I'm seeing multiple failures running two threads of v3.08 (opencl_nvidia) x86_64-pc-linux-gnu on an NVIDIA GeForce GTX 1660 Ti with 6144MiB GPU memory and 32gb ram.

Amicable/Amicable_OpenCL.cpp, line 302: Kernel size for NVIDIA GPU has been set to 23
Amicable/Amicable_OpenCL.cpp, line 577: Self-test passed
Amicable/Amicable_OpenCL.cpp, line 592: clEnqueueWriteBuffer returned error -4
2024-03-27 17:33:48 (44894): called boinc_finish(-1)

amicable_10_21_15837_1711569302.291305_143_0

Tommy

Error code -4 means not enough GPU memory. Try to run only 1 WU at a time. You can also try to reduce kernel size to 21, it will reduce memory usage.

That was it - these recent wu are using almost twice the GPU memory:

0 N/A N/A 44957 C ...inc_Amicable/amicable_OpenCL_v_3_08 1724MiB |
0 N/A N/A 46310 C ...inc_Amicable/amicable_OpenCL_v_3_08 1724MiB |

0 N/A N/A 53400 C ...inc_Amicable/amicable_OpenCL_v_3_08 3362MiB |

Thanks
Tommy

https://sech.me/boinc/Amicable/workunit.php?wuid=33000654

I abandoned this task once. 8 hours of pointless waste of electricity and then only 12.5% ​​of the task was completed. Please remove it from the mailing list as this task has already caused problems for many people.