GRAW running with and without a PhysX PPU: The outcome is inside this thread

[Farid]

http://forums.overclockers.co.uk/showthread.php?t=17568825

10-16 FPS slower with hardware PPU

There's some pictures showing the "differences" visible with and without PPU.

I thought that accelerating an unquantifiable process such as Physics with a fixed function hardware running on a proprietary API wasn't a great idea at all. But actually, the same harware slowing down the game is something else all together.

And the worst part is that the visible "improvement" from the PPU is not impressive at all.
Some would argue that teh drivers are still early, or that adding another driver layer to a game wasn't such a great idea, but then again, this is a S3 Virge type of deal, except that linear filtering is more interesting , from a visual stand point than a few more particles, if you ask me, that is.

[Geo]

Not exactly a propitious beginning. You'd think, if anything, there would have been extra special attention to getting this right at launch to put the entire enterprise in the best possible light. . .

[Jawed]

v1.0 - doncha just love it, eh?

Jawed

[Cartoon Corpse]

probably take devs awhile to develop to it's capabilities?

first tries, not invested to heavily into that?

i'd prefer to see a wider variety of differences. not just explosions.

gothic3 is gonna use it, but they've already claimed it's just some additional eye candy, won't impact game experience really.

i hope it's capable of more than just putting more rubble in an explosion. hardly worth a seperate card if that's all it will be used for.

[digitalwanderer]

It wouldn't be bad if it put more rubble into the scene without slowing it down, but if it slows it down I just don't see the point/advantage. :???:

[Sxotty]

Well I would have to see it in motion, I mean it just isn't fair to judge it like that...

[digitalwanderer]

Well I would have to see it in motion, I mean it just isn't fair to judge it like that...
Fair, no; fun, yes!

Speculate-speculate-speculate-speculate!

[Zengar]

Well, really, what did you expect? Maybe some CPU power is saved off by offloading the physics to a PPU, but it is neglected by additional setup and rendering/batching cost. The correct way would be improving the CPUs(adding additional thread units).

[Geo]

What did we expect? Since when did adding a $200+ piece of hardware to your system suddenly get defined as expecting a free lunch when the darn system runs slower afterwards? If this was their expectation, they did a fine job of hiding it from the casual observer (that would be me). Did any of you who where following this closely hear anything from them that suggested slower overall fps with physx in place?

[tEd]

Well all it does is help with physics of the objects or in the case just having more objects without the cpu doing additional work and if you are already graphiccard limited then rendering more objects it is no surprise fps goes down.

ppu probbaly will show it strength when a game is cpu limited

[digitalwanderer]

Did any of you who where following this closely hear anything from them that suggested slower overall fps with physx in place?
No, but I'm pretty well known for being stupid and missing the obvious a lot. http://mastabeta.com/forum/images/smilies/yep.gif

[BRiT]

What did we expect? Since when did adding a $200+ piece of hardware to your system suddenly get defined as expecting a free lunch when the darn system runs slower afterwards?

You mean like anyone who purchased a Creative EAX-enabled sound card, only to have it negatively impact performance when EAX features are enabled?

[Geo]

You mean like anyone who purchased a Creative EAX-enabled sound card, only to have it negatively impact performance when EAX features are enabled?

That would be a good example. Hopefully they got reamed for it. But they also had a m/o/n/o/p/o/l/y strong position in an established niche when they did it. This makes it somewhat easier to get away with, wouldn't you agree? Particularly when comparing an add-on feature that can be disabled vs the core function of the hardware involved?

[BRiT]

This makes it somewhat easier to get away with, wouldn't you agree? Particularly when comparing an add-on feature that can be disabled vs the core function of the hardware involved?

Oh, I completely agree. I was not expecting the performance in GR:AW PC to be worse with the added enhancements, especially after getting sucked in by the Cell Factor video. So now I'm left asking questions... Can accelerate frame-rates given identical workloads? Does it need to be designed for from the start? What is their killer-app? Did they reverse the benchmarks? Is that a pre-alpha stage? Can they improve performance; if so is it PhysX drivers or App-code?

