Poking through my work units, it seems various NVIDIA cards are having trouble validating:
(I am running an AMD 5870 (atiOpenCL) on Win 7 x64, SP1)


Validates:
WU 24389 atiOpenCL + BRP3SSE
WU 24563 atiOpenCL + BRP3cuda32 (GTX 550 Ti)
WU 24633 atiOpenCL + BRP3cuda32OSX (Quadro 4000)
WU 24650 atiOpenCL + BRP3cuda32 (GTX 570)
WU 24702 atiOpenCL + BRP3cuda32 (GeForce 8700M GT)
WU 24715 atiOpenCL + BRP3cuda32 (GeForce 9600 GT)
WU 24735 atiOpenCL + BRP3cuda32 (GeForce GTX 580)
WU 24739 atiOpenCL + BRP3cuda32 (GeForce GTX 260)


Here's where things get weird...

Does Not Validate:
WU 20316 atiOpenCL + (GTX 570) + (GTX 295)
The GTX 295 did not validate with either the atiOpenCL or the GTX 570, but the atiOpenCL & GTX 570 validated together. Digging though the GTX 295 host history shows that they have a history of hit and miss validation errors with both NVIDIA & ATi.


WU 20274 atiOpenCL + (AMD 5800 [Linux]) + (AMD 6250/6310) + (GTX 580) + (AMD A8-3850 APU)
The Linux platform AMD 5800 & the AMD 6250/6310 errored out while computing. The A8-3850 APU did not validate with GTX 580, but did validate with my 5870. Digging through the AMD A8-3850 APU host history, I found two other workunits where the GTX 580 does not validate with it:
WU 21874
WU 24782


WU 23559 atiOpenCL + (AMD 5800 [Linux]) + (GTX 580) + (GTX 470)
The (BRP3cuda32) GTX 580 did not validate against the (BRP3cuda32nv270) GTX 470. The the unit was then sent to Bikeman (AMD 5800 [Linux]) and myself. The two atiOpenCL app's validated against each other.


WU 23524 atiOpenCL + (GTX 580) + (GTX 580)
I validated against a GTX 580, but the other GTX 580 did not validate.


As I finished typing this I noticed that the GTX 580 that kept invalidating was the same host http://albert.phys.uwm.edu/show_host_detail.php?hostid=1210 and he has 193 invalid results. But it is still strange that in WU 23559 that the GTX 470 did not validate against the AMD 5800 and it took another atiOpenCL to validate. That would point to the atiOpenCL app, however, I have validated against plenty of NVIDIA cards (shown above) previously.

---