GPU\u6545\u969c\u7684\u8868\u73b0\u5f62\u5f0f\u901a\u5e38\u4e0e\u5176\u6839\u6e90\u76f8\u5173\uff1a<\/p>\n

\n
• \u9a71\u52a8\u7248\u672c\u4e0d\u5339\u914d<\/strong>\u9999\u6e2f\u670d\u52a1\u5668\u90e8\u7f72\u4e2d\uff0c\u5185\u6838\u66f4\u65b0\u540e\u9a71\u52a8\u53ef\u80fd\u51fa\u73b0\u4e0d\u517c\u5bb9\u3002\u901a\u8fc7uname -r<\/code>\u67e5\u770b\u5185\u6838\u7248\u672c\uff0c\u5e76\u4e0eNVIDIA\u9a71\u52a8\u53d1\u5e03\u8bf4\u660e\u6bd4\u5bf9\uff0c\u786e\u8ba4\u517c\u5bb9\u6027\u3002<\/li>\n
• \u8d44\u6e90\u8fc7\u5ea6\u5206\u914d<\/strong>\u5728\u5bb9\u5668\u5316\u73af\u5883\u4e2d\uff0cnvidia-container-cli stats<\/code>\u7b49\u5de5\u5177\u53ef\u67e5\u770bDocker\/Kubernetes\u5bb9\u5668\u662f\u5426\u8d85\u51fa\u5206\u914d\u7684GPU\u5185\u5b58\u2014\u2014\u8fd9\u662f\u5171\u4eab\u670d\u52a1\u5668\u79df\u7528\u573a\u666f\u4e2d\u7684\u5e38\u89c1\u95ee\u9898\u3002<\/li>\n
• \u786c\u4ef6\u8001\u5316<\/strong>\u9999\u6e2f\u6e29\u6696\u6c14\u5019\u4e0b\uff0cGPU\u53cd\u590d\u70ed\u8282\u6d41\u53ef\u80fd\u5bfc\u81f4\u6c38\u4e45\u6027\u635f\u574f\u3002\u8fd0\u884cnvidia-smi -f \/tmp\/gpu_stats.log -l 60<\/code>\uff0c\u53ef\u6355\u6349\u957f\u671f\u6027\u80fd\u8870\u51cf\u8d8b\u52bf\u3002<\/li>\n<\/ul>\n

  [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””]<\/p>\n

  \u9999\u6e2f\u670d\u52a1\u5668\u57fa\u7840\u8bbe\u65bd\u4e2dGPU\u6545\u969c\u7684\u5b9e\u7528\u4fee\u590d\u65b9\u6848<\/strong><\/h2>\n

  1. \u9a71\u52a8\u7ba1\u7406\u4e0e\u8f6f\u4ef6\u4fee\u590d<\/h3>\n

  \u66f4\u65b0\u6216\u91cd\u88c5\u9a71\u52a8\u9700\u9075\u5faa\u7cfb\u7edf\u5316\u6d41\u7a0b\uff1a<\/p>\n

  \n
  1. \u5f7b\u5e95\u5378\u8f7d\u65e7\u9a71\u52a8<\/strong>\n

     apt-get remove --purge nvidia-*\r\nrm -rf \/etc\/nvidia \/usr\/lib\/nvidia<\/code><\/pre>\n<\/li>\n
     \u7248\u672c\u5339\u914d\u5b89\u88c5<\/strong>\u4ece\u5b98\u65b9\u4ed3\u5e93\u4e0b\u8f7d\u9a71\u52a8\uff0c\u786e\u4fdd\u4e0eLinux\u5185\u6838\u53ca\u670d\u52a1\u5668\u67b6\u6784\u5339\u914d\u3002\u9488\u5bf9\u9999\u6e2f\u6570\u636e\u4e2d\u5fc3\uff0c\u4f18\u5148\u9009\u62e9\u65e0GUI\u7684\u65e0\u5934\u9a71\u52a8\u5305\uff0c\u51cf\u5c11\u56fe\u5f62\u51b2\u7a81\uff1a\n
