search

wilicc / gpu-burn

Multi-GPU CUDA stress test
BSD 2-Clause "Simplified" License
1.37k stars 295 forks source link

Low mem for result Error #77

Open pratik98 opened 1 year ago

pratik98 commented 1 year ago

I am getting this error. How do i fix this?

Initialized device 0 with 3956 MB of memory (117 MB available, using 105 MB of it), using FLOATS Results are 268435456 bytes each, thus performing 68719476734 iterations Couldn't init a GPU test: Low mem for result. aborting.

wilicc commented 1 year ago

Something goes haywire when you don't have much free memory available for the program. The problem you have is that all of your GPU's memory has already been allocated. Are you perhaps running another instance of GPU burn with full memory utilization? If you want to run multiple instances concurrently, reduce the amount of memory they are allowed to use with the -m parameter.

pratik98 commented 1 year ago

No i am only running one instance. i tried limiting the amount of memory. [i am using Jestson Nano By the way] ./gpu_burn -m 5% output run length not specified in the command line. Using compare file: compare.ptx Burning for 10 seconds. sh: 1: nvidia-smi: not found Could not invoke nvidia-smi, no temps available Segmentation fault (core dumped) Initialized device 0 with 3956 MB of memory (1539 MB available, using 76 MB of it), using FLOATS Results are 268435456 bytes each, thus performing 68719476734 iterations Couldn't init a GPU test: Low mem for result. aborting.

alexmyczko commented 8 months ago

check using nvtop or nvidia-smi how much memory is really left?