{"id":50164,"date":"2023-11-20T18:02:54","date_gmt":"2023-11-20T10:02:54","guid":{"rendered":"https:\/\/www.ioiotimes.com\/?p=50164"},"modified":"2023-11-20T18:03:04","modified_gmt":"2023-11-20T10:03:04","slug":"%e8%8b%b1%e7%89%b9%e7%88%be%e5%8a%a0%e9%80%9fhpc%e5%92%8cai%e6%8a%80%e8%a1%93%e6%87%89%e7%94%a8%e6%96%bc%e7%a7%91%e5%ad%b8%e7%a0%94%e7%a9%b6%e7%99%bc%e5%b1%95%ef%bc%8c2023%e5%b9%b4super-computing","status":"publish","type":"post","link":"https:\/\/www.ioiotimes.com\/?p=50164","title":{"rendered":"\u82f1\u7279\u723e\u52a0\u901fHPC\u548cAI\u6280\u8853\u61c9\u7528\u65bc\u79d1\u5b78\u7814\u7a76\u767c\u5c55\uff0c2023\u5e74Super Computing\u5e74\u5ea6\u5c55\u6703(SC23)\u4e0a\u5c55\u793a\u5176\u6280\u8853\u5982\u4f55\u52a0\u901f\u79d1\u5b78\u5275\u65b0\u4e26\u63a8\u52d5\u65b0\u4e16\u4ee3\u904b\u7b97"},"content":{"rendered":"\n

\u82f1\u7279\u723e\u5728\u7f8e\u570b\u4e39\u4f5b Super Computing<\/a> \u5e74\u5ea6\u5c55\u6703\uff08SC23\uff09\u4e0a\u5c55\u793a\u4e86\u85c9\u7531 AI \u52a0\u901f\u7684\u9ad8\u6548\u80fd\u904b\u7b97\uff08HPC\uff09\uff0c\u7522\u54c1\u7d44\u5408\u6a6b\u8de8 Intel\u00ae Data Center GPU Max \u7cfb\u5217\u3001Intel\u00ae Gaudi\u00ae2 AI \u52a0\u901f\u5668\u3001Intel\u00ae Xeon\u00ae\u8655\u7406\u5668\uff0c\u7686\u5c55\u73fe\u51fa HPC \u548c AI \u5de5\u4f5c\u8ca0\u8f09\u7684\u9818\u5148\u6548\u80fd\u3002\u82f1\u7279\u723e\u4e5f\u5206\u4eab\u8207\u7f8e\u570b\u963f\u8ca2\u570b\u5bb6\u5be6\u9a57\u5ba4\uff08Argonne National Laboratory, ANL\uff09\u5408\u4f5c Aurora \u751f\u6210\u5f0f AI \u8a08\u756b\u7684\u76f8\u95dc\u9032\u5c55\uff0c\u5167\u5bb9\u5305\u62ec\u5728 Aurora \u8d85\u7d1a\u96fb\u8166\u4e0a\u53c3\u6578\u91cf\u9ad8\u9054 1 \u5146\u7684 GPT-3 \u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff08LLM\uff09\u6700\u65b0\u767c\u5c55\uff0c\u9019\u4e9b\u9032\u5c55\u7372\u76ca\u65bc Max \u7cfb\u5217 GPU \u7368\u7279\u67b6\u69cb\u548c Aurora \u8d85\u7d1a\u96fb\u8166\u7684\u7cfb\u7d71\u80fd\u529b\u3002\u82f1\u7279\u723e\u548c\u963f\u8ca2\u570b\u5bb6\u5be6\u9a57\u5ba4\u5408\u4f5c\uff0c\u900f\u904e Aurora \u65e9\u671f\u79d1\u5b78\u8a08\u756b\uff08Aurora Early Science Program\uff09\u548cExascale \u904b\u7b97\u5c08\u6848\uff08Exascale Computing Project, ECP\uff09\u7684\u5404\u9805\u61c9\u7528\uff0c\u52a0\u901f\u79d1\u5b78\u767c\u5c55\uff0c\u4e26\u540c\u6b65\u63ed\u793a Intel\u00ae Gaudi3 AI \u52a0\u901f\u5668\u548c Falcon Shores \u7684\u767c\u5c55\u85cd\u5716\u3002<\/p>\n\n\n\n

