Hallo together,
on almost one of two WU's I get an error on CUDA Computing. It seams that there are tree kind of errors, some lines of the task result file looks like this:
=============================================================================================================================== Task Start and 1. kind of error
[20:47:20][8469][INFO ] Application startup - thank you for supporting Einstein@Home!
[20:47:20][8469][INFO ] Starting data processing...
[20:47:20][8469][INFO ] CUDA global memory status (initial GPU state, including context):
------> Used in total: 144 MB (1905 MB free / 2049 MB total) -> Used by this application (assuming a single GPU task): 0 MB
[20:47:20][8469][INFO ] Using CUDA device #0 "GeForce GT 630" (0 CUDA cores / 0.00 GFLOPS)
[20:47:20][8469][INFO ] Version of installed CUDA driver: 6000
[20:47:20][8469][INFO ] Version of CUDA driver API used: 3020
...
[00:12:40][8469][INFO ] Checkpoint committed!
[00:13:37][8469][ERROR] Error freeing CUDA HS device memory (error: 719)
[00:13:37][8469][ERROR] Demodulation failed (error: 1010)!
===============================================================================================================================
2. kind of error
[01:34:21][3169][INFO ] Checkpoint committed!
[01:34:57][3169][ERROR] Error during CUDA host->device HS thresholds data transfer (error: 719)
[01:34:57][3169][ERROR] Demodulation failed (error: 1007)!
===============================================================================================================================
3. kind of error
[21:34:13][2412][INFO ] Checkpoint committed!
[21:34:45][2412][ERROR] Error during CUDA device->host time series length transfer (error: 719)
[21:34:45][2412][ERROR] Demodulation failed (error: 1008)!
CUDA Computing for Milkiway and Asteroids works fine on the machine.
Has anyone an idea what went wrong?
Regards,
Klaus
Copyright © 2024 Einstein@Home. All rights reserved.
Demodulation failed Errors on CUDA - Ubuntu Linux with GeForce
)
CUDA error 719 seems to be caused by invalid memory access usually. With your version of BOINC not properly detecting NVIDIA memory on 64 bit systems I'd first try to rule that out as the source of failure. You should upgrade BOINC to 7.0.65 or later and see if that helps.
I'm still getting Errors 1007
)
I'm still getting Errors 1007 and 1008. I'm on Ubuntu Vivid (development release) with BOINC 7.4.23.
[11:55:25][3443][INFO ] Using CUDA device #0 "GeForce GTX 750 Ti" (0 CUDA cores / 0.00 GFLOPS)
[11:55:25][3443][INFO ] Version of installed CUDA driver: 7000
[11:55:25][3443][INFO ] Version of CUDA driver API used: 3020
...
------> Used in total: 411 MB (1637 MB free / 2048 MB total) -> Used by this application (assuming a single GPU task): 223 MB
[13:39:34][3443][ERROR] Error during CUDA host->device HS thresholds data transfer (error: 715)
[13:39:34][3443][ERROR] Demodulation failed (error: 1007)!