Message boards :
Generalized Fermat Prime Search :
GFN=15 (32768) Switching to OCL2 soon
Posted 1689 days ago by Captiosus
That's not expected.
Those errors (i.e., repeated "MaxErr exceeded") you're seeing are typical of the "overclocking syndrome" we've seen all along on Nvidia GPUs. If that GPU is overclocked (even if it's merely "factory overclocking") you could try reducing to stock speeds to see if the problems go away. Or, since OCL3 seemed to be more reliable for you, run 65536 instead. It will be running OCL3 for a while.
So, OCL2 is hyper-sensitive to overclocking, even slightly? I'm asking because I've never been able to run OCL2 without it freezing about 15-25% into the workunit. Once it freezes, compute work for those work units on either of my GPUs stops, and the application simply sits there twiddling its thumbs, not releasing the GPU and failing the workunit. It does this with both GFN-15 and GFN-17-Mega units.
I get the following message in the event log when it happens though:
12/20/2015 12:04:20 PM | PrimeGrid | Task genefer15_32768_2352665_3 exited with zero status but no 'finished' file
12/20/2015 12:04:20 PM | PrimeGrid | If this happens repeatedly you may need to reset the project.
Resetting the project did not help.
I'm also not sure if it has anything to do with the fact I have mixed generation GPUs in my system (GTX 660 for the main card, GT 440 OEM as the secondary), and there's some funky conflict between the two going on. I know both cards have the required compute capability, as they are on the list (GTX 660 is CC 3.0, GT 440 OEM, being a variant of the GTS450 (same silicon), is CC 2.1), so I'm pretty sure thats not the cause.