\u82f1\u7279\u723e\u4f01\u696d\u526f\u7e3d\u88c1\u66a8\u8cc7\u6599\u4e2d\u5fc3 AI \u89e3\u6c7a\u65b9\u6848\u7e3d\u7d93\u7406 Deepak Patil \u8868\u793a\uff1a\u300c\u82f1\u7279\u723e\u6301\u7e8c\u81f4\u529b\u65bc\u63d0\u4f9b\u5275\u65b0\u6280\u8853\u89e3\u6c7a\u65b9\u6848\uff0c\u4ee5\u6eff\u8db3 HPC \u548c AI \u793e\u7fa4\u7684\u9700\u6c42\u3002Xeon CPU\u3001Max GPU \u548c CPU \u5353\u8d8a\u7684\u6548\u80fd\u8868\u73fe\u90fd\u6709\u52a9\u65bc\u63a8\u9032\u76f8\u95dc\u7814\u7a76\u548c\u79d1\u5b78\u767c\u5c55\uff0c\u642d\u914d Gaudi \u52a0\u901f\u5668\u66f4\u80fd\u5168\u9762\u5c55\u73fe\u51fa\u82f1\u7279\u723e\u7684\u9818\u5148\u6280\u8853\u80fd\u5920\u70ba\u5ba2\u6236\u5e36\u4f86\u66f4\u503c\u5f97\u4fe1\u8cf4\u7684\u9078\u64c7\uff0c\u4ee5\u6eff\u8db3\u591a\u5143\u7684\u5de5\u4f5c\u8ca0\u8f09\u9700\u6c42\u3002\u300d<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

<\/p>\n\n\n\n

\u70ba\u4ec0\u9ebc\u91cd\u8981\uff1a<\/strong>\u91dd\u5c0d\u79d1\u5b78\u7814\u7a76\u7684\u751f\u6210\u5f0f AI \u53ca\u5176\u6700\u65b0\u6548\u80fd\u548c\u57fa\u6e96\u6e2c\u8a66\u7d50\u679c\uff0c\u7a81\u986f\u51fa\u82f1\u7279\u723e\u63d0\u4f9b\u7684\u5ba2\u88fd\u5316\u89e3\u6c7a\u65b9\u6848\u8db3\u4ee5\u6eff\u8db3 HPC \u548c AI \u5ba2\u6236\u7684\u7279\u5b9a\u9700\u6c42\u3002\u82f1\u7279\u723e\u900f\u904e\u8edf\u9ad4\u5b9a\u7fa9\u7684\u65b9\u5f0f\uff0c\u63a1\u7528 oneAPI \u898f\u7bc4\u8207 HPC \u548c AI \u589e\u5f37\u5de5\u5177\u5957\u4ef6\uff0c\u5354\u52a9\u958b\u767c\u4eba\u54e1\u8de8\u8d8a\u67b6\u69cb\u6846\u67b6\uff0c\u7121\u7e2b\u8f49\u79fb\u7a0b\u5f0f\u78bc\uff0c\u52a0\u901f\u79d1\u5b78\u7814\u7a76\u3002\u6b64\u5916\uff0cMax \u7cfb\u5217 GPU \u548c CPU \u4e5f\u5c07\u5e03\u7f72\u5728\u591a\u7d44\u5373\u5c07\u4e0a\u7dda\u7684\u8d85\u7d1a\u96fb\u8166\u4e4b\u4e2d\u3002<\/p>\n\n\n\n

