Hi,

since yesterday I have loads of computing errors on my cuda WU's. Lot's of them reported a new error, I did not find in the board yet.

Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 915 : unknown error.
Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 920 : unknown error.
Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 920 : unknown error.
Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 925 : unknown error.
Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 925 : unknown error.
Cuda error 'cudaMemcpy(&flags, dev_find_pulse_flag, sizeof(*dev_find_pulse_flag), cudaMemcpyDeviceToHost)' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_

Some examples:
Task 1347461231
Task 1347461186
Task 1347461176
Task 1347461170

System:
Win Vista 32 SP2, BOINC client version 6.6.36 for windows_intelx86
Lunatics Win32 v0.2 CudaV12
CUDA device: GeForce 8500 GT (driver version 18585, compute capability 1.1, 512MB, est. 5GFLOPS)

Looks like my wingman have no problems. Does anyone have an idea of the cause?

Ralf

---

Free-DC | BOINCstats
89642h S@h classic CPU time

Hi,

just noticed "Questions and answers:Cuda" may be a better place for this post.
Can someone please move it?

Thank you,
Ralf

---

Free-DC | BOINCstats
89642h S@h classic CPU time

Hmm.. strange..

If you used the whole files of the Lunatics Installer and you make errors.. the only thing I could say.. take the new nVIDIA_driver_190.x and look if they continue.


BTW.
If you use the new CUDA_V2.3_.dll's with nVIDIA_driver_190.x you will get ~ 30 % speed up.

CUDA_V2.3:
[http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=208]-[2x .dll's]

---

Ahh BTW. ..

Maybe better to ask at the opt. crew site: [http://lunatics.kwsn.net]

---

It's look like you have the same errors like Fred W had:

[http://setiathome.berkeley.edu/forum_thread.php?id=54599]

IIRC, this errors came because his graphic card RAM was damaged.

---

Ralf

The first look is you are using the unified 0.2 with 185.xx drivers, You should be at the 190.xx drivers.

The next part would be have you Rebooted since this started. I would say there is a chance that something is stuck in video memory.

Lastly Number Crunching is the place to talk about problems with Optimized Cuda issues.

Regards

Hi,

since yesterday I have loads of computing errors on my cuda WU's. Lot's of them reported a new error, I did not find in the board yet.

Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 915 : unknown error.
Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 920 : unknown error.
Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 920 : unknown error.
Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 925 : unknown error.
Cuda error 'find_pulse_kernel' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_exp/client/cuda/cudaAcc_pulsefind.cu' in line 925 : unknown error.
Cuda error 'cudaMemcpy(&flags, dev_find_pulse_flag, sizeof(*dev_find_pulse_flag), cudaMemcpyDeviceToHost)' in file 'd:/BoincSeti_Prog/sinbad_repositories/LunaticsUnited/SETI_CUDA_MB_

Some examples:
Task 1347461231
Task 1347461186
Task 1347461176
Task 1347461170

System:
Win Vista 32 SP2, BOINC client version 6.6.36 for windows_intelx86
Lunatics Win32 v0.2 CudaV12
CUDA device: GeForce 8500 GT (driver version 18585, compute capability 1.1, 512MB, est. 5GFLOPS)

Looks like my wingman have no problems. Does anyone have an idea of the cause?

Ralf

---

Please consider a Donation to the Seti Project.

@Pappa

Update to 190 is done.

As I became aware of the error when Cuda queue already was empty, I could not check if rebooting (#1 problem solving action ;-) would have cleared the problem. But only 50% of the WU ended with "Error while computing", the rest ran fine. I will have a closer look to the system over the weekend.

Thanks, good to know beeing in the right board.

@Sutaru Tsureku

Thinking it is an actual problem I set search limit to 30 days. So I did not find Fred W's post. Following your advice, I already downloaded the Memory Tester for GPU he mentioned in his posting and will start some cycles when the problem reappears.
Currently I got two...and now 42 new Cuda WU's and the first one is running normal so far.

Regards,
Ralf

---

Free-DC | BOINCstats
89642h S@h classic CPU time

Hi,

1.5 days of error free Cuda crunching after upgrading and rebooting now.
Going through the board, I think 'find_pulse_kernel' is only an effect, caused by "cudaMemcpy" error. There are lots of postings of graphic card memory errors (defect memory, out of memory, memory fragmentation,...) in combination with different WU computing errors.
Many thanks to Sutaru Tsureku and Pappa for your help.


Regards,
Ralf

---

Free-DC | BOINCstats
89642h S@h classic CPU time