Username:
Password:
Remember Me?
   Lost your password?
Search

Physics acceleration via the GPU
Continuation: Mirror's Edge: Physx effects benchmark review


Mirror's Edge - Physx effects reviewed: Benchmark sequence
Mirror's Edge: 3.0 GHz Quad with additional effects without GPU Physx
 
Mirror's Edge: 3.0 GHz Quad with additional effects without GPU Physx [Source: view picture gallery]
To check how the additional physics effects make themselves felt, we use the checkpoint B of the level "Heat”. On ground floor we slide through a corridor, while the SWAT team chasing us blasts nine big glass panes. This is a worst-case scenario.

Mirror's Edge: 3.0 GHz Quad with additional effects with GPU Physx
 
Mirror's Edge: 3.0 GHz Quad with additional effects with GPU Physx [Source: view picture gallery]
Mirror's Edge - Physx effects reviewed: Benchmarks with and without additional physics
The fps are directly related to the additional effects. If you have a Geforce 9800 GTX+ or a better card, you should use GPU Physx and the additional effects. This way Mirror's Edge is running noticeably slower, but still quite smooth with a GTX 260-216 for example. If you also use a quad-core instead of a dual-core processor at the same clock speed, you also get about 20 percent more performance.

With a Radeon HS 4870/1G you should better do without the additional effects: As soon as the window panes shatter, the performance is reduced and the fps drop to single digit values - this way Mirror's Edge is unplayable even on a 4.0 GHz quad-core processor. But without the additional effects the DICE title runs smooth although the Radeon doesn't benefit as much from a faster CPU as the Nvidia cards do.

During our benchmarks with the review code, the Radeons were faster than the Geforce cards even with activated physics effects. The reason is simple: The benchmark sequence we used for those tests, didn't use a lot of physics effects.







Mirror's Edge - Physx effects reviewed: Benchmarks with dedicate Geforces as Physx accelerators
Many users ask themselves if a second Geforce as a Physx accelerator increases the performance. Based on the results of a GTX 260-216 with activated GPU Physx it becomes clear: A weak card like the 8400 GS is absolutely overstrained by the additional effects and slows the system down. But even 8600 GT can boost the system by about 15 percent, a 9600 Gt by about 30 percent. If you even use a 9800 GTX+ as Physx card, you can get about 5 percent more.




Mirror's Edge - Physx effects reviewed: Conclusion
The PC exclusive, additional physics effects make Mirror's Edge unique and the game benefits in matters of visualization as well as atmosphere. But practically this statement applies for owner of a Geforce 8 or better graphics card. Mirror's Edge becomes unplayable slow if the physics have to be calculated by the processor. But the additional effects, which can be smoothly displayed with GPU Physx only, are not essential for the gaming experience in Mirror's Edge. If this will change in future title, remains to be seen.

Picture gallery  (enlarge to view source)



--
Author: Marc Sauter (Jan 22, 2009)


You’re looking for more information related to AMD/ATI products like Phenom, Athlon and Radeon? Then please check our special AMD landing page with news, reviews and guides.




Advertisement

Comments (7)

Read all comments here!
ruyven_macaran Re: Mirror's Edge: Physx effects benchmark review
Super Moderator
15.05.2009 02:50
Considering wider and wider Memory-Interfaces (up to two times 448Bit with the new GTX295 single PCB :wow: ): Never.
Personally I keep an eye on the case manufacturers returning to full-length-compatible cases, as it was standard a decade ago. (but somehow todays cases always have to put some 5,25" bays in front of this, so they get to deep for my setup while wasting free space in height)
ruyven_macaran Re: Mirror's Edge: Physx effects benchmark review
Super Moderator
12.05.2009 12:04
You should wait for the GTX260 (might also try a GTS250, which isn't that much worse and a huge improvement from a 8600GTS.), as for example Vista does not allow different graphic-card-drivers to be active, so you can only run several cards of the same manufacturer.
Also keep in mind, that a 8600+4870 might actually be slower, then a GTX260 by herself, as can been seen here:
www.pcgameshardware.de/ai...
(don't think we had an english edition of this article, but the benchmarks should be unilingual )
As you can see: 4870+PPU (this actually works on modern operating systems, as the ageia physX card doesn't need any graphic driver. but getting new drivers at all is also getting hard, so not realy an option) are hardly any better then the GTX260 by herself, while the gtx260+ppu pairing is definetly faster then gtx260+8600gt. take this difference and subtract it from the 4870+ppu results and this combination marks the lowest end of the list.
pcghx_Kristoffer Re: Mirror's Edge: Physx effects benchmark review
Administrator
10.02.2009 16:02
Quote: (Originally Posted by Unregistered - Giz)
Hi Marc,
Am I reading your graph on page 2 correctly and there is NEGATIVE performance by using a 8400GS as a Physx processor (as compared to not having an additional card, and using the 260 alone?) It looks like you get HALF the performance if you do this!

Your results:
G260-216 GPU, 9600GT PPU: 66.6
G260-216 GPU, 8600GT PPU: 60.1
G260-216 GPU AND PPU (single card): 52.3
***G260-216 GPU, 8400GS PPU: 24.7***

Thanks,
giz


Yes that's right. The 8400 is so overstressed by the calculations, that it slows the whole system down.

Copyright © 2010 by Computec Media AG      About/Imprint  •  Terms/Conditions