     chmod +x NVIDIA-Linux-x86_64-525.89.02.run\r\n.\/NVIDIA-Linux-x86_64-525.89.02.run --no-opengl --silent<\/code><\/pre>\n<\/li>\n
     \u5bb9\u5668\u8fd0\u884c\u65f6\u4fee\u590d<\/strong>\u5728Kubernetes\u96c6\u7fa4\u4e2d\uff0c\u66f4\u65b0nvidia-device-plugin<\/code>\u4ee5\u5339\u914d\u9a71\u52a8\u7248\u672c\u3002\u9a8c\u8bc1daemonset<\/code>\u914d\u7f6e\uff0c\u907f\u514d\u9999\u6e2f\u591a\u8282\u70b9\u90e8\u7f72\u4e2d\u7684GPU\u5206\u914d\u5931\u8d25\u3002<\/li>\n<\/ol>\n
     2. \u73af\u5883\u4e0e\u786c\u4ef6\u4fee\u590d<\/h3>\n
     \u5e94\u5bf9\u9999\u6e2f\u7684\u6c14\u5019\u6311\u6218\u662f\u9884\u9632\u6545\u969c\u590d\u53d1\u7684\u5173\u952e\uff1a<\/p>\n
     \n
     \u5373\u65f6\u6563\u70ed\u63aa\u65bd<\/strong>\u5728\u670d\u52a1\u5668\u673a\u67b6\u4e2d\u90e8\u7f72\u9ad8\u98ce\u91cf\u8f74\u6d41\u98ce\u6247\u589e\u5f3a\u6c14\u6d41\u3002\u5728\u670d\u52a1\u5668\u6258\u7ba1\u673a\u623f\u4e2d\uff0c\u9ad8\u6e29\u671f\u95f4\u9700\u4e0e\u6570\u636e\u4e2d\u5fc3\u5de5\u4f5c\u4eba\u5458\u534f\u8c03\uff0c\u8c03\u6574\u901a\u9053\u5bc6\u95ed\u7cfb\u7edf\u3002<\/li>\n
     \u786c\u4ef6\u66f4\u6362\u6d41\u7a0b<\/strong>\u9488\u5bf9\u670d\u52a1\u5668\u79df\u7528\u73af\u5883\u4e2d\u7684\u6545\u969cGPU\uff0c\u9075\u5faa\u4ee5\u4e0b\u6b65\u9aa4\uff1a\n
     \n
     \u62c6\u5378\u524d\u4f7f\u7528nvidia-smi -e 1<\/code>\u5907\u4efd\u56fa\u4ef6<\/li>\n
     \u786e\u4fdd\u66ff\u6362GPU\u4e0e\u539f\u578b\u53f7\u4e00\u81f4\uff0c\u907f\u514dPCIe\u901a\u9053\u914d\u7f6e\u95ee\u9898<\/li>\n
     \u4e3a\u9999\u6e2f\u670d\u52a1\u5668\u90e8\u7f72\u4e2d\u7684\u65b0\u786c\u4ef6\u91cd\u65b0\u914d\u7f6eBIOS\/UEFI\u8bbe\u7f6e<\/li>\n<\/ol>\n<\/li>\n
     \u957f\u671f\u6563\u70ed\u4f18\u5316<\/strong>\u5bf9\u4e8eAI\u5de5\u4f5c\u8d1f\u8f7d\uff0c\u53ef\u8003\u8651\u4e3a\u670d\u52a1\u5668\u6539\u88c5\u6db2\u51b7\u7cfb\u7edf\u3002\u6d78\u6ca1\u5f0f\u51b7\u5374\u5728\u9999\u6e2f\u9ad8\u6e7f\u5ea6\u73af\u5883\u4e2d\uff0c\u53ef\u4f7fGPU\u6e29\u5ea6\u964d\u4f4e30-40\u2103\u3002<\/li>\n<\/ul>\n
     [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””]<\/p>\n
