPU, I'll take GPU PhysX

http://www.firingsquad.com/hardware/...ance/page4.asp
http://www.firingsquad.com/hardware/...ance/page5.asp
http://www.firingsquad.com/hardware/...ance/page6.asp
http://www.firingsquad.com/hardware/...ance/page7.asp

Advantage of having a second card:
http://www.firingsquad.com/hardware/...ance/page8.asp

end of article has list of upcoming games with PhysX support

http://www.driverheaven.net/articles...d=122&pageid=4
yes overall compared to no physx, the games slow down, but the increased realism is worth it, just dont do it without a second card.

[willardjuice]

Originally Posted by Florin

Still beats DX10.1 nicely

Uh what? Go back into your hole until you have something useful to say.

at least half the gamers will eventually have an nVidia card that supports PhysX

I don't buy that half the market has a video card capable of PhysX (maybe in a year or two, but only if Intel fails to make a big enough splash) but I'll play along. As a so called "developer", you have to admit you wouldn't create a game that only half the market can play. Until almost the whole market has a PhysX capable device, we will never see anything more than extra "eye candy" physics from it.

[Florin]

Originally Posted by willardjuice

Uh what? Go back into your hole until you have something useful to say.

Considering that that 19% is owners of relatively high end hardware, it is a *massive* installed base. Obviously the point is that if you're looking at installed base like you do, then you're effectively advocating that DX10.1 support right now is a complete waste of time. I'm sure you didn't want to say that.

[Skrying]

Originally Posted by XMAN26

Compared to running it on the CPU, I'll take GPU PhysX

http://www.firingsquad.com/hardware/...ance/page4.asp
http://www.firingsquad.com/hardware/...ance/page5.asp
http://www.firingsquad.com/hardware/...ance/page6.asp
http://www.firingsquad.com/hardware/...ance/page7.asp

Advantage of having a second card:
http://www.firingsquad.com/hardware/...ance/page8.asp

end of article has list of upcoming games with PhysX support

http://www.driverheaven.net/articles...d=122&pageid=4
yes overall compared to no physx, the games slow down, but the increased realism is worth it, just dont do it without a second card.

Those are not exactly fair. We're talking about something setup to particularly sell a card. We'll see how this goes months from now. But I have this funny feeling PhysX won't produce anything truly worth it.

[XMAN26]

Originally Posted by Skrying

Those are not exactly fair. We're talking about something setup to particularly sell a card. We'll see how this goes months from now. But I have this funny feeling PhysX won't produce anything truly worth it.

In what way are they not fair. They show No PhysX, PhysX by CPU, PhysX by 1 GPU(GPU doing game and PhysX) and PhysX by 1 dedicated GPU(1 for game and 1 for PhysX). How much more fair do you want it to be?

[Skrying]

Originally Posted by XMAN26

In what way are they not fair. They show No PhysX, PhysX by CPU, PhysX by 1 GPU(GPU doing game and PhysX) and PhysX by 1 dedicated GPU(1 for game and 1 for PhysX). How much more fair do you want it to be?

Build a physics library and system around a quad core CPU and I'm willing to bet you'll get similar quality results. But a physics library and system around a PPU and you get these results. Nothing so far from any PhysX using game has shown effects beyond additional non-gameplay changing uses. Until I'm shown something in a major title that actually shows how useful the additional acceleration will radically further gameplay I don't care. I'm gameplay > graphics and effects every time, that's why from the start I've consider PhysX boring because it's shown nothing worth caring about for me.

[willardjuice]

Originally Posted by Florin

Considering that that 19% is owners of relatively high end hardware, it is a *massive* installed base.

What color is the sky in your world?

Obviously the point is that if you're looking at installed base like you do, then you're effectively advocating that DX10.1 support right now is a complete waste of time. I'm sure you didn't want to say that.

I don't see how this has any relevance with PhysX's market share. I would not recommend to anyone to base their decisions around PhysX nor DirectX 10.1 until they become pertinent factors. PhysX adds small "eye candy" effects to a few games and DirectX 10.1 adds a decent performance boost in a game that's not patched. Neither are earth shattering and should not be factored into a purchasing decision. Now I have no idea who you are, but would you kindly go away until you bring relevant and meaningful discussion to this thread.

[Florin]

Originally Posted by willardjuice

Neither are earth shattering and should not be factored into a purchasing decision. Now I have no idea who you are, but would you kindly go away until you bring relevant and meaningful discussion to this thread.

I'm not sure who you are either, but I do know you're being completely obnoxious.

You jumped into the topic offering your opinion that Physx isn't going to matter because sophisticated audio didn't either. This is a poor analogy, as was already pointed out (quite unlike DX10.1 support - which along with Physx, is just one factor that you might chose to take into account at your next video card purchase) but it remains the core of your argument. You've not really responded to anyone at all other than to claim that their opinions are irrelevant, so go on and listen to your own advice.

[willardjuice]

Physx isn't going to matter because sophisticated audio didn't either

but it remains the core of your argument

That's not what I said and it's certainly not my core argument. I used it as an example: EAX (hardware audio) is more profound right now than PhysX and yet it's still fairly irrelevant in the grand scheme of things. My core argument is this (pay attention closely) PhysX needs an overwhelming majority of market share to be successful and without AMD/Intel support, I have doubts that it will achieve this (it might just end up adding eye candy physics to a few games). Is that so hard to comprehend? You came in and started blabbering about DirectX 10.1 while making no connection whatsoever to the PhysX discussion. I don't mind Scali's arguments because I can at least see where he is coming from, your comments should be deleted because they offer no value.

