• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Mirror's Edge: PPUs still useful

I'm not so sure... I think we need to see a lot more benchmarks with real games before we can make any kind of judgement on whether a PPU is still worth it now and in the near future.

Kyle, Brent, where are you guys? :D
 
I'm not so sure... I think we need to see a lot more benchmarks with real games before we can make any kind of judgement on whether a PPU is still worth it now and in the near future.

Kyle, Brent, where are you guys? :D

+1000
we need a good ol' [H]ard look at the current state of ppu's and gpu physx.

Comparo of PPU, the lowest gpu supported (8400gso??), 9600gt, 88/9800gt/x, 260, 280 as dedicated physx cards

It would also be cool to see how these cards perform as dedicated physx when they are run in a pcie x1 slot since not everyone will have a free x16/x8 slot especially if they are running an SLI/Crossfire setup.
 
Speaking from someone that owns the PPU, and have used it with the 8800GTX, GTX260, & GTX280....


I guess that if you care about PhysX, using an Ageia PPU will still be faster when you only have a single GPU.

I agree, if you have no other alternative (i.e. if you are on the Red Team, or if you're running a game thats maxing the programmable shaders) running the Ageia PPU is better than nothing on a game that has hardware PhysX.


Even GTX 260-216 + PPU is as fast as GTX260-216 + 9800GTX+.

I disagree with this... in most cases I found a single 260 running double duty was faster in most cases than the 260+Ageia PPU combo.
 
+1000
we need a good ol' [H]ard look at the current state of ppu's and gpu physx.

Comparo of PPU, the lowest gpu supported (8400gso??), 9600gt, 88/9800gt/x, 260, 280 as dedicated physx cards

It would also be cool to see how these cards perform as dedicated physx when they are run in a pcie x1 slot since not everyone will have a free x16/x8 slot especially if they are running an SLI/Crossfire setup.

+1. Would also like to add to definitely compare the 9600GT and 9600GSO, as someone here has suggested that the particular specs of the 9600GSO may make it a faster PPU card than the 9600GT, despite the 9600GT being the better graphics card otherwise.
 
Well I was running ME with a single GTX 260 OC'd (specs in sig) using Physx and saw noticable drops. So I picked up a 9600GT and the drops are completely gone. So it's definitely a great combo GTX + 9600GT. The 9600GT is about the best bang for your buck with Physx, it's alot faster than Aegia - but gives you great FPS bump for the least $$$.

Its rumored there are 600 games coming with Physx. EA is using it in most games they have arriving this year. Definitely a good time to jump on board. That said my buddy just picked up a used 9600GSO for Physx but he's waiting on his GTX to test it.

here's some proof the 9600GT is about the best bang for the buck, and shows this easily beating the GTX alone.

MirrorsEdge_PhysX_Skalierung.png
 
good. I got one cheap to go with my 4850

My buddy couldnt get that working with his 4850 and 9600GSO even in Windows 7. Apparently build 6519 is the only build it works in right now. Will be fixed by RTM, but it def wouldnt work in build 7000 the public beta.
 
My buddy couldnt get that working with his 4850 and 9600GSO even in Windows 7. Apparently build 6519 is the only build it works in right now. Will be fixed by RTM, but it def wouldnt work in build 7000 the public beta.

I'm using a PPU, not an Nvidia card. It works perfectly for me under Vista x64
 
oops, realized not needed...

Was tying to figure out where the last chart came from, but realized it was a summary
 
Here's the full review from Firingsquad:
http://firingsquad.com/hardware/mirrors_edge_physx_performance/

No wonder why developers chose Havok in the past, seeing how poorly PhysX performs on CPU. Its looks like a crappy physics engine, considering that even old games like Splintercell does much of the same with higher FPS on Havok (clothes effects etc.). Even when GPU is used, the performance hit is big.

I'm not impressed.
 
I'm using a PPU, not an Nvidia card. It works perfectly for me under Vista x64

yea, he's since bought my GTX 260 and has it working with a 9600GSO without a problem.

I upgraded to a nicely overclocked (756 core, 1566 shader, 1260 ram) GTX 280 ;)
 
I'm not so sure... I think we need to see a lot more benchmarks with real games before we can make any kind of judgement on whether a PPU is still worth it now and in the near future.

Kyle, Brent, where are you guys? :D

We are working on setting up previews now of games that actually have physics that impact gameplay. Not this Mirror's Edge trash add-in window dressing. That looks idiotic IMO.
 
Yeah, we shouldn't forget that PhysX hasn't been hardware accelerated (in the mainstream market) for more than a year so far (technically nVidia enabling PhysX on GF8 series GPUs at the end of last year was the start of this), so there was no way that mainstream games would feature HW-accelerated, PhysX-based gameplay elements. This year that might all change.
 
your opinion is duly noted. although, i disagree.