     \u9999\u6e2f\u670d\u52a1\u5668\u8fd0\u7ef4\u4e2d\u7684GPU\u6545\u969c\u4e3b\u52a8\u9884\u9632\u7b56\u7565<\/strong><\/h2>\n
     1. \u5b9e\u65f6\u76d1\u63a7\u67b6\u6784<\/h3>\n
     \u90e8\u7f72\u53ef\u9760\u7684\u76d1\u63a7\u6808\u662f\u9884\u6d4b\u6027\u7ef4\u62a4\u7684\u6838\u5fc3\uff1a<\/p>\n
     \n
     Prometheus\u914d\u7f6e<\/strong>\u4f7f\u7528\u4ee5\u4e0b\u8bb0\u5f55\u89c4\u5219\u76d1\u63a7GPU\u7279\u5b9a\u6307\u6807\uff1a\n
     record: gpu_temp_warning\r\nexpr: nvidia_gpu_temp_celsius > 80\r\nrecord: gpu_memory_alert\r\nexpr: nvidia_gpu_memory_used_bytes \/ nvidia_gpu_memory_total_bytes * 100 > 90<\/code><\/pre>\n<\/li>\n
     Grafana\u4eea\u8868\u76d8\u8bbe\u7f6e<\/strong>\u521b\u5efa\u5305\u542b\u4ee5\u4e0b\u5185\u5bb9\u7684\u76d1\u63a7\u9762\u677f\uff1a\n
     \n
     24\u5c0f\u65f6GPU\u6e29\u5ea6\u8d8b\u52bf<\/li>\n
     \u670d\u52a1\u5668\u96c6\u7fa4\u4e2d\u9a71\u52a8\u7248\u672c\u4e00\u81f4\u6027<\/li>\n
     \u9999\u6e2f\u6570\u636e\u4e2d\u5fc3\u9ad8\u5cf0\u8d1f\u8f7d\u671f\u95f4\u7684GPU\u5185\u5b58\u5e26\u5bbd\u4f7f\u7528\u7387<\/li>\n<\/ol>\n<\/li>\n
     \u544a\u8b66\u7b56\u7565<\/strong>\u914d\u7f6e\u591a\u7ea7\u544a\u8b66\u2014\u201475\u2103\u89e6\u53d1\u8b66\u544a\uff0c85\u2103\u89e6\u53d1\u7d27\u6025\u544a\u8b66\uff0c\u4ee5\u9002\u5e94\u9999\u6e2f\u73af\u5883\u6e29\u5ea6\u7684\u6ce2\u52a8\u3002<\/li>\n<\/ul>\n
     2. \u7248\u672c\u63a7\u5236\u4e0e\u7ef4\u62a4\u6d41\u7a0b<\/h3>\n
     \u7ba1\u7406\u9a71\u52a8\u4e0e\u7cfb\u7edf\u66f4\u65b0\uff0c\u51cf\u5c11\u4e1a\u52a1\u4e2d\u65ad\uff1a<\/p>\n
     \n
     \u7ef4\u62a4\u7a97\u53e3<\/strong>\u5c06GPU\u9a71\u52a8\u66f4\u65b0\u5b89\u6392\u5728\u975e\u9ad8\u5cf0\u65f6\u6bb5\uff0c\u907f\u514d\u5f71\u54cd\u9999\u6e2f\u7684\u56fd\u9645\u4e1a\u52a1\u8fd0\u8425\u3002<\/li>\n
     \u7248\u672c\u517c\u5bb9\u6027\u77e9\u9635<\/strong>\u7ef4\u62a4\u8868\u683c\u8bb0\u5f55\u4ee5\u4e0b\u5bf9\u5e94\u5173\u7cfb\uff1a\n
     \n
     \u5185\u6838\u7248\u672c\u4e0e\u517c\u5bb9GPU\u9a71\u52a8<\/li>\n
     \u5bb9\u5668\u8fd0\u884c\u65f6\u7248\u672c\u4e0envidia-container-toolkit\u7248\u672c<\/li>\n