\u963f\u8ca2\u570b\u5bb6\u5be6\u9a57\u5ba4\u5206\u4eab\u4e86\u4ee5 Aurora \u8d85\u7d1a\u96fb\u8166\u63a8\u52d5\u79d1\u5b78\u7814\u7a76\u751f\u6210\u5f0f AI \u8a08\u756b\u7684\u9032\u5c55\u3002\u900f\u904e Aurora \u751f\u6210\u5f0f AI \u8a08\u756b\uff0c\u963f\u8ca2\u570b\u5bb6\u5be6\u9a57\u5ba4\u3001\u82f1\u7279\u723e\u53ca\u5408\u4f5c\u5925\u4f34\u5c07\u651c\u624b\u5275\u5efa\u6700\u5148\u9032\u7684 AI \u6a21\u578b\uff0c\u5960\u5b9a\u79d1\u5b78\u767c\u5c55\u57fa\u790e\u3002\u9019\u4e9b\u6a21\u578b\u5c07\u4f7f\u7528\u8d85\u904e 1 \u5146\u53c3\u6578\u898f\u6a21\u7684\u79d1\u5b78\u6587\u672c\u3001\u7a0b\u5f0f\u78bc\u548c\u79d1\u5b78\u8cc7\u6599\u96c6\u9032\u884c\u8a13\u7df4\uff0c\u6db5\u84cb\u591a\u9580\u79d1\u5b78\u9818\u57df\u3002\u751f\u6210\u5f0f AI \u8a08\u756b\u5c07\u5229\u7528 Megatron \u548c DeepSpeed \u7684\u57fa\u790e\u6280\u8853\uff0c\u670d\u52d9\u751f\u7269\u3001\u764c\u75c7\u7814\u7a76\u3001\u6c23\u5019\u79d1\u5b78\u3001\u5b87\u5b99\u79d1\u5b78\u548c\u6750\u6599\u79d1\u5b78\u7b49\u591a\u9580\u79d1\u5b78\u9818\u57df\u3002<\/p>\n\n\n\n

Intel Max \u7cfb\u5217 GPU \u67b6\u69cb\u548c Aurora \u8d85\u7d1a\u96fb\u8166\u7cfb\u7d71\u64c1\u6709\u512a\u79c0\u7684\u6027\u80fd\uff0c\u53ea\u9700 64 \u7bc0\u9ede\u5373\u53ef\u9ad8\u6548\u7387\u8655\u7406 1 \u5146\u500b\u53c3\u6578\u7684\u6a21\u578b\uff0c\u9060\u4f4e\u65bc\u4e00\u822c\u6240\u9700\u3002\u963f\u8ca2\u570b\u5bb6\u5be6\u9a57\u5ba4\u4ee5 256 \u500b\u7bc0\u9ede\u8655\u7406\u56db\u500b\u5be6\u4f8b\uff0c\u5c55\u73fe\u51fa Aurora \u5e73\u884c\u8655\u7406\u591a\u5be6\u4f8b\u7684\u80fd\u529b\uff1b\u4e5f\u80fd\u5920\u52a0\u901f\u8a13\u7df4\u64f4\u5c55\u672a\u4f86\u8d85\u904e 1 \u842c\u7bc0\u9ede\u4e0a\u6578\u5146 token \u7684\u53c3\u6578\u6a21\u578b\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

<\/p>\n\n\n\n

\u95dc\u65bc\u82f1\u7279\u723e\u548c\u963f\u8ca2\u570b\u5bb6\u5be6\u9a57\u5ba4\uff1a<\/strong><\/p>\n\n\n\n

\u82f1\u7279\u723e\u548c\u963f\u8ca2\u570b\u5bb6\u5be6\u9a57\u5ba4\u5c55\u793a Aurora \u8d85\u7d1a\u96fb\u8166\u4e0a\u5982\u4f55\u900f\u904e\u7cfb\u7d71\u529f\u80fd\u548c\u8edf\u9ad4\u5806\u758a\u5be6\u73fe\u5927\u898f\u6a21\u52a0\u901f\u79d1\u5b78\u7814\u7a76\u3002\u5de5\u4f5c\u8ca0\u8f09\u5be6\u4f8b\u5305\u62ec\uff1a<\/p>\n\n\n\n