Its ok to disagree. :) What I find interesting, is that you say Mirrors edge have "higher level physics effects. What I see is: Ragdoll, minor use of smoke, cloth and debris/decals. These have all been used successfully on Havok without the huge performance hit.
Come on, 2.5X framerate hit with the use of CPU? For that little physics that Mirrors Edge possesses?
Nothing there impresses. I wouldn't even bother getting a PPU for that.

Mirrors edge doesn't show how good physics are on GPU, but how poorely optimized PhysX is on CPU. Its so poorly that it almost seem deliberate.

I mean, a GTX285, the fastest single GPU card of today, goes from 62fps(framerate cap) to 25fps when running this on CPU according to extremetech? Even Crysis with its custom physics engine (based on Havok) seem to perform better on very high and Crysis had more then just decoration effects.
Here you can see a bit what the engine supports, on CPU... :
http://www.geo2web.com/2008/12/07/crysis-physics-engine-movies/

More impressive then Mirrors edge IMHO! Now, with this on GPU, giving higher framerate without loss of PQ, then I might be more impressed. What Crysis deliver on eyecandy while also giving physics is much more then what Mirrors edge does.

So we disagree. I find that Mirrors edge is nothing but an example of physics that is. Even if PhysX wasn't propritary, I wouldn't want developers using it. In shader intensive games (like crysis), I need all the shader power I get and then its nice with a choice of CPU physics. I'd rather have more realistic models then some cloth moving and acting correctly in physics.

GPU physics is not a bad idea, please don't misunderstand that. I just don't find Mirrors edge as a good example of it. Its rather a good example of bad CPU optimization. I'm not fond of propritary tech in games either, but it has to start somewhere, so I applaud Nvidia's initiative.

Cryostatis might be more of an example of GPU PhysX. Lets hope they do something about the CPU PhysX as well. :)

As Kyle Bennett, I want more interactive physics thats more then eyecandy. I'll take Gordon Freemans gravgun over this any day. ;)
 
Its ok to disagree. :) What I find interesting, is that you say Mirrors edge have "higher level physics effects. What I see is: Ragdoll, minor use of smoke, cloth and debris/decals. These have all been used successfully on Havok without the huge performance hit.
Come on, 2.5X framerate hit with the use of CPU? For that little physics that Mirrors Edge possesses?
Nothing there impresses. I wouldn't even bother getting a PPU for that.

Mirrors edge doesn't show how good physics are on GPU, but how poorely optimized PhysX is on CPU. Its so poorly that it almost seem deliberate.

I mean, a GTX285, the fastest single GPU card of today, goes from 62fps(framerate cap) to 25fps when running this on CPU according to extremetech? Even Crysis with its custom physics engine (based on Havok) seem to perform better on very high and Crysis had more then just decoration effects.
Here you can see a bit what the engine supports, on CPU... :
http://www.geo2web.com/2008/12/07/crysis-physics-engine-movies/

More impressive then Mirrors edge IMHO! Now, with this on GPU, giving higher framerate without loss of PQ, then I might be more impressed. What Crysis deliver on eyecandy while also giving physics is much more then what Mirrors edge does.

So we disagree. I find that Mirrors edge is nothing but an example of physics that is. Even if PhysX wasn't propritary, I wouldn't want developers using it. In shader intensive games (like crysis), I need all the shader power I get and then its nice with a choice of CPU physics. I'd rather have more realistic models then some cloth moving and acting correctly in physics.

GPU physics is not a bad idea, please don't misunderstand that. I just don't find Mirrors edge as a good example of it. Its rather a good example of bad CPU optimization. I'm not fond of propritary tech in games either, but it has to start somewhere, so I applaud Nvidia's initiative.

Cryostatis might be more of an example of GPU PhysX. Lets hope they do something about the CPU PhysX as well. :)

As Kyle Bennett, I want more interactive physics thats more then eyecandy. I'll take Gordon Freemans gravgun over this any day. ;)

no, i do agree with you that mirrors edge specifically could be better optimized, but what i meant was hardware physx games in general in regards to higher level physics effects. mirrors edge is probably not the best example of what hardware physx is capable of. i actually think cellfactor would better fit the bill at this point in time, and that includes interactive gameplay physics as well. and crysis, while still relatively impressive in terms of physics compared to all the other games out there, still feels quite contrived in many instances of gameplay, imo.