Why are the numbers so low? Does MS have anything to do with this?

[DudeMiester]

It's just ... particles. :???:

[Zengar]

I was not following the developements at all, but maybe someone could tell me if it is possible to have a look at PhysX API. Is it free aviable? I don't really understand how they are dong it, because physics usually requires lots of memory access. Does this card have a onboard memory? So you basically upload the world information and then download the updated data and send it to GPU? That's a huge setup overhead.

[Guden Oden]

Anyone who was surprised that the FPS drops when the game draws tons more junk on the screen probably needs to get his reality specs adjusted at the optrician... A more reasonable test would be how the FPS is affected by having the CPU do all the extra physics calcs on top of running the game. Seeing as how HL2 sags just by blowing up a couple wooden crates, I would expect the impact to be a fair sight worse than just a 16FPS drop.

As for the 'improvement not being visually impressive', is that really the fault of the PPU? C'mon, if you're gonna bash it, at least use rational arguments.

This thread is crap. Haven't we learned the dangers of drawing far-reaching conclusions and making snap judgements from not just ONE example, but also the FIRST example by now? :roll: Maybe wait until more games come out first, yes? :roll:

[Rodéric]

Agreed.
To have a FAIR comparison, you would need run the game at EXACT SAME SETTINGS, w/ and w/o Physics acceleration, and see the change in FPS.

Having FPS drop when having more going on on-screen is expected...

[Rys]

Indeed. I've got a PhysX board (and full GRAW, cheers Ubi!) and the game uses it to add 'wow' effects rather than increase performance at the same physics 'detail'. And in motion it does look better (don't listen to the Frenchman!) as you play, just you take a framerate hit since the game is graphics bound (pretty severely it seem) and more's being rendered graphically because of it.

Pretty much the same happens with Bet on Soldier (the only other PhysX title I've scared up, which is also part of the current appeal), too.

It'll take time (as always) for the concept to take hold, both in games and developers using it to do meaningful stuff in their games (wow effects included).

I have to say the API is pretty decent from a programmers standpoint, so hopefully that won't be a barrier to entry for this go at accelerated physics.

[Jawed]

I still think that the specific geometry shading functionality of D3D10 is going to make this stuff entirely redundant. Effects-physics do not belong on PPU. Not when they then have to be routed via the CPU to the GPU.

The quicker Aegia gets compelling gameplay physics to the market, the better chance they have. One way I see this happening is that games ported from PS3, say, are only playable on PCs that have a PPU installed. Sure it marginalises the appeal of the PC version - but in comparison with the PS3 version of the game, the PC version will be marginal anyway just in terms of volume.

I can't say I'm in favour of PPU, per se - but I think this is Aegia's only chance.

Jawed

[Rys]

The quicker Aegia gets compelling gameplay physics to the market, the better chance they have.
Ageia have largely enabled this already. It's a developer thing now, providing they are keen on working with the hardware in the first place.

[blakjedi]

We compared this to the 360 version last month... no way this beats the 360 version. It may be art, it may be that the 360 was the lead system... no one knows why but the 360 looks a lot better and more realistic...

[mjtdevries]

Ok, so maybe we should have expected somewhat lower performance because the GPU has to draw more objects.

But we are only talking about something like 30 or 40 tiny and simple objects, that are added to the scene.
And that would be so much extra work for the GPU that the framerate drops with 40%??

I find that hard to believe...

[K.I.L.E.R]

Let me attempt to summarise this.

Devs get PPU board.
Devs go nuts adding bagillions of entities and effects into game.
Devs enable these effects when PPU is enabled otherwise they are disabled.
People with PPU boards complain of performance hit.

What would happen if all the extra effects were disabled and the PPU was being used?
I would like to see those results.

So far no one can comment on the PPU's performance until we see some equal ground benchmarks.
Also what about the fact that you need to store more data(extra particles in GRAW's extra detail mode) and send it across from RAM into the video card AND the PPU?
For particles the same data would be sent twice!
I had previously made a thread about this, about sending the data ONCE both to the PPU and the video card.

