Most demanding PhysX game? and what ded. card for 60fps in 3d surround?
Whats the most demanding currently, Mafia 2? Batman Arkham Asylum? Cryostasis?

My understanding is that higher resolutions are more demanding of PhysX (though I dont understand why, why should display res matter? Or does it just use up the shaders normally no longer shared with physx processing?), I could understand how increased FOV from 3 monitors could slow it if more is visible on screen and potentially animated there.

Since AFAIK even multi SLI graphics systems can be brought to their knees by certain games (Metro 2033) in 3d surround i'm assuming dedicated PhysX may improve performance as well.

But i'm curious where the radical improvements level off, i'm not looking for a 4% improvement spending twice as much, just to eliminate bottlenecks. (ie jumping from 13fps to 39fps is worth it, from 36fps to 39fps is not) My guess is more shaders = better and DX11 doesn't matter... should a 285GTX keep framerates at 60fps (assuming it's a PhysX bottleneck) no matter what happens in all current top end games? I'm guessing in the future a 480gtx or 580gtx will at some point, due to 2x as many shaders, double performance further?

If something manages 30fps on one card, doubling the shaders (at same core speed) I assume should mean 60fps, and 192 shaders on a 260gtx is probably the same framerate as 192 shaders on a 550ti, is that a fair guess? Or do different cards have different 'efficiency' in running PhysX?
Whats the most demanding currently, Mafia 2? Batman Arkham Asylum? Cryostasis?



My understanding is that higher resolutions are more demanding of PhysX (though I dont understand why, why should display res matter? Or does it just use up the shaders normally no longer shared with physx processing?), I could understand how increased FOV from 3 monitors could slow it if more is visible on screen and potentially animated there.



Since AFAIK even multi SLI graphics systems can be brought to their knees by certain games (Metro 2033) in 3d surround i'm assuming dedicated PhysX may improve performance as well.



But i'm curious where the radical improvements level off, i'm not looking for a 4% improvement spending twice as much, just to eliminate bottlenecks. (ie jumping from 13fps to 39fps is worth it, from 36fps to 39fps is not) My guess is more shaders = better and DX11 doesn't matter... should a 285GTX keep framerates at 60fps (assuming it's a PhysX bottleneck) no matter what happens in all current top end games? I'm guessing in the future a 480gtx or 580gtx will at some point, due to 2x as many shaders, double performance further?



If something manages 30fps on one card, doubling the shaders (at same core speed) I assume should mean 60fps, and 192 shaders on a 260gtx is probably the same framerate as 192 shaders on a 550ti, is that a fair guess? Or do different cards have different 'efficiency' in running PhysX?

#1
Posted 03/14/2012 07:30 PM   
The most demanding PhysX game I ever played is Alice:Madness Returns with physX on high.
It brings my single GTX 580 fps down to the teens while playing in 3D.
Had to turn physX down to medium. Even medium physX is sometimes too much (pepper grinder), but no way will I play with physX on low.
http://www.youtube.com/watch?v=m1sI8ovRIFY

mafia 2 benchmark is very physX heavy but usually gameplay is less physX heavy.
batman:AA is physX heavy in Scarecrow's world but rest of the game is less physX heavy.
Have not played batman:AC yet.
The most demanding PhysX game I ever played is Alice:Madness Returns with physX on high.

It brings my single GTX 580 fps down to the teens while playing in 3D.

Had to turn physX down to medium. Even medium physX is sometimes too much (pepper grinder), but no way will I play with physX on low.

/>


mafia 2 benchmark is very physX heavy but usually gameplay is less physX heavy.

batman:AA is physX heavy in Scarecrow's world but rest of the game is less physX heavy.

Have not played batman:AC yet.

[Acer GD245HQ - 1920x1080 120Hz] [Nvidia 3D Vision]
[Asus P7H55 - core i3 530 @4.15GHz ~1.39V and Zalman CNPS5X] [Transcend 2x2GB DDR3]
[Palit GTX 260 @ 695MHz (580 died)] [Intel SSD 330 - 240GB]
[Creative Titanium HD + Beyerdynamic DT 880 (250ohm) headphones] [Windows 7 64bit]

#2
Posted 03/15/2012 01:54 PM   
[quote name='Partol' date='15 March 2012 - 01:54 PM' timestamp='1331819696' post='1383268']
The most demanding PhysX game I ever played is Alice:Madness Returns with physX on high.
It brings my single GTX 580 fps down to the teens while playing in 3D.
Had to turn physX down to medium. Even medium physX is sometimes too much (pepper grinder), but no way will I play with physX on low.
http://www.youtube.com/watch?v=m1sI8ovRIFY

mafia 2 benchmark is very physX heavy but usually gameplay is less physX heavy.
batman:AA is physX heavy in Scarecrow's world but rest of the game is less physX heavy.
Have not played batman:AC yet.
[/quote]