consider also, that a single gpu in mirrors edge is already handling the graphics workload, which at high resolutions + AA is already heavy, so accelerating hardware physx is just going to overload it...hence the substantial fps drop you pointed out. so you can either utilize another gpu to offload the physics workload, or lower the graphics workload on the single gpu to improve fps, in most cases. just like in crysis, i can decide to lower my resolution, other graphics settings, or turn off AA to improve my fps. i can expect that you will see in some instances in the near future where the games will show you that truly impressive looking hardware physics effects run alot more efficiently on a decent gpu at this point than even a high end cpu due to the highly parallel architecture. but when interactive gameplay physics does get into the mainstream, you can expect the cpu to choke heavily. i don't see this changing unless something comes out that can prove otherwise.

so mirrors edge is what is it is. dice decided to port the game to pc from the consoles which were already utilizing software based physx, and then layered some "higher level" physics effects on top of it for the pc version. considering the short time frame they may have had to implement it, i would say it isn't great performance wise, but still satisfactory. perhaps if they developed the game with this in mind from the start, we might have had something truly amazing and also efficient in performance as well. so who knows? i am sure more games in the near future will pick up the mantle and deliver that which you're asking for. (if you love football, the game backbreaker hopefully coming out this year might give you something to cheer about.) but one thing i wouldn't expect is heavy cpu optimization on most of these future games, considering the track record. perhaps the games will at least be more balanced in terms of workload for both the graphics and physics. why not have the best of both worlds if possible? i guess we shall have to wait and see. but yeah, bring on the interactive physics!
 
Its ok to disagree. :) What I find interesting, is that you say Mirrors edge have "higher level physics effects. What I see is: Ragdoll, minor use of smoke, cloth and debris/decals. These have all been used successfully on Havok without the huge performance hit.

Care to list any real world evidence of this?
 
While Crysis has some very nice physics effects, if I run the CPU benchmarks included with the game, my E6600@3 GHz slows down to < 10 fps aswell. And Crysis doesn't even use cloth or fluid effects (which are very, VERY computationally expensive, and which is what Mirror's Edge uses quite a lot).

Heck, even good old Half Life 2 (Havok) takes a huge framerate hit when you throw a grenade into a stack of crates and other assorted rubbish. It will go well below 50 fps... while the game runs at 150+ fps most of the time.
And those physics aren't anywhere near as impressive as Crysis or Mirror's Edge.
 
no, i do agree with you that mirrors edge specifically could be better optimized, but what i meant was hardware physx games in general in regards to higher level physics effects. mirrors edge is probably not the best example of what hardware physx is capable of. i actually think cellfactor would better fit the bill at this point in time, and that includes interactive gameplay physics as well. and crysis, while still relatively impressive in terms of physics compared to all the other games out there, still feels quite contrived in many instances of gameplay, imo.

Cellfactor might be a better example. :) If to compare Mirrors edge, with its repeatative scenery and limited physics effects, to as example Far Cry 2 's world with interactive physics elements and enviromental effects, Mirrors edge falls short with its cloth and decals. If you haven't played Far cry 2, just look at the video's in youtube or something showing the effects. For those who wants to show the impact (Atech?), physics is scalable to some degree in Farcry 2.

Having a physX wrapper for some added eyecandy in games have already been in the market through Aegia. Nothing new and exciting.

consider also, that a single gpu in mirrors edge is already handling the graphics workload, which at high resolutions + AA is already heavy, so accelerating hardware physx is just going to overload it...hence the substantial fps drop you pointed out. so you can either utilize another gpu to offload the physics workload, or lower the graphics workload on the single gpu to improve fps, in most cases. just like in crysis, i can decide to lower my resolution, other graphics settings, or turn off AA to improve my fps.

Which is much of my point. Sometimes games are very shader intensive for added photorealism and other things. Hence the need for optimized CPU offloading of physx. Mirrors edge showed how it shouldn't be done. Us gamers want much and with an as little framerate hit as possible. Games are also becoming more multithreaded and takes more advantage out of several cores. Shaving 40+ fps going software/CPU isn't acceptable. Seeing the small gain with some added effects makes it even less worth it.

