Context switching between CUDA and Graphics should only be a matter of microseconds as you mentioned.
In one frame, the context switching only needs to happen once. Do the CUDA stuff first and then render your results. This is a lot less than transferring your results over the PCIe bus though.
well yeah, anything can happen in a matter of microseconds
and switching over to CUDA at the start of the frame doesn't incur any performance penalty? or there is simply no switching involved?
since pcie transfers are involved with a dedicated card, it might be illuminating to know how the frame times compare and the overall smoothness of gameplay.
I think that there is an overlap between memory transfers and computations.
Looks fantastic, will enjoy seeing it in games 5+ years from now.
Thanks for the link. I found one more review which is interesting : http://www.youtube.com/watch?v=s5IxYyeppU0
Metro 2033 benchmark results :
GTX 680 + GTX 560Ti (PhysX with PCIe x16) : 61 fps
GTX 660 + GTX 560Ti (PhysX with PCIe x8) : 54 fps
GTX 680 (Render + PhysX) : 52 fps
PCIe seems to be the main bottleneck in this case as performance scales more with bandwidth changes. Since there is a performance improvement by adding a dedicated device I think that there is an overlap between memory transfers and computations. This will be different across different applications though.
Ahemm. Reaction time.
Yes that is again one switch. It should be around 10-20 microseconds I have heard. Modern games do context switch between Compute Shader and Graphics so it's not a big deal.
This should be FCATed I concur.
There is a lot of stuff ihv's can do to hide latency
No the performance is from running a GTX 680 over a GTX 660....
You clearly lack any practical know how on this subject so best to leave it at that as everything you've said is completely wrong.
PhysX FleX is a particle based simulation technique for real-time visual effects. So far we showed examples for rigid body stacking, particle piles, soft bodies and fluids. In this video we'll be showing some of the new advancements within PhysX FleX.