[Jawed]

It's not just about the GPU drawing, it's about the extra "conditional" workload on the CPU introduced by executing the extra effects physics, and moving the resulting data from PPU to GPU. Well, that's what I think.

Jawed

[K.I.L.E.R]

You cannot move directly from PPU to GPU.
It needs to go:
RAM -> PPU -> RAM ->GPU

It's not just about the GPU drawing, it's about the extra "conditional" workload on the CPU introduced by executing the extra effects physics, and moving the resulting data from PPU to GPU. Well, that's what I think.

Jawed

Just to mention my stance on PPUs:
They are currently ill-equiped with current PC hardware.
PCI-X will not fix the problem of multiple dispatching of the same data down the pipeline.
PC hardware needs to be redesigned in order to accomodate the PPU.

[czekon]

i was expecting to see a lot more of fireworks on the screen :???:

[pjbliverpool]

Here's a small preview on the PPU including some new GRAW and Cell factor vids:

http://www.pcper.com/article.php?aid=245

[Tridam]

It's been weeks I'm playing with PhysX and I can't say I'm enthousiast by what I saw. I planned to publish a review last month but I quickly forgot that idea when I realized I had nothing more to say than "PhysX board is a great cooler placed between the GeForce 7900 in SLI". Hopefully there is now more PhysX enabled content. But nothing worth buying it.

Driver structure sucks. Every app uses a specific driver rev. Each driver release is supposed to be included into the new ones but it's not always the case and you can't install an older driver rev. You have to remove the newer one and then install the old one. Then if you reinstall that newer rev it removes the older you need. It's a nightmare when dealing with a lot of beta apps. Even with games : unless you tweak driver files manually you have to switch from a driver to the other to run GRAW and BOS.

I've benchmarked GRAW and BOS in CPU limited scenes. As soon as a physics scene is handled by the PPU, there is a huge framerate drop. It's not at all because of the visual details increase. It's CPU limited (maybe because of a bad latency hiding or data conversion in a PPU friendly format). Moreover these extra details are just "visual". They don't interact with you. It looks nice in GRAW but IMHO awful in BOS.

[Farid]

This thread is crap. Haven't we learned the dangers of drawing far-reaching conclusions and making snap judgements from not just ONE example, but also the FIRST example by now? :roll: Maybe wait until more games come out first, yes? :roll:
Get a grip, chillax, no need to serve us the "I'm pissed at the internet, I swear" card, Guden.
No one said this was a fair review of the PPU and of its capabilities.

This thread is exactly what's it's advertised to be, a single example of the PPU (non) awesomeness. People can either discuss these particular results or speculate some more (with these numbers as their baseline or not, it's still speculation).
Nobody did set up a bait and switch type of deal in this thread.

If specualtion was a danger, this whole board would be under a Defcon1 alert 24/7.

Also, no theses PhysX games are not GPU bound, as stated by Tridam:
I've benchmarked GRAW and BOS in CPU limited scenes. As soon as a physics scene is handled by the PPU, there is a huge framerate drop. It's not at all because of the visual details increase.


Besides, as I said, it's not visually impressive, and that's a perfectly valid point, physics add realism and/or details to a scene. And in the case of a shooter like GRAW, all I'm seeing is just some more particles during explosions and a subtential framerate drop.
If that's fine for some of you, so be it, if some of you guys are not convinced but still want to give more time to the thing to show its usefulness, so be it and if some of you already consider this PPU as a joke, then, once again, so be it.

No one's speculations, or claims, are supposed to negate one another's opinion on a subject. If one's opinion is to say that it's too early to call anything on the PPU subject, then this one can say that in this thread, that's the point of a discussion, or else the thread title would have been "PPU's All Laughing and Hating Circle Jerk: No Constructive Opinions Required".


On the PPU subject, now, to me the thing is still nothing but a fixed function array of simple Vector Units, accelerating a barely quantifiable, across the different games, process, running a propietary API and adding another layer of drivers to the games it's supposed to accelerate.
The only idea of adding what is a closed platform IC to accelerate something like physics sounds inane to me.
Now, if, what I consider, a stupid idea, conceptually wise (read even if it worked perfectly and had great results), doesn't even do a good job, in its first showings, do not expect me to take a wait and see stance with regards to it.