Gimme a gravgun for physics and rather use the GPU for some heavy shader photorealism and I'll take that over some glass debree and banners. ;)
 
i think an ideal game to utilize gpu physx might be a next gen rts that isn't necessarily cutting edge graphically, but offers incredibly realistic physics in-game that actually affected the way you would play the game against other opponents. maybe something like full-blown destructability that would damage your units, and you wouldn't have to worry that much over fps like in a twitchy competitive fps, lol. it would certainly give the genre a shot in the arm considering how stale it has become recently, at least for me personally. and it would have to be a pc exclusive since it would make use of hardware-accelerated physics. hmm, too bad this hypothetical game doesn't exist, otherwise i might have been a killer app for the ppu, lol. sorry, just thinking out loud...

p.s. - (some other cool things to add in this game: real smoke/fog that works like the "fog of war" in most rts games but meanders and billows realistically to make it hard to see through, natural events/ disasters like realistic tornadoes that flip over vehicles to prevent them from being used & rivers of magma from volcanoes that destroy tank treads to make them immobile, buildings that never blow up or collapse in the same fashion & buildings that caught on fire could be affected by wind currents which could cause the fire to spread without the assistance of soldiers armed with flame throwers, and supersoldiers would lose pieces of armor and clothing as they are shot and damaged which would cause them to be killed more easily...all of it accelerated by the power that gpu phyx would provide. now, that would be a cutting edge rts game. wow, who wants to help me make this game, lmao. it would own all rts games)
 
i think an ideal game to utilize gpu physx might be a next gen rts that isn't necessarily cutting edge graphically, but offers incredibly realistic physics in-game that actually affected the way you would play the game against other opponents. maybe something like full-blown destructability that would damage your units, and you wouldn't have to worry that much over fps like in a twitchy competitive fps, lol. it would certainly give the genre a shot in the arm considering how stale it has become recently, at least for me personally. and it would have to be a pc exclusive since it would make use of hardware-accelerated physics. hmm, too bad this hypothetical game doesn't exist, otherwise i might have been a killer app for the ppu, lol. sorry, just thinking out loud...


Now we are talking about useful physics. :D

Next gen games doesn't have to be cutting edge graphics if they have an added playablility.
Far cry 2, since I've used that as example, have fire physics which interacts with wind physics. wind direction matters to how fire spreads and it can be used as tactics. These are useful things for physics.
GPU accelarated physics could potentially make a whole world act upon such and that would be worthy to sacrifice FPS and eyecandy for! It needs to be worthy though.
Look at Starwars - force unleached:
http://www.youtube.com/watch?v=lHHV77u0bUQ
Thats fun physics and done on CPU.

If GPU accelerated physics should be anything but a buzz word, it needs to give us more then what CPU physics can do. Some extra eyecandy or "water moving the correct way" is simply not worth it, then I would prefer that the shaders are used for more global illumination effects.

Havoks new engine gives us a lot on CPU optimized for multicore:
http://www.havok.com/content/view/72/57/

What we need is a reason to move it over to GPU and use some of our shader power. Otherwise, I prefer to keep my shaders available and rather use my CPU more efficiently.
 
Now we are talking about useful physics. :D

Next gen games doesn't have to be cutting edge graphics if they have an added playablility.
Far cry 2, since I've used that as example, have fire physics which interacts with wind physics. wind direction matters to how fire spreads and it can be used as tactics. These are useful things for physics.
GPU accelarated physics could potentially make a whole world act upon such and that would be worthy to sacrifice FPS and eyecandy for! It needs to be worthy though.
Look at Starwars - force unleached:
http://www.youtube.com/watch?v=lHHV77u0bUQ
Thats fun physics and done on CPU.

If GPU accelerated physics should be anything but a buzz word, it needs to give us more then what CPU physics can do. Some extra eyecandy or "water moving the correct way" is simply not worth it, then I would prefer that the shaders are used for more global illumination effects.

Havoks new engine gives us a lot on CPU optimized for multicore:
http://www.havok.com/content/view/72/57/

What we need is a reason to move it over to GPU and use some of our shader power. Otherwise, I prefer to keep my shaders available and rather use my CPU more efficiently.

the main problem, which has been mentioned before in other threads, is that of developers not willing to cut out half of their potential sales (for financial reasons, obviously) due to the gpu in a customer's pc. so that's a big reason why we don't have tons of games coming out with uber interactive physics running off the gpu (and that also isn't possible to do on the cpu feasibly), notwithstanding graphical complexity in a game. the second that happens, radeon users will be up in arms due to the game essentially being a nvidia exclusive title. a reason that dice realized with mirrors edge, most likely. it's much safer for them to just add "exclusive" physics effects than to alienate half their potential sales by turning the game into a crazy game physics playground (which would have been awesome, lol). they couldn't have anway, unless the game was built exclusively for the pc and gpu physx for nvidia from the ground up. so unless nvidia threw tons of money at a developer to build an exclusive game for its' gpus (maybe futuremark studios?), it might be awhile before this happens. though, that's not to say more gameplay physics won't happen right now, just not on the same scale that could be possible if gpu physx existed for both gpu camps. which is why i'm curious as to what game previews [h] will have for us to see.

*see my post script above - i should start my own dev studio, lol.
 
Back
Top