Calling airsquirrels

[airsquirrels]

1.6 ms/iteration or so. Not too shabby at all...

[James Heinrich]

Originally Posted by Madpoo

James is doing a check with his CudaLucas as well to see if he encounters anything along the way. So far his residues have been matching mine.

I started my DC well before madpoo, but he got it done in 33h and I'm still only at 59M (on a GTX 580, 6.5ms/it), about 27h left. But so far residues are matching madpoo's.

[Madpoo]

Originally Posted by James Heinrich

I started my DC well before madpoo, but he got it done in 33h and I'm still only at 59M (on a GTX 580, 6.5ms/it), about 27h left. But so far residues are matching madpoo's.

I feel like if I weren't running my tests on servers with ECC, my error rate would be higher.

[James Heinrich]

Originally Posted by Madpoo

James is doing a check with his CudaLucas as well to see if he encounters anything along the way. So far his residues have been matching mine.

And (finally) done.

M( 73850519 )C, 0xa4a0717244d30cf6, 4096K, CUDALucas v2.05.1, estimated total time = 133:45:10

Triple-check complete.

[airsquirrels]

Looks like this issue is dead very dead. I will keep an eye out to see if it happens again before checking in results as I ramp up my CUDA work.

[chalsall]

Originally Posted by airsquirrels

Looks like this issue is dead very dead. I will keep an eye out to see if it happens again before checking in results as I ramp up my CUDA work.

Just to put on the table...

GPUs are much better than CPUs at TF'ing. We are still under a bit of a crunch to optimally feed the LL''ers and P-1'ers (but it's slowly getting better).

Everyone should do whatever they enjoy, of course. But LL'ing or P-1'ing with a GPU is currently suboptimal for the project as a whole.