How to Burn Multi-GPUs using CUDA stress test memo
2 likes3,370 views
The document outlines a procedure for conducting a multi-GPU CUDA stress test on CentOS 7.3 using NVIDIA Tesla K80 GPUs. It provides installation steps for necessary software and libraries, as well as commands to monitor GPU performance and temperatures during the test. The results indicate successful operation with high performance metrics and no errors.
How to Burn Multi-GPUs using CUDA stress test memo
- 1. How to Burn Multi-GPUs using CUDA stress test memo (2017/05/20) 2017/05/20 SAKURA Internet, Inc. Research Center. SR / Naoto MATSUMOTO (C) Copyright 1996-2017 SAKURA Internet Inc
- 2. Multi-GPUs CUDA Stress Test on CentOS7.3 2 # uname -sr; cat /etc/redhat-release Linux 3.10.0-514.16.1.el7.x86_64 CentOS Linux release 7.3.1611 (Core) # grep proc /proc/cpuinfo | wc -l 12 # yum groupinstall "Development Tools" -y # yum install epel-release -y # yum install mesa-libGLU.x86_64 mesa-libGLU-devel.x86_64 freeglut-devel -y # wget https://developer.nvidia.com/compute/cuda/8.0/Prod2/local_installers/cuda_8.0.61_375.26_linux-run # bash ./cuda_8.0.61_375.26_linux-run # cd /usr/lib64; ln -sfn libGL.so.1 libGL.so # wget http://wili.cc/blog/entries/gpu-burn/gpu_burn-0.7.tar.gz # tar xzvf ./gpu_burn-0.7.tar.gz # vi Makefile NVCC=/usr/local/cuda/bin/nvcc # make # ./gpu_burn 100 GPU 0: Tesla K80 (UUID: GPU-45ee2619-60c8-d519-8fe3-ef2747abef4b) GPU 1: Tesla K80 (UUID: GPU-421a091f-7faf-6764-44ce-168be00d5c83) Initialized device 0 with 11439 MB of memory (11359 MB available, using 10223 MB of it), using FLOATS Initialized device 1 with 11439 MB of memory (11359 MB available, using 10223 MB of it), using FLOATS 100.0% proc'd: 12720 (2102 Gflop/s) - 13356 (2283 Gflop/s) errors: 0 - 0 temps: 72 C - 57 C
- 3. Multi-GPUs CUDA Stress Test on CentOS7.3 3 # watch -n 1 nvidia-smi Every 1.0s: nvidia-smi Sat May 20 07:16:10 2017 Sat May 20 07:16:11 2017 +-----------------------------------------------------------------------------+ | NVIDIA-SMI 375.26 Driver Version: 375.26 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | |===============================+======================+======================| | 0 Tesla K80 Off | A11D:00:00.0 Off | 0 | | N/A 81C P0 132W / 149W | 10290MiB / 11439MiB | 100% Default | +-------------------------------+----------------------+----------------------+ | 1 Tesla K80 Off | ACF7:00:00.0 Off | 0 | | N/A 60C P0 141W / 149W | 10290MiB / 11439MiB | 100% Default | +-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+ | Processes: GPU Memory | | GPU PID Type Process name Usage | |=============================================================================| | 0 12039 C ./gpu_burn 10286MiB | | 1 12037 C ./gpu_burn 10286MiB | +-----------------------------------------------------------------------------+