What physics need, if you ask me, that is, is indeed more FP power, but more programmable FP power. Be it on the CPU or on the GPU.
When these ressources are not needed by a game for the physics, since not all the games need the same amount of physics calculations, they can be used for something else.

[Rys]

It's CPU limited (maybe because of a bad latency hiding or data conversion in a PPU friendly format).
Did some testing in CPU limited situations in GRAW last night to check. The 'cooking' phase of the PhysX simulation (where meshes are generated for the PPU to work on) does seem cause the drop in CPU limited situations, yes, but it seems to move on to the GPU at higher graphical settings (the drop is lower in my testing).

[IgnorancePersonified]

I want to believe :)

Any chance its the pci bus holding things up? Surely they are not that stoopid...

[K.I.L.E.R]

That would depend on how much the devs are sending down the bus, wouldn't it?

I want to believe :)

Any chance its the pci bus holding things up? Surely they are not that stoopid...

[Farid]

Rys's review is up:

http://www.hexus.net/content/item.php?item=5492&page=1

I remember the feeling the first time I fired up GLQuake on my Voodoo Graphics accelerator. I get almost the same feeling when I fire it up now. I don't get that feeling when I play PhysX-enabled titles, since the effects presented in the first wave of supporting titles isn't as overwhelming and persistent.

...

GPU effects physics (just Havok FXâ„¢ right now) is the easier bet for a developer looking to tack on effects-based stuff to an existing engine, too, especially if they already license Havok Physicsâ„¢. That means we pretty much just wait and see what happens.

Currently there's no reason to spend the £200+ to acquire one. The current effects in the supported games aren't worth the price and potential performance drop.

[Geo]

I liked the "under the covers speculation", tho winced at, err. . .266mhz or 533mhz! :smile: But then I suspect the author did too in having to write it.


The parallel elements are themselves made up of multiple fully FP32 processing units in MIMD, each with a SIMD array of vector units. Think of the hardware as an NxM array of vector units in that sense (if the patents are to be believed), likely 4x4 for this first iteration of PhysX hardware, which fits in with their likely makeup and implementation in silicon.

. . .

In terms of its data rate, the hardware is supposedly capable of 6 5D vector MADDs per cycle, per vector unit. In the 4x4 design we suspect first hardware to have, that's a near 50Gflop (all FP32 flops) rate when fully utilised, at 250MHz.



Well, given that 266mhz vs 533mhz plays a not-insignificant-role in the question I'm about to ask. . . would one of our stalwarts like to make a comparison of physx vs R580/G71 on math power?

And is it likely they've got any significantly-more-capable-than-usual-ALUs baking their cake under there?

[Jawed]

would one of our stalwarts like to make a comparison of physx vs R580/G71 on math power?
Ageia is fucked.

Jawed

[Geo]

Ageia is fucked.

Jawed

Awww, you jumped right to the conclusion! :lol:

I was looking for the breakdown.

[Jawed]

To be honest, this isn't about GLOPs per se.

Brute force, combined with bad algorithms/data-structures, is not a win (Geforce FX) - in other words just because a GPU has more GFLOPs doesn't mean it's better at the job - ATI admits as much saying that a GPU is best for effects physics. Ageia is really trying to define a new market, and effects physics isn't really it.

Ageia (I will get the spelling down, one day) needs to get past effects physics and get devs on board to make compelling gameplay-physics games. In GRAW, for example, that would be like blowing a gaping hole in the side of a building - not just scattering a few lumps of crap around and bending a lamppost (though the latter is cute, noticed it while playing last night).

I dare say this is all blindingly obvious and has been discussed to death. It's just a shame to see how even mere effects physics has completely defeated the devs of GRAW - if you're gonna do effects, at least produce a pall of smoke that hangs around and a layer of dust, and cars whose constituent bits actually look like bits, not whole doors. Ageia has had a long time to put together some devrel, and this is the result?

