Hello everyone. I got tired of reading all the forum posts where someone asks if a dedicated physx card will be worth having, and then saying they would give it a try without coming back to let us know how it went, so I did it. I had to modify the drivers of course, but before I found something else to do with my 580, I put it in along with my 680. GTX 680 gave me avg 54 fps in Arkham City, but jumped to 71 with the 580 handling the physx. Unfortunately, the Metro 2033 benchmark always gives me 33 fps no matter what settings I use, so if anyone knows a good way to bench that game let me know. I will do some more testing to make sure that score wasn't a fluke. I will also try with my old GTX 275 instead of the 580.
The current driver for the 680 doesn't work well with Metro 2033, and it's really subtle with the use of Physx. Have any other titles? Alice Madness Returns had a lot of effects, and overall it's a pretty great game
I think I got Alice, but I never installed it. Even on my 580, metro 2033 gave me 33 fps with or without physx, with or without DOF, AAA or 4x MSAA. I think the benchmark that came with the game sucks.
Obviously a dedicated physx card helps, a lot, specially in batman arkham city, but the question is, can you live with the extra heat and noise? If yes, then you're good to go. If you don't have a spare card and want to buy a new one, I'd suggest to get a cheap low-end card like GT 440. It can handle physx just fine.
Can it though? The GTX 680 has a 256-bit bit bus, more shaders and higher clocks than a GT 440 so is the lower-end card likely to bottleneck PhysX more than if you used the GTX 680? That's what I've always wondered. You'd think that offloading PhysX to another card would allow the GTX 680 to devote 100% of its resources to rendering graphics but benchmarks I've read suggest that isn't the case as older hardware tends to be less efficient/less optimised. PhysX, unfortunately, is rarely tested in graphics cards reviews due to the AMD cards not being able to handle it in hardware and it's even rarer to find a review where PhysX is tested on dedicated cards. As such it's confusing knowing whether it is actually worth buying or using a separate PhysX card. My own feeling is that there are not enough PhysX games to justify buying a dedicated card but if you have an older card lying around wasted then it might be worth it. I have a GTX 280 and a GTX 580 but with me only having an 850 W PSU I'm not sure installing one of those for PhysX testing would be a good idea. I guess you always underclock them to reduce power consumption/heat as either would surely be plenty for PhysX alone?
give it a try man,on my old i7920 rig i used a 750watt psu an ran a 480 an a 260 for physx so your psu will be fine
Well, I'm off work next week and my PC is long overdue for a good clean inside so I may fit the GTX 280 if NVIDIA release a new driver that supports all their cards. Currently, the 301.10 drivers I'm using will not recognise my GTX 580 nevermind my old GTX 280.
I'm running a gtx680 with a gtx280 EVGA FTW edition for physx with no problems using the 301.10 drivers with a modded inf. Works without an issue.
No doubt. When I got my 580, I used my 285 for PhysX for quite a while on my Oc'd Phenom II - all on my PC P&C Silencer 750. There's no doubt that system drew more power than my current one. I'm keeping my 580 for my other rig and will also use it for PhysX for Batman very soon.
The Adv DOF is the big FPS eater in metro 2033, disabling that will give you your FPS back. To me DOF didn't add close to any thing in the game. So I disabled it when I played. Edit: I know some one gonna see my sig and try to figure out how i got metro 2033 to run on it. Of course I didn't play it with a 9800 GT. That would be silly.
Correct, I was wrong. Maybe the physx gpu support is the reason why Metro 2033 is such a hard title for nvidia cards compared to ATI.
Well, I got off work early today so I did some more testing. I am still only using the Batman game, but the Unreal engine gets a lot of use and most of the time uses Physx, so why not? I did five runs on the benchmark. GTX 680: Min - 13 - 21 - 21 - 19 - 16 Max - 101 - 101 - 102 - 104 - 98 Avg. - 53 - 53 - 55 - 54 - 53 GTX 680 + GTX 580: Min - 11 - 7 - 34 - 8 - 30 Max - 117 - 116 - 119 - 115 - 118 Avg. - 71 - 72 - 72 - 72 - 72 GTX 680 + GTX 275: Min - 7 - 14 - 15 - 19 - 19 Max - 116 - 115 - 114 - 116 - 113 Avg. - 64 - 65 - 63 - 66 - 63 These tests were done @ 1080p so I would think that at higher resolution having a dedicated Physx card would be quite nice as even while playing through the game I noticed that it was smoother (yes, I know that's subjective). HOWEVER... The cards I used are both bigger than the 680. Optimally, you would need a smaller card and I don't know what kinda of performance you would get. I am going to guess that a 560 would be faster than my 275, and I am tempted to go buy a 550 ti for testing. Bottom line for me is the price. I would not pay more than 100 USD for a dedicated physx card, especially when it's actually a video card. I don't see why NVIDIA doesn't just make a card strictly for physx when it would just run off their video card drivers anyway. How hard could it be to make a video card without the video?
Well I just did some test with batman and mafia 2 with only a 680 and 450 here. Seems a cheap 450 works good. http://forums.guru3d.com/showthread.php?t=361100 Results hope that helps 680 Mafia 2 no physx 60.5 FPS 680 Mafia 2 + 450 for physx 81.8 FPS 680 Batman Arkham City no physx MIn=7 Max=118 AVG=64 680 Batman Arkham City 450 physx MIn=12 Max=124 AVG=74