To me, the PhysX seemed less heavy in Arkham City than it was in Arkham Asylum. In fact, AA has seemed like the heaviest PhysX of any game I've played. I had absolutely no issues with Alice. It ran extremely well with everything maxed out.
[quote name='Partol' date='15 March 2012 - 01:54 PM' timestamp='1331819696' post='1383268']

The most demanding PhysX game I ever played is Alice:Madness Returns with physX on high.

It brings my single GTX 580 fps down to the teens while playing in 3D.

Had to turn physX down to medium. Even medium physX is sometimes too much (pepper grinder), but no way will I play with physX on low.

/>


mafia 2 benchmark is very physX heavy but usually gameplay is less physX heavy.

batman:AA is physX heavy in Scarecrow's world but rest of the game is less physX heavy.

Have not played batman:AC yet.





To me, the PhysX seemed less heavy in Arkham City than it was in Arkham Asylum. In fact, AA has seemed like the heaviest PhysX of any game I've played. I had absolutely no issues with Alice. It ran extremely well with everything maxed out.

Core i5 4670K @ 4.3GHz with 4GHz uncore
Corsair H100 water cooler
8GB Patriot Viper DDR3 @ 2400MHz
Gigabyte Z87X-OC mobo
SLI eVGA GTX670 FTW (1227MHz boost 6800MHz VRAM)
Soundblaster Z with 5.1 speakers
112GB OCZ Vertex 3 SSD
240GB OCZ Vertex 4 SSD
Corsair Carbide 500R case (white)
Akasa PowerXtreme 1200W PSU
HP ZR2740W 2560 x 1440 27" IPS panel
Windows 7 Pro 64bit

#3
Posted 03/19/2012 02:16 PM   
I just beat Arkham City and I'm still playing it to solve all the Riddler challenges. I can say that the PhysX usage is pretty heavy, as I have GTX 480 installed as a PPU (just for this game, LOL) and it gets pegged at around 20%-50% usage on average. I'd say it's at least equal to the first game in regards to PhysX effects.

Arkham City is definitely more taxing than the first game if you have DX11 and PhysX enabled, although most of the performance issues in AC can be attributed to the unoptimized DX11 features.

The boss fight with Joker at the steel mill (in AC) brings my PC to it's knees (10-15 fps), so I have to disable PhysX or DX11 to get through this particular fight. I think this issue is a bug in the game though, because my primary GPU's usage levels are very low during this fight when I have PhysX enabled (30% on average). From what I've heard, even GTX 580 SLI setups cannot achieve a playable framerate during this fight if everything's cranked.
I just beat Arkham City and I'm still playing it to solve all the Riddler challenges. I can say that the PhysX usage is pretty heavy, as I have GTX 480 installed as a PPU (just for this game, LOL) and it gets pegged at around 20%-50% usage on average. I'd say it's at least equal to the first game in regards to PhysX effects.



Arkham City is definitely more taxing than the first game if you have DX11 and PhysX enabled, although most of the performance issues in AC can be attributed to the unoptimized DX11 features.



The boss fight with Joker at the steel mill (in AC) brings my PC to it's knees (10-15 fps), so I have to disable PhysX or DX11 to get through this particular fight. I think this issue is a bug in the game though, because my primary GPU's usage levels are very low during this fight when I have PhysX enabled (30% on average). From what I've heard, even GTX 580 SLI setups cannot achieve a playable framerate during this fight if everything's cranked.

EVGA E758 A1 X58 // Core i7 920@4Ghz // OCZ Platinum DDR3 1600 // EVGA GTX 670 SLI // Seasonic X Series Gold 1050w // Corsair 800D // Dual Dell Ultrasharp U2410 displays // Dell Ultrasharp 2408WFP

#4
Posted 03/19/2012 02:37 PM   
Cryostasis is the most demanding PhysX game I've played as of yet. Of course I played it with my old rig which was a Q6600 and GTX260's, but any time water came in to play my machine would drop to it's knees. It would be interesting to play it now and see how my current setup would handle it. I'm sure it would be much smoother, but the performance hit would still be crazy.

I'm pretty sure I talked about it quite a bit in my review. Might be worth a read.