Jawed

[Sobek]

Ageia is fucked.

Jawed


I COL'd. (Chuckled Out Loud).

I would say that temporarily, you're right. :)

[PiNkY]

AnandTech
link (http://anandtech.com/video/showdoc.aspx?i=2751)

[_xxx_]

Ageia is fucked.

Jawed

It is, but regardless of the performance issues. The idea in itself is nonsensical, who needs yet another piece of specialized HW for something that can be sufficiantly well done without it? And with HW becoming ever more powerful with such a rapid development rate (multicore CPU's, GPU's supporting effects-physics), it'll be rendered obsolete before it takes off anyway.

[pc999]

If this is all it can do it doesnt worth IMO, but maybe we should wait for a game(s) that really uses it.

Anyway as I said before this and Intia processor integrated in a gamers motherboard would be very good as long as they can get relatively low price, at least AIseek is going the right way IMO.

[Titanio]

I want to believe :)

Any chance its the pci bus holding things up? Surely they are not that stoopid...

I was sure I read a comment from John Carmack raising this concern, that the bandwidth to the rest of the system would likely negatively affect framerate.

He also said this:

Once of the problems though it’s likely to actually decrease your wall clock execution performance, and this is one of the real issues with all sorts of parallel programming is that it’s often easy to scale the problem to get higher throughput, but it often decreases your actual wall clock performance, because of inefficiencies with dealing with it.

Basically, don't look for improvement in the rate at which things are done. The improvement would/could come in the amount you do per frame, not how many frames you do. It's up to devs to push a card to do more per frame such that if you were to do the same on a regular CPU, you'd wind up with totally unacceptable performance.

(And of course, there are the other issues mentioned of the GPU having more to render etc.)

[poopypoo]

Well I would have to see it in motion, I mean it just isn't fair to judge it like that...
http://images.anandtech.com/reviews/physics/asus/physx/grawbench.zip

...honestly, my impression after seeing it in motion is worse... :P it seems they apparently won't let us view the game with just CPU emulation of PhysX, so we can't judge. the improvement of the hardware in their own (shouldn't we assume this is best-case scenario?) benchmark is only ~10-30% according to anandtech's numbers (though I'd be curious to see some lower-end CPU performance numbers). I think it's pretty awful for $300, but i'm obviously a cheapish gamer. Maybe in a few generations this might be ok.

[czekon]

it's not worth this money, crap

[SugarCoat]

To be honest, this isn't about GLOPs per se.

Brute force, combined with bad algorithms/data-structures, is not a win (Geforce FX) - in other words just because a GPU has more GFLOPs doesn't mean it's better at the job - ATI admits as much saying that a GPU is best for effects physics. Ageia is really trying to define a new market, and effects physics isn't really it.

Ageia (I will get the spelling down, one day) needs to get past effects physics and get devs on board to make compelling gameplay-physics games. In GRAW, for example, that would be like blowing a gaping hole in the side of a building - not just scattering a few lumps of crap around and bending a lamppost (though the latter is cute, noticed it while playing last night).

I dare say this is all blindingly obvious and has been discussed to death. It's just a shame to see how even mere effects physics has completely defeated the devs of GRAW - if you're gonna do effects, at least produce a pall of smoke that hangs around and a layer of dust, and cars whose constituent bits actually look like bits, not whole doors. Ageia has had a long time to put together some devrel, and this is the result?

Jawed


Cell factor is apparently going to require a PPU to run so you'll get your game designed around physics soon enough. The driver problem mentioned though is a HUGE turn off for me. Having to load specific drivers for each title? Wow, talk about an early phase. Rather then seeing the same physics run on CPUs for comparison, i'd really like to hear what Havok thinks about the effects.

[tEd]

Not a good start.

Hopefully everybody can drop the voodoo card reference now :wink:

[Geo]

http://www.techreport.com/onearticle.x/9923

Gee, and people wonder why I keep saying that we should be discussing this in 3Dtech, rather than in the low-rent (for B3D anyway) district.


