Here is a new PhysX benchmark (available since around 2 weeks!) derived from an upcoming online game called Mars. This benchmark features NVIDIA APEX for all PhysX based effects (destruction, clothes, particles, fluids) and is based on the Unreal Engine 3.
You can download the benchmark from these links:
This Physx benchmark is convenient because it’s short: around 2 minutes. Then I quickly tested several PhysX configurations:
- ASUS Matrix GTX 580 @ 926MHz alone (3D + PhysX)
- ASUS Matrix GTX 580 @ 926MHz (3D) + EVGA GTX 580 @ 797MHz (PhysX)
- ASUS Matrix GTX 580 @ 926MHz (3D) + CPU PhysX (Core i7 @ 3.2GHz)
- ASUS Matrix GTX 580 @ 926MHz (3D) + MSI GTX 460 Cyclone @ 727MHz (PhysX)
- ASUS Matrix GTX 580 @ 926MHz (3D) + Zotac GT 240 @ 550MHz (PhysX)
Settings: 1920×1080 fullscreen, MSAA 1X, quality high (see screenshot below).
49.14 FPS – GTX 580 (3D) + GTX 580 (PhysX) |
| 45.26 FPS – GTX 580 (3D) + GTX 460 (PhysX) |
| 42.45 FPS – GTX 580 (3D + PhysX) |
| 30.12 FPS – GTX 580 (3D) + GT 240 (PhysX) |
| 15.75 FPS – GTX 580 (3D) + CPU Core i7 @ 3.2GHz (PhysX) |
PhysXInfo has also some results as well as additional information about this PhysX benchmark.
PhysX system: ASUS ROG MATRIX GTX 580 (3D) + EVGA GTX 580 SC (PhysX)
Settings:
I want see game like this unreal engine 3+ physx like alice
This test clearly shows how poorly implemented nVidia Physx really is. I mean a beefy core i7 almost grinds to a halt when trying to use it for the physics. Use the test but replace it with Havok. The visual results would be pretty much the same, and you can use any combination of GPU-CPU you like: Nvidia-Intel, Nvidia-AMD, AMD-Intel or AMD-AMD. With Physx you can only do Nvidia-Nvidia 😛
@HavokFan, 100% true !
@HavokFan I think there is no question that PhysX is optimized for NVIDIA hardware, yes, but keep in mind that clothes and especially fluid simulation are still pretty expensive and not exactly the sort of computation that a general purpose processor shines at. How much they add to the gameplay is arguably debatable, but the fact still remains that stream processors are much better at these sorts of computations.
Guys, I am sorry, but this PhysX thing has gotten ridiculous a long time now. I mean ok you get physics particles, some nice clothing animation etc. So f*cking what?
Have you guys seen what the engine of Red Fuction Armageddon does? Have you used the magnet gun? I mean come on who needs physx anyway!
That being said, the demo runs on software mode on my i7+570 setup. Don’t know why and I don’t see any physx option to enable inside the benchmark.
It gave around 15fps on my stock 860. Ugly demo to say the least.
@Psolord RFA is an excelent example of Havok in action. And I’m running it smoothly maxed out on an AMD-AMD combo. If the game had Physx in it, I’d be playing it at 5 fps, not 45.
Heh, tell me about it man.
I was against PhysX since day one. People were telling me back then “You are an ATI fanboy that’s why you say these things”.
Yeah well newsflash, I have GTX 570 SLI now. So much for an ATI fanboy! PhysX still sucks. Nothing changed.
To give a more illustrative example, watch this video I recorded the other day.
http://www.youtube.com/watch?v=z3RUKzqADmA&feature=player_detailpage#t=923s
From this exact time stamp onwards, physx just shows its weakness. The game dives to 30fps from a solid 60, when more than a handful of physx effects are on screen and that’s on two GTX 570s, with the second one dedicated to PhysX. I mean what is that?
In case anyone wonders if this is due to fraps recording, not it’s not. The same thing happens without it.
Also the game is artificially capped at 30fps, so to hide PhysX’s inadequacy. I had to unlock it through the config file, only to discover why this frame cap was there in the first place.
I mean come on, is that what we are to expect from TWIWMTBP? Seriously now? They are destroying our framerates, to promote their crap? Jesus Christ where has everyone’s vision has gone anyway?
Many people make the same mistake. HAVOK is cpu physics . NVIDIA physics is gpu physics. If you want to see the difference you have to have nvidia latest generation gtx4…5… physic card. The picture quality is much better and the games are more realistic with gpu physics. But you have to see !Also there is bullet gpu physics but is not so good as Nvidia, it is free and was used at movies ,but now CUDA make the difference !Finally dont look only the frames,look and how enjoy can be a game by physics.!
Havok can’t do such physics effects like PhysX. So STFU AMD deps.
so basically the physx part is a waste of money (5 fps for a £150 gpu (460) and bearly more fps for a pair of 580 ‘s)
Here is my result:
http://i53.tinypic.com/71ixc5.png
U jelly?
Run not 1080p for Dual core. Problem?
I run with GTX460 in SLI 49.24 fps in average – with the same settings as in this article. Pretty good compared to a GTX580 I would say.
http://i342.photobucket.com/albums/o405/Grnkjr0/Untitled-37.jpg
Hey guys check out intel’s Cloth Simulation
http://software.intel.com/en-us/articles/soa-cloth-simulation-with-256-bit-intel-advanced-vector-extensions-intel-avx/
Nothing against Havoc but if you really want an independent physics engine better use Bullet!
bullet is shit.
anyway this test is built off the unreal engine which uses single threaded physx
Have any of you really run this test with your rigs setup right?
i was reading one persons comment and from what he said i could understand that he has a driver problem as his options where not showing of physics SECOND every computer that i have seen setup by someone else never has the cpu parking turned off Seriously people Turn the thing off and you will never get lockups with your cpu running anything!.
especially on a i7 and thirdly there is two versions or two paths of I7 one is the desktop version most people have that sure its newer chip ect but still does not even come near the 1366 i7 with Tripple channel mem i will post a TEST of this benchmark running fine on i7 970 with 560TI with 6GB DRR3 Ram @ 1600Mhz as from what i seen of the screen shots it looks the normal to me.
OH and another thing i have never seen any Gigabyte product work well i noticed there was a Gigabyte board there.
To the guys who talked about Havok or other SW physics, yes in some cases i have seen games like Just Cause 2 that only used cuda and supported 3D run really well and looked like it has physx only to find out that it does not use PhysX at all but yet Renders some really good Water effects, debris ect.(Water Foam could have been better), my personal Opinion of the tech right now is Nvidia works on just about 100% of games and apps out there hence i use them ATI/AMD still needs to work out some Compatablity with programs and games but Hardware wise both companys have similar results in REAL Gaming or Graphics that a LCD or plasma or even the CRT ( real time ) can display, ATI has many procs in there gpu running at lower clock, Nvidia has fewer procs running at higher clock Results around about the same. Fluid Mark on the intel i7 970 in cpu mode can do more PhysX then in GPU mode due to particle Limitation i will post these results soon.