     \u56fa\u4ef6\u7248\u672c\u4e0e\u9999\u6e2f\u670d\u52a1\u5668\u578b\u53f7\u7684\u786c\u4ef6\u517c\u5bb9\u6027<\/li>\n<\/ul>\n<\/li>\n
     \u81ea\u52a8\u5316\u6d4b\u8bd5<\/strong>\u4f7f\u7528CI\/CD\u6d41\u6c34\u7ebf\u9a8c\u8bc1\u66f4\u65b0\u540e\u7684GPU\u529f\u80fd\u3002\u8fd0\u884cCUDA\u57fa\u51c6\u6d4b\u8bd5\u548c3D\u6e32\u67d3\u811a\u672c\uff0c\u786e\u4fdd\u6027\u80fd\u4e00\u81f4\u6027\u3002<\/li>\n<\/ul>\n
     3. \u5197\u4f59\u4e0e\u6545\u969c\u8f6c\u79fb\u673a\u5236<\/h3>\n
     \u4e3a\u9999\u6e2f\u670d\u52a1\u5668\u67b6\u6784\u6784\u5efa\u5f39\u6027\uff1a<\/p>\n
     \n
     GPU\u6545\u969c\u8f6c\u79fb\u811a\u672c<\/strong>\u521b\u5efasystemd\u5355\u5143\u76d1\u63a7GPU\u5065\u5eb7\u72b6\u6001\uff1a\n
     [Unit]\r\nDescription=GPU Health Monitor\r\nAfter=multi-user.target\r\n\r\n[Service]\r\nType=simple\r\nExecStart=\/usr\/bin\/bash -c \"while true; do \r\n  if nvidia-smi | grep -q 'Failed'; then \r\n    systemctl restart gpu-failover.target; \r\n  fi; sleep 30; \r\ndone\"<\/code><\/pre>\n<\/li>\n
     \u5730\u57df\u5197\u4f59<\/strong>\u5bf9\u4e8e\u6838\u5fc3\u4e1a\u52a1\u5e94\u7528\uff0c\u5728\u9999\u6e2f\u4e0d\u540c\u533a\u57df\u7684\u6570\u636e\u4e2d\u5fc3\u95f4\u590d\u5236\u5de5\u4f5c\u8d1f\u8f7d\u3002\u4f7f\u7528BGP\u8def\u7531\u5b9e\u73b099.99% SLA\u8981\u6c42\u5185\u7684\u6545\u969c\u8f6c\u79fb\u3002<\/li>\n
     \u5907\u7528\u786c\u4ef6\u50a8\u5907<\/strong>\u5728\u670d\u52a1\u5668\u6258\u7ba1\u90e8\u7f72\u4e2d\uff0c\u6bcf\u4e2a\u673a\u67b6\u81f3\u5c11\u50a8\u5907\u4e00\u5757\u5907\u7528GPU\uff0c\u4ee5\u7f29\u77ed\u786c\u4ef6\u6545\u969c\u540e\u7684\u5e73\u5747\u4fee\u590d\u65f6\u95f4\uff08MTTR\uff09\u3002<\/li>\n<\/ul>\n
     [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””]<\/p>\n
     \u7ed3\u8bed\uff1a\u9999\u6e2f\u670d\u52a1\u5668\u751f\u6001\u4e2dGPU\u7ba1\u7406\u7684\u6700\u4f73\u5b9e\u8df5<\/strong><\/h2>\n
