Dissimilar GPUs for PhysX and PCIe limitations?

Daniel Watstein

Honorable
Jun 9, 2013
1
0
10,510
I've read a few posts as well as a TH review back in '09 or '10 about 16x vs 8x vs 4x PCIe 2.0 performance but the conclusions were that you'd only notice with a high-end card. Does anyone have an idea of the bandwidth usage of a gtx 680 vs PCIe 3.0 16x/8x? I want to stick a 460 in a new build as a dedicated physX card but I want to make sure I'm not choking off the 680's bandwidth by switching to 8x/8x. Any experience?
 
You shouldn't notice any difference by cutting down the bandwidth, but it is a complex issue. It is possible the bandwidth is only halved when installing two cards in SLI, rather than adding another for dedicated PhysX, which does not directly involve the graphics rendering subsystem (similar to adding a sound card). You could also consider installing the 460 in a x4 PCIe slot without any loss in its PhysX performance versus x8 or x16.

What PhysX games do you intend to play?