使用gpu-burn测试nvidia显卡显存

公司内部环境的GPU卡无法使用，重启服务器后频繁异常，使用gpu-burn工具对显卡进行测试。

操作步骤

下载源代码

git clone https://github.com/wilicc/gpu-burn.git

依赖安装

需要提前安装cuda，否则无法编译和使用。

yum install -y gcc-c++

编译

cd gpu-burn/
make

使用

编译完成后即可本地执行

[root@lolicp gpu-burn]# ./gpu_burn 
Run length not specified in the command line. Using compare file: compare.ptx
Burning for 10 seconds.
GPU 0: Tesla T4 (UUID: GPU-91e6dfa9-4514-b9a7-d868-442a1b821a62)
GPU 1: Tesla T4 (UUID: GPU-6863b855-2e09-6ab0-a62f-e1b224a625c9)
Initialized device 1 with 15109 MB of memory (6207 MB available, using 5586 MB of it), using FLOATS
Results are 268435456 bytes each, thus performing 19 iterations
Initialized device 0 with 15109 MB of memory (14801 MB available, using 13321 MB of it), using FLOATS
Results are 268435456 bytes each, thus performing 50 iterations
100.0%  proc'd: 0 (0 Gflop/s) - 19 (1805 Gflop/s)   errors: 0 - 0   temps: 36 C - 59 C 
    Summary at:   Tue May 14 13:21:17 CST 2024

Killing processes with SIGTERM (soft kill)
Freed memory for dev 1
Uninitted cublas
Freed memory for dev 0
Uninitted cublas
done
Tested 2 GPUs:
    GPU 0: OK
    GPU 1: OK

END