Havok's livelihood is no doubt threatened by Ageia's push into physics, but this hyper-aggressive approach has NVIDIA's sweaty fingerprints all over it, in my view.

[WaltC]

It is, but regardless of the performance issues. The idea in itself is nonsensical, who needs yet another piece of specialized HW for something that can be sufficiantly well done without it? And with HW becoming ever more powerful with such a rapid development rate (multicore CPU's, GPU's supporting effects-physics), it'll be rendered obsolete before it takes off anyway.

Agreed. I see these two factors primarily as the bane of the PPU in its present articulation:

(1) Price. $250 for an add-in physics PCI card is ridiculous, and imo, it costs as much as it does because for one thing the PPU has to be clocked and volted and heatsinked to try and make it fast enough in MHz to attempt to keep up with the current crop of 2-3GHz cpus. Just imagine what will happen to it when 65nm comes into full swing for cpu deployment. As we can see, even working in tandem with current cpus the PPU creates a measurable drag. IE, the PPU doesn't calculate nearly as fast as it needs to. How much faster isn't clear to me--I'd like to see some tests with much slower cpus that I think might shed some light on the subject. Maybe--hope not--but maybe it's just too slow, period. A price of $99 for a retail PPU would be much better--$49 would be just about perfect.

(2) Developer interest. IMO, from here on out game devs will be concentrating if not struggling to master multithreaded code so as to successfully leverage multi-core cpu environments. I can see ahead a time when extra physics can be managed by a second cpu core in a scalable manner, right from within a game itself. And so, imo, a PPU would have to be a good deal faster than that to justify itself. I just don't think cutting edge developers are going to have the time to invest in supporting a PPU, much less optimizing around one, when they have so much work to do and such a learning curve ahead just to properly support multi-core cpu environs.

[Mat3]

...I'd like to see some tests with much slower cpus that I think might shed some light on the subject....

I think test sheds some light on CPU vs PPU physics...

http://www.anandtech.com/video/showdoc.aspx?i=2751&p=4

Rather disappointing results actually. It seems like the processing cores just aren't powerful enough, and the only real advantage the physx processor has is the high internal bandwidth.

[IgnorancePersonified]

After finaly getting time to play the demo and coming to the realisation that it is in no way a showcase for 3D graphics nor a showcase for AI nor a showcase for FPS gameplay nor a showcase for a fast Dual cpu system with 2gb of memory I find it hard to understand why it should be a showcase for a much newer technology like a PPU?? It is a peice of shit demo that runs like a pig in every aspect (sorry if anyone was involved and reading this) Having a run of the mill sound card I left that out but it didn't sound that good either. Maybe Agiea *should* have made it a showcase for thier technology but it's my understanding they didn't code the entire game and possibly only provided input. As it is I think I will wait a little before reaching any solid conclusions on its(PPU) usefulness across a wide range of gaming applications and implementations.

[pjbliverpool]

It looks like the latest Ageia drivers seriously boost performance of the PPU in GRAW:

http://firingsquad.com/hardware/ageia_physx_ppu/page4.asp

The games no looses no performance on a single XTX system averaging 40.6fps and never falling below 33fps with all details maxed out and PPU physics on.

I personally think the demo looks great but only when you get into the more open areas where you can start to appreciate the massive scale of the city.

[Frank]

Want great physics on a PC? Wait for a PS3 on a PCIe card that will run the whole game by itself. Or for the quadro CPUs that also have lots of local storage. Or the DX10 GPUs that support a physics API.

;)

[Geo]

It looks like the latest Ageia drivers seriously boost performance of the PPU in GRAW:

http://firingsquad.com/hardware/ageia_physx_ppu/page4.asp

The games no looses no performance on a single XTX system averaging 40.6fps and never falling below 33fps with all details maxed out and PPU physics on.

I personally think the demo looks great but only when you get into the more open areas where you can start to appreciate the massive scale of the city.

