Physx Benchmarks comparing 480GTX to 650GTX in Batman AC.

Hi,

Yesterday I was bored and picked up a PNY GTX650 card and benchmarked it against one of my GTX480's in Batman AC. The two 480 GTXs are in 8X slots while the 650GTX is in a 4X slot. The system is a P8Z68 Deluxe 2600K build. I ran it in 1920x1200 resolution testing in low detail mode vs high quality mode with all the eye candy and AA turned up all the way. Pretty much shows that the 480GTX creams the 650 GTX in Physx performance when the gpus don't have much work to do but when the heat is on the GTX650 is helpful.

Here are my results detailing the low score - high score - average scores. I ran the tests a couple times and averaged the results.

Lowest Quality Batman AC

Low High Average
40.5 127.5 81.0 480GTX w/ded 480 Physx

29.0 120.5 70.5 480GTX w/ded 650GTX Physx

31.5 121.0 70.5 480SLI w/ded 650 Physx

40.0 125.0 77.0 480SLI shared Physx

High Quality Max Settings

9.5 76.0 55.5 480SLI shared Physx

10.5 93.5 58.5 480SLI w/ded 650 Physx

18.5 48.0 39.0 480GTX w/ded 650GTX Physx

8.0 52.0 39.5 480GTX w/ded 480 Physx

So as you can see with the highest quality game settings if I didn't have the 650GTX and was using the 480s in SLI and Physx shared the highest score suffers a bit though the average being just 2-3 % better. I'm hoping that would translate in higher fps and smoother performance.

Just wanted to share those numbers. Maybe save someone $100. Its interesting though in the lowest quality settings adding a 650GTX in there hurts performance by like 10% compared to just having 480s in SLI running Physx as well.

Thanks.

 

Nice work, i find the best way to run physx is with a single card setup. But i really appreciate those numbers, dude.

 

Thanks

I just was looking for a reason to justify the purchase. Thanks.

 

Just curious which drivers where you using? Recently had to RMA a Msi GTX480 lightning in my spare rig, due to it not reading temps and fan speed any more in Afterburner. Decided to pick up a GTX660Ti in the mean time. Also, the rig is using a GTX460 256bit 1Gb card for Physx.

Anyway, installed the new card and 310.90 drivers, ran FluidMark to make sure all was okay,and got horrible performance. I was only using 30% of the GTX460, 99% of the GTX660ti, and getting horrible frame rates. Tried everything, from settings to reinstall of the drivers, no go.

Just for giggles I went back to 306.97 drivers I had before. Guess what? Now the GTX460 was being use between 60-70%, the GTX660ti 80-90%, and frame rate double!

 

I also did similar tests involving 3xGTX470 in SLI, then I used 2 in sli and one dedicated for physx, performance was similar. Then I decided to get a GT640 for 3SLI+ physx and it helped a little bit but not justifying the money.
Now, I am testing 3xGTX660ti in SLI and the performance is almost as 2GTX660ti SLI+ 1x470 for physx. For some reason the automatic physx on 3way sli is worse then using 2sli_+_physx dedicated

 

dual card setups are more optimized than triple card setups, that'll be why, i'ts due to the not so optimal triple card scaling.

 

I just bought Batman AC and now idk did they really fix the game or did newest driver do some magic or is it my cpu or all together.. It still kinda stuttered by changing transitions though, but ok..


Physx - high; rest highest dx11 with fxaa - high

cpu and gpu OC in profile.


min 43fps
max 113fps
avg 81fps

Also by that freezer gun part - gpu usage max, with ~ 66-80fps.




Physx - normal, rest highest dx11 with fxaa


min 51fps
max 121fps
avg 90fps




physx - off, rest highest dx11 with fxaa-high


min 60fps
max 137fps
avg 112fps


EDIT: that lowest fps appears to be happening by scene transitions (by all 3 tests), so its more of a str3aming bottleneck (game installed on sata2 std HDD).


Vs what nvidia reviewed back then


http://www.geforce.com/whats-new/gu...ty-graphics-breakdown-and-performance-guide#3

or any other site
http://physxinfo.com/news/6922/batman-arkham-city-physx-benchmarks-roundup/

Even with addon gpu its slower then my single 570gtx

 

Sorry for the off-topic question but did they end up fixing DX11 yet? It was a stuttering mess when released.

 

I didnt play it yet, but whole dx11 benchmark looked a lot smoother then what it was at first (tested leaked version patched to 1.2).

Now only 2-3 hickups at the beginning by few scenes, before it was during the scene and especially by changing scenes.

 

Damn Nvidia cards love this game, on my 7970 I get massive drops to 28fps with MAX settings (physX off of course) and some stutters. I also get CPU usage of 100% on a few threads so it seems its CPU bottlenecked.