A Questions about dedicated PhysX card.

[A Really Hoopy Frood]

Okay so I have two questions really.

A) Does this program make use of a dedicated PhysX card? I recently got a new GPU and ran a test. Then I put my old GPU back in and made it a dedicated card. (Being different chipsets, I couldn't SLI the two together.) There seemed to be no improvement in the Physics scores. Oddly enough, there were some improvements graphically. I think that was just because there was also a driver update.

B) When running Ice Storm Unlimited it just shows a little box in the middle of the screen which whips through a bunch of pictures really quickly. Am I assuming correctly that it's still rendering and preforming the tests? Why then do I not get a fullscreen, movie-like experience that all other tests and modes give me?

[Astori Heichalot]

A. physics is not physX. physics score is determining your CPU (&RAM) performance. while physX is a technology exclusively built for nvidia card, which gives additional graphic experience in games. While not all games are promised to have different experience on physX, but few specific games do give significant graphic changes (Batman:AC, Alice:Madness Returns, Mafia 2 & Unreal Tournament)
so regardless which GPU you use, as long you have the same CPU & RAM setup, physics score will remain the same.

B. I have not give Ice Storm Unlimited a shot, but those are meant for mobile devices (android/apple) platform me thinks
so, lower resolution maybe

[A Really Hoopy Frood]

Originally posted by Astori Heichalot:

A. physics is not physX. physics score is determining your CPU (&RAM) performance. while physX is a technology exclusively built for nvidia card, which gives additional graphic experience in games. While not all games are promised to have different experience on physX, but few specific games do give significant graphic changes (Batman:AC, Alice:Madness Returns, Mafia 2 & Unreal Tournament)
so regardless which GPU you use, as long you have the same CPU & RAM setup, physics score will remain the same.

B. I have not give Ice Storm Unlimited a shot, but those are meant for mobile devices (android/apple) platform me thinks
so, lower resolution maybe

Hmm, that's what I thought. Honestly I probably don't even need ot use Ice Storm, but I'm a sucker for Spaceships. xD

I suppose I'll find a game in my library that uses PhysX and see how much better it works now. Thanks for the info.

[UnReal-4-Life]

Borderlands 2 you will notice try it. I can tell. When running 3 srceens you want all gpu and useing a old video card for a dedicated you can see the diff.

[Clo1Crew]

u can crossfire amd cards and they dont have to b the same chipset.for instance..hd 7850 with hd7870 works no problem...prefer it to gaming that nividia...u get more bang for your buck

[Clo1Crew]

Originally posted by Archangel:

Originally posted by Astori Heichalot:

A. physics is not physX. physics score is determining your CPU (&RAM) performance. while physX is a technology exclusively built for nvidia card, which gives additional graphic experience in games. While not all games are promised to have different experience on physX, but few specific games do give significant graphic changes (Batman:AC, Alice:Madness Returns, Mafia 2 & Unreal Tournament)
so regardless which GPU you use, as long you have the same CPU & RAM setup, physics score will remain the same.

B. I have not give Ice Storm Unlimited a shot, but those are meant for mobile devices (android/apple) platform me thinks
so, lower resolution maybe

Hmm, that's what I thought. Honestly I probably don't even need ot use Ice Storm, but I'm a sucker for Spaceships. xD

I suppose I'll find a game in my library that uses PhysX and see how much better it works now. Thanks for the info.

physX does work with amd cards its not just made for nividia,,,thats a mith..u can download it and it will run on amd cards

[Astori Heichalot]

Originally posted by Turtle Soup:

physX does work with amd cards its not just made for nividia,,,thats a mith..u can download it and it will run on amd cards

lol

[Clo1Crew]

Originally posted by Astori Heichalot:

Originally posted by Turtle Soup:

physX does work with amd cards its not just made for nividia,,,thats a mith..u can download it and it will run on amd cards

lol

laugh as much as u like im using it on mirrors edge with amd card

[Astori Heichalot]

Originally posted by Turtle Soup:

Originally posted by Astori Heichalot:

lol

laugh as much as u like im using it on mirrors edge with amd card

I just find it hilarious since you are so sure about something technical which is actually incorrect

ofcourse you can use any AMD card to play ANY physX games

but whether the physX is actually running by the card or the cpu, is a whole different matter, which you seemed not to understand that

remember, this is a fact, not opinion lulz

[Clo1Crew]

Originally posted by Astori Heichalot:

Originally posted by Turtle Soup:

laugh as much as u like im using it on mirrors edge with amd card

I just find it hilarious since you are so sure about something technical which is actually incorrect

ofcourse you can use any AMD card to play ANY physX games

but whether the physX is actually running by the card or the cpu, is a whole different matter, which you seemed not to understand that

remember, this is a fact, not opinion lulz

thanks for the advice

[Cakefish]

No, this benchmark is hardware neutral i.e. it doesn't discriminate between AMD and NVIDIA and give one an unfair advantage over the other. If it used hardware PhysX then that would produce bias towards NVIDIA GPUs in the results.

[UL_Jarnis]

3DMark does not use PhysX at all for anything.

I'm not sure what you are babbling about NVAPI? NVAPI is Nvidia's interface for querying various graphics card information bits. ATI has similar API. Both are used solely for determining hardware configuration (and if you disable SystemInfo scan from help tab, neither is called at all during the run)

[UL_Jarnis]

Combined test is separate from the graphics and physics tests and it pushes the system in a different way. It could easily score better than another card even if Graphics and Physics scores would initially suggest otherwise.

Why? I'll try to explain;

Graphics tests have very little CPU load so it purely maxes out the GPU. Physics test pushes the processor alone, GPU plays no part. Combined test stresses the CPU heavily and then puts a complex GPU-heavy scene on top of that. In Graphics test the GPU driver can eat up all the processor cycles in the world and score probably wouldn't change much. In Physics test the GPU driver is essentially idling (the scene is very simple, only the physics calculations on the CPU are very complex). Neither score is therefore substantially influenced by the efficiency (CPU usage) of the video driver.

Combined test score is definitely influenced by how much CPU is spent by the driver to render the very complex scene. Low CPU usage by driver would mean more CPU time left for the physics calculations and probably also slightly higher framerate with the complex graphics (driver not bogged down so badly by the physics simulation).

[UL_Jarnis]

Also if you want to ask detailed questions, ether open a support case at

http://support.futuremark.com

or send an email to info [at] futuremark.com

[UL_Jarnis]

Like I explained, the difference (which is, frankly, fairly minor here - less than 200 points in combined test which is bit over 2% in this case) is easily explained by the different CPU usage of AMD and NVIDIA drivers - effect that would not show in the other tests in any way.

Both cards have essentially identical graphics scores (40 point difference from 26000 points a fraction of a percent) but the Radeon-based system, even with a faster processor (300MHz higher overclock), can't quite match the NVIDIA system in combined test due to this.

And yes, Combined test was added especially to ferret out situations like this - differences in overall graphics performance when CPU is heavily stressed. Doesn't happen in every game, but can happen - good examples would probably be games like Company of Heroes 2 or even Battlefield 4 (multiplayer) which both tend to stress CPU very hard.

[UL_Jarnis]

Memory bandwidth, differences in architectures, far more complex shaders. You can think 3DMark 11 as "first generation simpler DX11" while 3DMark Fire Strike is "second generation very complex DX11 with GPU-compute fluid simulations" (the smoke effects).

You can find similar differences in game performance - simpler graphics -> GK104 is fine when compared to Tahiti. Very complex DX11 graphics -> suddenly Tahiti doing better.

For kinda similar reasons AMD cards tend to demolish NVIDIA cards in bitcoin mining... some specific types of math operations are just so much faster on AMD cards. Architectural differences showing performance differences in specific use cases. GPU performance comparison isn't always so simple... :)

GK104 vs GK110 could be for example due to GK104 having much lower double precision math performance - GK104 shader units have 8 FP64 CUDA cores while GK110 has 64... "minor" difference in floating point performance...