I find that page to be a puzzlement in more ways than one:


In this case we saw a 3% decline in performance, with the minimum frame rate lowered by 3 fps. A 3 fps drop in minimum frame rates isn’t ideal, but it’s still a lot better than what users were reporting earlier, so it certainly looks like AGEIA’s working hard on getting the frame rate issues resolved. Unfortunately, we weren’t able to test PhysX out with SLI as NVIDIA’s current ForceWare 84.43 driver doesn’t have a profile for GRAW. When we tried to force SLI manually we actually got lower frame rates, so we’ll just have to wait for NVIDIA to issue a new driver with the proper SLI profile for GRAW (remember, GRAW is an ATI Get In The Game title, which probably explains why it has out-of-the-box support for ATI CrossFire).


1). I thot the consensus around here was GITG is no more and hasn't been for awhile?
2). Why does NV SLI "require a profile" for this title? Or rather, a profile beyond what FS could go make for themselves? Isn't that the point of user profiles in SLI?

[Cartoon Corpse]

maybe it would be better/happier on my 512M 1900xtx and fx60 dual core.

[karlotta]

maybe it would be better/happier on my 512M 1900xtx and fx60 dual core. it will be, that is the low to medium H/W to run the game.It needs to be at least at 1600/1200 , but is realy a 30in display game, so a CF R600 and a topline CPU iswhat this game needs ... let alone extra physics.. the havok stuff is the best i have seen to date in a game.. just the effects...wow.

[EasyRaider]

but is realy a 30in display game,
WTF? You can't be serious...

[karlotta]

WTF? You can't be serious...
Iam completely serious. The lack of AA requires a uber high res. my lil LCD is a 1600/1200 and its just passable. my 21in CRT at 2048/1536 is the way the game should be seen....I believe my lil 1800xt and a X24400 is the minimum spec for this game... Grinn made a choice, for the lowend today your fukd.

[Fred da Roza]

Iam completely serious. The lack of AA requires a uber high res. my lil LCD is a 1600/1200 and its just passable. my 21in CRT at 2048/1536 is the way the game should be seen....I believe my lil 1800xt and a X24400 is the minimum spec for this game... Grinn made a choice, for the lowend today your fukd.

Is there a way to force AA? I can't believe a modern game was designed without the option of AA.

This game definately is power hungry and I consider my PC better than average for gaming.

AMD 3000+
1.5 GB RAM
XFi sound
6600 GT

It runs but at the lowest res setting. And you definately need to crank up the res with this game. Need to upgrade that video card. So what's a good buy nowadays?

[karlotta]

Is there a way to force AA? I can't believe a modern game was designed without the option of AA.

This game definately is power hungry and I consider my PC better than average for gaming.

AMD 3000+
1.5 GB RAM
XFi sound
6600 GT

It runs but at the lowest res setting. And you definately need to crank up the res with this game. Need to upgrade that video card. So what's a good buy nowadays?
No, there is noway to have AA, or MSAA, with the lighting Grinn use's. If your going to upgrade is all about 512mb of vram and a 1900, 0r 7900... No other real choice.

[IgnorancePersonified]

The other choice is not to play the POS.

[Mintmaster]

It looks like the latest Ageia drivers seriously boost performance of the PPU in GRAW:

http://firingsquad.com/hardware/ageia_physx_ppu/page4.asp
Man, those guys are idiots. They run in a very GPU limited resolution, and conclude that because there's no framerate drop, the PPU must be great. :lol:

Then when they go CrossFire - which is probably still pretty GPU limited - the performance drop is because the "driver wasn’t built with CrossFire setups in mind." It's absolutely pathetic how so many performance oriented sites have no understanding of performance basics.

If you want to compare CPU/PPU speed, reduce the f***ing resolution.

What's worse is that they didn't even compare the old drivers. At least Anandtech did in their PPU driver update article:
http://www.anandtech.com/video/showdoc.aspx?i=2759&p=5
Result: New drivers do almost nothing.

[IgnorancePersonified]

Isn't that "Hangar of Doom" demo the same world they have been demoing previously? Means they have been playing around with UnrealEngine 3 for quite some. Wish they had come up with more though in that time if that is true.

[karlotta]

http://patches.ubi.com/ghost_recon_advanced_warfighter