Also you should dig up ChrisRay's Physx write up which should be stickied around here somewhere. Loads of great information in there that is still valid. The rule of thumb is that if you are going to be running a dedicated PPU it should more then 50% the shaders (cuda cores) then your main rendering card. The more the better to avoid any type of bottleneck. From my personal experience though a dedicated PPU isn't exactly the best way to go. If you are running SLI you can always test running with SLI disabled and the second card as a PPU Vs running strictly with SLI enabled and PhysX being processed at the same time.
Cryostasis is the most demanding PhysX game I've played as of yet. Of course I played it with my old rig which was a Q6600 and GTX260's, but any time water came in to play my machine would drop to it's knees. It would be interesting to play it now and see how my current setup would handle it. I'm sure it would be much smoother, but the performance hit would still be crazy.



I'm pretty sure I talked about it quite a bit in my review. Might be worth a read.



Also you should dig up ChrisRay's Physx write up which should be stickied around here somewhere. Loads of great information in there that is still valid. The rule of thumb is that if you are going to be running a dedicated PPU it should more then 50% the shaders (cuda cores) then your main rendering card. The more the better to avoid any type of bottleneck. From my personal experience though a dedicated PPU isn't exactly the best way to go. If you are running SLI you can always test running with SLI disabled and the second card as a PPU Vs running strictly with SLI enabled and PhysX being processed at the same time.

Image

#5
Posted 03/22/2012 01:36 AM   
Is there any reason 50% is a magic number? My guess is it would all depend upon the game and future games will need more... or is it just more that any game that stresses more than 50% of your main graphics shaders will be worse off or only marginally improved with a dedicated card... I suppose the math makes a kind of sense. I'm wondering why they don't make it so PhysX could be run on multiple cards though, including a full time dedicated but accessing other graphics cards when necessary?

[s]So does an SLI rig run PhysX on all cards? Should the dedicated PhysX for an SLI rig literally match the primary cards to have enough shaders?[/s]
INSERT: NVM, I found the article I assume you wanted me to find. http://forums.nvidia.com/index.php?showtopic=150597 However it doesn't seem to cover 4 way dedicated cards (not dual 590 type) which i'm still curious about.

It's mostly wondering which GPU I should design around as a dedicated PhysX card that will never drop below 60fps, ever, in any game that has it (as long as my video card keeps up with video at 60fps of course) without ridiculously overpaying. (a 580gtx seems overkill but will anything less let me play all the above mentioned games at 60fps in the worst parts?) I may not have the ideal right now (i'll have a 260gtx for PhysX during the next upgrade to whatever graphics i'll use, like 580gtx SLI hopefully) but hope to in future. Though i'm guessing the future will be even worse in the Kepler era so more will always be better.


Is it a rule of thumb that PhysX scales linearily? Like if I get 30fps with 100 shaders on a dedicated card, it should get 60fps with 200 shaders on a dedicated card of similar speed? Ie - if I test a game and get 20fps, will going up 20% speed and a card with 2.5x the shaders get me 60fps...
Is there any reason 50% is a magic number? My guess is it would all depend upon the game and future games will need more... or is it just more that any game that stresses more than 50% of your main graphics shaders will be worse off or only marginally improved with a dedicated card... I suppose the math makes a kind of sense. I'm wondering why they don't make it so PhysX could be run on multiple cards though, including a full time dedicated but accessing other graphics cards when necessary?



So does an SLI rig run PhysX on all cards? Should the dedicated PhysX for an SLI rig literally match the primary cards to have enough shaders?

INSERT: NVM, I found the article I assume you wanted me to find. http://forums.nvidia.com/index.php?showtopic=150597 However it doesn't seem to cover 4 way dedicated cards (not dual 590 type) which i'm still curious about.



It's mostly wondering which GPU I should design around as a dedicated PhysX card that will never drop below 60fps, ever, in any game that has it (as long as my video card keeps up with video at 60fps of course) without ridiculously overpaying. (a 580gtx seems overkill but will anything less let me play all the above mentioned games at 60fps in the worst parts?) I may not have the ideal right now (i'll have a 260gtx for PhysX during the next upgrade to whatever graphics i'll use, like 580gtx SLI hopefully) but hope to in future. Though i'm guessing the future will be even worse in the Kepler era so more will always be better.





Is it a rule of thumb that PhysX scales linearily? Like if I get 30fps with 100 shaders on a dedicated card, it should get 60fps with 200 shaders on a dedicated card of similar speed? Ie - if I test a game and get 20fps, will going up 20% speed and a card with 2.5x the shaders get me 60fps...

#6
Posted 03/23/2012 12:51 AM   
Scroll To Top