[pjbliverpool]

Originally Posted by willardjuice

PhysX adds small "eye candy" effects to a few games and DirectX 10.1 adds a decent performance boost in a game that's not patched. Neither are earth shattering and should not be factored into a purchasing decision.

There might not be much that uses PhysX right now but it has huge potential. NV have allowed you to take your old GPU which would have otherwise been useless and add it into you your PC as a very powerful physics processor which as well as expanding hugely the potential for games on your PC, also if used correctly would free up a huge chunk of CPU power. Hell, if all games used PhysX acceleration for "regular" physics that would have otherwise been run on the CPU then it analogous to getting an upgrade from a 4 core CPU to an 8 core for FREE.

Even if you don't have an older CUDA capable GPU its still your choice whether you turn on these extra effects or not so nobody loses. This is free, you can turn it off if you don't want it and its potentially a huge value add, especially if you already have a GF8/9. Why this isn't seen as a universally great thing is beyond me.

[Scali]

Originally Posted by willardjuice

I don't buy that half the market has a video card capable of PhysX (maybe in a year or two, but only if Intel fails to make a big enough splash) but I'll play along. As a so called "developer", you have to admit you wouldn't create a game that only half the market can play. Until almost the whole market has a PhysX capable device, we will never see anything more than extra "eye candy" physics from it.

Actually there's a very good market for people with slower CPUs, but a reasonable videocard to offload PhysX to.
So I could design a game that either requires a quadcore CPU to run, or a PhysX-capable GPU (or a PPU ofcourse).
People that don't meet the minimum requirements are just out of luck, just as they've always been. Up until now we just had to have a fast GPU *and* a fast CPU. Physics is by far the most computationally expensive part of modern games, aside from the actual rendering. Offloading that from the CPU to the GPU could mean that perhaps people can stick to dualcores as long as they have a good GPU.

Aside from that, the line between gameplay physics and eyecandy is extremely vague.
Take Crysis for example. At different physics levels, you have different levels of destruction of trees, houses etc. However, you can use these destructive objects in gameplay, using it to kill opponents or take cover, and such.

[trinibwoy]

Originally Posted by willardjuice

As a so called "developer", you have to admit you wouldn't create a game that only half the market can play. Until almost the whole market has a PhysX capable device, we will never see anything more than extra "eye candy" physics from it.

That's true....developers will not push gameplay physics that can't be adequately run on a CPU. But they can certainly do more with scalable eye-candy and effects physics. It's still better than what we have now and a step forward. Not every card out there can run with maxed out detail or AA either....

[Scali]

Originally Posted by Florin

Considering that that 19% is owners of relatively high end hardware, it is a *massive* installed base. Obviously the point is that if you're looking at installed base like you do, then you're effectively advocating that DX10.1 support right now is a complete waste of time. I'm sure you didn't want to say that.

I'm quite sure he is, because the only people that could possibly have anything against GPU PhysX are the ones that can't make use of this free technology: the ATi users.

[willardjuice]

but it has huge potential.

Just because something has huge potential doesn't mean it will be a success.

People that don't meet the minimum requirements are just out of luck

That's a quite the market share you're leaving out. I don't know of any developer that would do this.

I'm quite sure he is

As I stated before (that you clearly just ignored), I put DirectX 10.1 in the same lot as PhysX.

[Scali]

Originally Posted by willardjuice

Just because something has huge potential doesn't mean it will be a success.

Or a failure.
I think his point is that people here are shooting it down every way they possibly can, while there's no way to tell if it's going to be a success or not. Just wait and see.
Being overly negative about something with a huge potential like this is just... weird.

Originally Posted by willardjuice

That's a quite the market share you're leaving out. I don't know of any developer that would do this.

Not at all. There are already games (I believe Assassin's Creed is one of them) that refuse to run on anything with less than two cores.
Games pushing the boundaries of hardware requirements is just a natural thing.
I've had to upgrade my hardware to meet the minimum requirements of a game many times...
I had to upgrade to 32-bit processors, 3d cards, more memory, new OSes etc, and ofcourse the overall performance requirements had to be met aswell. Just being able to run the game doesn't always mean it is actually playable.
My guess is that physics is the next big thing in games. Games like Max Payne 2 and Half-Life 2 gave us a first glimpse. Crysis pushed the envelope as far as it could go on current CPUs. Physics are becoming a big part of the overall gameplay, and developers will one-up eachother in the physics department. Currently you already need a dualcore to run a physics-heavy game like Crysis properly. A quadcore or PPU/GPU is the next logical step. It will happen, we just don't know exactly where and when.

Ofcourse, you could also take away most of the physics, so it runs on lower hardware, but then it probably will be an entirely different game altogether. Reminds me of Doom 3... It can be run without shadows on old/slow videocards. But the shadows are what Doom 3 is all about. It's what gives the game its dark and spooky atmosphere. Without the atmosphere it's about as scary as the original Doom.

[Skrying]

Originally Posted by Scali

I'm quite sure he is, because the only people that could possibly have anything against GPU PhysX are the ones that can't make use of this free technology: the ATi users.

Again, incredibly wrong assumptions. I guess my 8800GTS 512 was actually made by ATi according to you.

[Scali]

Originally Posted by Skrying

Again, incredibly wrong assumptions. I guess my 8800GTS 512 was actually made by ATi according to you.

Okay, then all you PhysX-haters... Explain in one sentence why you think a free, optional technology is bad.