     \u5728\u9999\u6e2f\u72ec\u7279\u7684\u670d\u52a1\u5668\u73af\u5883\u4e2d\u7ba1\u7406GPU\u5065\u5eb7\uff0c\u9700\u8981\u6280\u672f\u4e13\u4e1a\u77e5\u8bc6\u4e0e\u73af\u5883\u610f\u8bc6\u7684\u7ed3\u5408\u3002\u6280\u672f\u4eba\u5458\u7684\u6838\u5fc3\u8981\u70b9\u5305\u62ec\uff1a\u4f18\u5148\u5904\u7406\u6563\u70ed\u7ba1\u7406\u3001\u4e25\u683c\u63a7\u5236\u9a71\u52a8\u7248\u672c\u3001\u90e8\u7f72\u9002\u914d\u672c\u5730\u6c14\u5019\u6311\u6218\u7684\u4e3b\u52a8\u76d1\u63a7\u3002\u5c06\u8fd9\u4e9b\u7b56\u7565\u878d\u5165\u65e5\u5e38\u8fd0\u7ef4\uff0c\u56e2\u961f\u53ef\u6700\u5927\u9650\u5ea6\u51cf\u5c11GPU\u76f8\u5173\u505c\u673a\u65f6\u95f4\uff0c\u786e\u4fdd\u9999\u6e2f\u670d\u52a1\u5668\u79df\u7528\u4e0e\u6258\u7ba1\u670d\u52a1\u7684\u6700\u4f73\u6027\u80fd\u3002<\/p>\n
     \u5bf9\u4e8e\u6301\u7eed\u7684GPU\u7ef4\u62a4\uff0c\u5efa\u8bae\u5efa\u7acb\u4ee5\u4e0b\u5e38\u89c4\u6d41\u7a0b\uff1a<\/p>\n
     \n
     \u6bcf\u6708\u8fdb\u884c\u6e29\u5ea6\u68c0\u67e5\u4e0e\u7070\u5c18\u6e05\u7406<\/li>\n
     \u6bcf\u5b63\u5ea6\u5bf9\u7167NVIDIA\/AMD\u7684\u9999\u6e2f\u5730\u533a\u4e13\u9879\u5efa\u8bae\uff0c\u5ba1\u67e5\u9a71\u52a8\u7248\u672c<\/li>\n
     \u6bcf\u5e74\u89c4\u5212\u786c\u4ef6\u66f4\u65b0\uff0c\u5e94\u5bf9\u70ed\u5e26\u6c14\u5019\u4e0b\u7684\u8bbe\u5907\u8001\u5316\u95ee\u9898<\/li>\n<\/ul>\n
     \u5c06GPU\u6545\u969c\u9884\u9632\u89c6\u4e3a\u7cfb\u7edf\u6027\u6311\u6218\u800c\u975e\u88ab\u52a8\u5e94\u5bf9\u4efb\u52a1\uff0c\u4f01\u4e1a\u624d\u80fd\u5728\u9999\u6e2f\u7ade\u4e89\u6fc0\u70c8\u7684\u670d\u52a1\u5668\u79df\u7528\u4e0e\u6258\u7ba1\u5e02\u573a\u4e2d\uff0c\u7ef4\u6301\u9ad8\u6807\u51c6\u7684\u670d\u52a1\u53ef\u9760\u6027\u3002<\/p>\n
     [\/vc_column_text][\/vc_column][\/vc_row]<\/p>\n<\/div>","protected":false},"excerpt":{"rendered":"
     [vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””] \u5f15\u8a00\uff1a\u9999\u6e2f\u670d\u52a1\u5668\u73af\u5883\u4e2dGPU\u6545\u969c\u7684\u666e\u904d\u6027\u4e0e\u5f71\u54cd \u5728\u9999\u6e2f\u7684\u9ad8\u5bc6\u5ea6\u6570\u636e\u4e2d\u5fc3\u4e2d\u8fd0\u884c\u65f6\uff0cGPU\u9762\u4e34\u7740\u72ec\u7279\u6311\u6218\uff0c\u8fd9\u4e9b\u6311\u6218\u5f80\u5f80\u4f1a\u5f15\u53d1\u6545\u969c\u3002\u8fd9\u5ea7\u57ce\u5e02\u7684\u70ed\u5e26\u6c14\u5019\uff0c\u52a0\u4e4bAI\u8ba1\u7b97\u3001\u91d1\u878d\u4ea4\u6613\u7cfb\u7edf\u7b49\u9ad8\u5f3a\u5ea6\u5de5\u4f5c\u8d1f\u8f7d\uff0c\u5171\u540c\u9020\u5c31\u4e86\u4e00\u4e2a\u7279\u6b8a\u73