-C 1 Kepler Kernel

[TheoRettisch]

I just tested commit 567aea4b2d with Windows 8.

When i start Cudaminer with -C 1 on my 680 i get the following error:

GPU #0: cudaError 4 (unspecified launch failure) calling cudaStreamQuery(context_streams[stream][thr_id])' salsa_kernel.cu line 163

[Jamonek]

Same just happened to me as well.

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaDeviceSynchronize()' (salsa_kernel.cu line 520)

[2014-01-16 19:45:12] GPU #0:    0.00 khash/s with configuration K0x0
[2014-01-16 19:45:12] GPU #0: using launch configuration K0x0
[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaFree(d_V)' (salsa_kernel.cu line 690)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaStreamCreate(&tmp2)' (salsa_kernel.cu line 218)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaStreamCreate(&tmp2)' (salsa_kernel.cu line 219)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaEventCreateWithFlags(&tmp4, cudaEventDisableTiming)' (salsa_kernel.cu line 223)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaEventCreateWithFlags(&tmp4, cudaEventDisableTiming)' (salsa_kernel.cu line 224)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaEventRecord(context_serialize[1][thr_id])' (salsa_kernel.cu line 225)

[2014-01-16 19:45:12] GPU #0: cudaError 33 (invalid resource handle) calling 'cudaStreamWaitEvent(context_streams[stream][thr_id], context_serialize[(stream+1)&1][thr_id], 0)' (salsa_kernel.cu line 799)

Floating point exception (core dumped)

[cbuchner1]

This is is weird. I am running this code on two Ubuntu 12.04 machines right now, with no issues. There are Fermi, Kepler 3.0 and Kepler 3.5 devices involved.

I have had things crash on a Fermi device on Windows though.

2014/1/16 Jamone Kelly notifications@github.com

Same just happened to me as well.

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaDeviceSynchronize()' (salsa_kernel.cu line 520)

[2014-01-16 19:45:12] GPU #0: 0.00 khash/s with configuration K0x0 [2014-01-16 19:45:12] GPU #0: using launch configuration K0x0 [2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaFree(d_V)' (salsa_kernel.cu line 690)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaStreamCreate(&tmp2)' (salsa_kernel.cu line 218)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaStreamCreate(&tmp2)' (salsa_kernel.cu line 219)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaEventCreateWithFlags(&tmp4, cudaEventDisableTiming)' (salsa_kernel.cu line 223)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaEventCreateWithFlags(&tmp4, cudaEventDisableTiming)' (salsa_kernel.cu line 224)

[2014-01-16 19:45:12] GPU #0: cudaError 4 (unspecified launch failure) calling 'cudaEventRecord(context_serialize[1][thr_id])' (salsa_kernel.cu line 225)

[2014-01-16 19:45:12] GPU #0: cudaError 33 (invalid resource handle) calling 'cudaStreamWaitEvent(context_streams[stream][thr_id], context_serialize[(stream+1)&1][thr_id], 0)' (salsa_kernel.cu line 799)

Floating point exception (core dumped)

— Reply to this email directly or view it on GitHubhttps://github.com/cbuchner1/CudaMiner/issues/66#issuecomment-32528765 .

[Jamonek]

At this point I'm just having really bad luck with auto tune and will have to play around with the numbers to figure out the best configuration.

[ccomly]

a82ebc38f9c84404b882dcff3f2c38ab8fd88044 is the latest commit that doesnt crash when autotuning kepler.

on the plus side im at around 60kh/s faster than before on it. from 180 to 240-250+. On a 670 gtx with 4gb ram. my old 260 gtx has also been bumped up a few kh/s

[Shihira]

Similar problem occurs on my machine as well. my command is: cudaminer.exe -H 1 -i 0 -C 1 -m 1 -o stratum+tcp://dogepool.pw:3333 -u shihira.me -p * my card is: Geforce 210M a very old and slow card, but I did install the latest driver. then all kinds of “cudaError 2 (out of memory) ” came out. Please solve it, thank you.