Seti@Home optimized science apps and information
Optimized Seti@Home apps => Windows => GPU crunching => Topic started by: Pepi on 28 Apr 2010, 05:10:10 am
-
Before few days I have 37WU errors in one day. Since my comp is on open air, shaded from Sun, I think that temperature is not issue. Also errors are in 5 AM as in 2 PM. Day before and day after WU error are drooped to "normal" rate 1-2 per day.
clean install XP x64 SP2
190.38
optimized seti WU kill app ( reschedule whar to CPU)
nothing is OC
Errors from results ( first line)
Cuda error 'cudaAcc_CalcChirpData_kernel2' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_CalcChirpData.cu' in line 106 : unspecified launch failure
Cuda error 'cudaAcc_CalcChirpData_ke
Cuda error 'GaussFit_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_gaussfit.cu' in line 506 : unspecified launch failure.
Cuda error 'cufftExecC2C' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_fft.cu' in line 143 : unspecified launch failure.
Cuda error 'cudaAcc_CalcChirpData_kernel2' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_CalcChirpData.cu' in line 106 : unspecified launch failure.
Cuda error 'cufftExecC2C' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_fft.cu' in line 143 : unspecified launch failure.
Any suggestions?
-
upgrade to driver 197.45
-
That is not answer: because few days before crunch machine work like a charm. So I doubt that drivers is problem.
-
That is not answer: because few days before crunch machine work like a charm. So I doubt that drivers is problem.
It sortof could be, since there's been a lot of 'near VLAR' tasks in the last couple of days that get missed by the VlarKill app & possibly, to a lesser extent the rescheduler too, which could be crashing the driver. There are definite improvements in the latest driver that 'appear' to reduce the occurrences of driver crashes, that normally require a machine restart, and tend to exhibit exactly the symptoms you posted.
Eliminating that as a possibility 'properly' before going deeper would be a wise troubleshooting step, followed by checking for accumulated dust * then performing various card stress tests to make sure it hasn't taken Ill.
Jason
-
On one page I read: that some WU are full of "noise" and that can kill app. Today it gives only two errors, so I hope the worst is over :)