af\u5883\u2014\u2014\u5728\u6b64\u73af\u5883\u4e2d\uff0cGPU\u4e0d\u7a33\u5b9a\u4e0d\u4ec5\u662f\u5c0f\u9ebb\u70e6\uff0c\u66f4\u662f\u4e1a\u52a1\u98ce\u9669\u3002GPU\u76f8\u5173\u6545\u969c\u7684\u5e38\u89c1\u8ff9\u8c61\u5305\u62ec\u7cfb\u7edf\u7a81\u7136\u5d29\u6e83\u3001\u6e32\u67d3\u4efb\u52a1\u4e2d\u51fa\u73b0\u56fe\u5f62\u5f02\u5e38\uff0c\u4ee5\u53ca\u65e5\u5fd7\u4e2d\u6301\u7eed\u51fa\u73b0\u4e0e\u9a71\u52a8\u6545\u969c\u76f8\u5173\u7684\u9519\u8bef\u4fe1\u606f\u3002\u5bf9\u4e8e\u7ba1\u7406\u670d\u52a1\u5668\u79df\u7528\u6216\u670d\u52a1\u5668\u6258\u7ba1\u90e8\u7f72\u7684\u6280\u672f\u4eba\u5458\u800c\u8a00\uff0c\u638c\u63e1\u9ad8\u6548\u8bca\u65ad\u8fd9\u7c7b\u6545\u969c\u7684\u65b9\u6cd5\uff0c\u662f\u4fdd\u969c\u670d\u52a1\u53ef\u9760\u6027\u7684\u5173\u952e\u3002 \u9999\u6e2f\u670d\u52a1\u5668\u4e2dGPU\u6545\u969c\u7684\u5178\u578b\u75c7\u72b6\u5305\u62ec\uff1a \u9ad8\u8d1f\u8f7d\u573a\u666f\u4e0b\u968f\u673a\u51fa\u73b0\u5c4f\u5e55\u95ea\u70c1\u6216\u5b8c\u5168\u9ed1\u5c4f \u5e94\u7528\u7a0b\u5e8f\u5d29\u6e83\u5e76\u663e\u793a\u201cGPU\u8fdb\u7a0b\u7ec8\u6b62\u201d\u7b49\u9519\u8bef\u4fe1\u606f GPU\u6e29\u5ea6\u8bfb\u6570\u5f02\u5e38\u504f\u9ad8\uff08\u53d7\u76d1\u63a7\u7cfb\u7edf\u4e2d\u5e38\u8d85\u8fc785\u2103\uff09 \u7cfb\u7edf\u65e5\u5fd7\u663e\u793a\u5185\u6838\u5d29\u6e83\u6216\u9a71\u52a8\u521d\u59cb\u5316\u5931\u8d25 [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””] GPU\u6545\u969c\u5206\u6b65\u8bca\u65ad\uff1a\u4ece\u75c7\u72b6\u5230\u6839\u6e90 1. \u786c\u4ef6\u4e0e\u73af\u5883\u57fa\u7840\u68c0\u67e5 \u5728\u6df1\u5165\u8f6f\u4ef6\u8bca\u65ad\u524d\uff0c\u9488\u5bf9\u9999\u6e2f\u72ec\u7279\u7684\u670d\u52a1\u5668\u751f\u6001\uff0c\u7269\u7406\u68c0\u67e5\u81f3\u5173\u91cd\u8981\uff1a \u6e29\u5ea6\u8bc4\u4f30\u4f7f\u7528IPMI\u5de5\u5177\u8fdc\u7a0b\u68c0\u67e5GPU\u6e29\u5ea6\u3002\u70ed\u5e26\u6c14\u5019\u4e0b\uff0c\u5373\u4fbf\u662f\u901a\u98ce\u826f\u597d\u7684\u6570\u636e\u4e2d\u5fc3\u4e5f\u53ef\u80fd\u51fa\u73b0\u70ed\u91cf\u79ef\u805a\uff0c\u56e0\u6b6480\u2103\u5e94\u4f5c\u4e3a\u89e6\u53d1\u5373\u65f6\u6392\u67e5\u7684\u9608\u503c\u3002 \u8fde\u63a5\u6027\u9a8c\u8bc1\u5bf9\u4e8e\u670d\u52a1\u5668\u6258\u7ba1\u673a\u623f\u4e2d\u7684\u8bbe\u5907\uff0c\u68c0\u67e5PCIe\u63d2\u69fd\u548c\u7535\u6e90\u7ebf\u662f\u5426\u5b58\u5728\u8150\u8680\u8ff9\u8c61\u2014\u2014\u8fd9\u662f\u6f6e\u6e7f\u73af\u5883\u4e2d\u7684\u5e38\u89c1\u95ee\u9898\u3002\u8fde\u63a5\u677e\u52a8\u901a\u5e38\u4f1a\u8868\u73b0\u4e3aGPU\u95f4\u6b47\u6027\u68c0\u6d4b\u5931\u8d25\u3002 \u591aGPU\u914d\u7f6e\u68c0\u67e5\u5728\u96c6\u7fa4\u90e8\u7f72\u4e2d\uff0c\u901a\u8fc7\u4ea4\u6362\u673a\u7aef\u53e3\u8bca\u65ad\u786e\u4fddGPU\u95f4\u901a\u4fe1\u672a\u4e2d\u65ad\u3002PCIe\u901a\u9053\u914d\u7f6e\u9519\u8bef\u53ef\u80fd\u5bfc\u81f4\u8d44\u6e90\u4e89\u7528\u6545\u969c\u3002 2. \u7cfb\u7edf\u7ea7\u8bca\u65ad\u5de5\u5177\u4e0e\u547d\u4ee4 \u5bf9\u4e8e\u9999\u6e2f\u670d\u52a1\u5668\u79df\u7528\u73af\u5883\u4e2d\u7684Linux\u7cfb\u7edf\uff0c\u4ee5\u4e0b\u7ec8\u7aef\u547d\u4ee4\u53ef\u63d0\u4f9b\u5173\u952e\u8bca\u65ad\u4fe1\u606f\uff1a # NVIDIA GPU\u8bca\u65ad\uff08AMD\u663e\u5361\u53ef\u66ff\u6362\u4e3a\u5bf9\u5e94\u547d\u4ee4\uff09 nvidia-smi -q -d TEMPERATURE,PERFORMANCE # \u8be6\u7ec6GPU\u5065\u5eb7\u62a5\u544a dmesg | grep -i -E “nvidia|gpu|driver” # \u5185\u6838\u65e5\u5fd7\u5206\u6790 lspci | grep -i vga # \u786c\u4ef6\u68c0\u6d4b\u9a8c\u8bc1 nvidia-debugdump –dump-all # \u751f\u6210\u5b8c\u6574\u8c03\u8bd5\u65e5\u5fd7 \u5b9e\u7528\u63d0\u793a\uff1a\u5728\u9999\u6e2f\u591a\u79df\u6237\u670d\u52a1\u5668\u4e2d\uff0c\u4f7f\u7528nvidia-smi –loop=5\u76d1\u63a7GPU\u5b9e\u65f6\u4f7f\u7528\u60c5\u51b5\uff0c\u53ef\u5e2e\u52a9\u8bc6\u522b\u7279\u5b9a\u865a\u62df\u5b9e\u4f8b\u7684\u8d44\u6e90\u5360\u7528\u8fc7\u9ad8\u95ee\u9898\u3002 3. \u573a\u666f\u5316\u6545\u969c\u9694\u79bb […]<\/p>\n
     Read More…<\/a><\/p>\n","protected":false},"author":1,"featured_media":27559,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[62],"tags":[8979,8980,8489,265],"class_list":["post-27564","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-hong-kong-dedicated-server-sc","tag-gpu-bugs-diagnosis-sc","tag-gpu-driver-fix-sc","tag-server-gpu-optimization-sc","tag-265"],"acf":[],"yoast_head":"\n