\u900f\u904e Connectomics ML \u5be6\u73fe\u898f\u6a21\u5316\u7684\u5927\u8166\u9023\u63a5\u7d44\u91cd\u5efa\uff0c\u5728 500 \u591a\u500b Aurora \u7bc0\u9ede\u4e0a\u986f\u793a\u5177\u7af6\u722d\u529b\u7684\u63a8\u8ad6\u7522\u51fa\u7387\u3002<\/p>\n\n\n\n

\u901a\u7528\u539f\u5b50\u548c\u5206\u5b50\u96fb\u5b50\u7d50\u69cb\u7cfb\u7d71\uff08GAMESS\uff09\u63a1\u7528\u00a0Intel Max GPU \u53ef\u5be6\u73fe\u8d85\u904e Nvidia A100 \u5169\u500d\u7684\u7af6\u722d\u6548\u80fd\uff0c\u53ef\u91dd\u5c0d\u85e5\u7269\u548c\u50ac\u5316\u5291\u8a2d\u8a08\u7684\u8907\u96dc\u5316\u5b78\u904e\u7a0b\u9032\u884c\u5efa\u6a21\uff0c\u900f\u904e Aurora \u8d85\u7d1a\u96fb\u8166\u89e3\u958b\u5206\u5b50\u79d1\u5b78\u7684\u79d8\u5bc6\u3002<\/p>\n\n\n\n

Hardware\/Hybrid Accelerated Cosmology Code,\uff08HACC\uff09\u5df2\u5728\u8d85\u904e 1,500 \u500b Aurora \u7bc0\u9ede\u4e0a\u5c55\u793a\u904b\u884c\uff0c\u5be6\u73fe\u5177\u8c61\u5316\u7406\u89e3\u5b87\u5b99\u7269\u7406\u53ca\u6f14\u5316\u3002<\/p>\n\n\n\n

\u85e5\u7269\u7be9\u9078 AI \u63a8\u8ad6\u61c9\u7528\u662f Aurora \u85e5\u7269\u63a2\u7d22\u65e9\u671f\u79d1\u5b78\u8a08\u756b\uff08ESP\uff09\u7684\u4e00\u90e8\u5206\uff1b\u50c5\u4f7f\u7528 256 \u500b\u7bc0\u9ede\u5c31\u80fd\u5b8c\u6210\u8d85\u904e 200 \u5104\u7a2e\u6700\u5e38\u898b\u5316\u5b78\u5408\u6210\u7269\u8cea\u7684\u7be9\u9078\uff0c\u5be6\u73fe\u5de8\u91cf\u5316\u5b78\u8cc7\u6599\u96c6\u7684\u9ad8\u6548\u7be9\u9078\u3002<\/p>\n\n\n\n

<\/p>\n\n\n\n

\u82f1\u7279\u723e\u540c\u6642\u5c55\u793a\u65b0\u7684 HPC \u548c AI \u6548\u80fd\u4ee5\u53ca\u8de8\u786c\u9ad4\u548c\u61c9\u7528\u7a0b\u5f0f\u7684\u8edf\u9ad4\u512a\u5316\uff1a<\/p>\n\n\n\n

\u82f1\u7279\u723e\u548c\u6234\u723e\u5171\u540c\u767c\u5e03 STAC-A2 \u7684\u7d50\u679c\uff0cSTAC-A2 \u662f\u57fa\u65bc\u771f\u5be6\u5e02\u5834\u98a8\u96aa\u5206\u6790\u5de5\u4f5c\u8ca0\u8f09\u7684\u7368\u7acb\u57fa\u6e96\u6e2c\u8a66\u5957\u4ef6\uff0c\u5c55\u73fe\u61c9\u7528\u65bc\u91d1\u878d\u7522\u696d\u7684\u51fa\u8272\u8868\u73fe\u3002\u8207 8 \u9846 Nvidia H100 PCIe GPU \u7684\u67b6\u69cb\u76f8\u6bd4\uff0c4 \u500b\u9846 Intel\u00ae Data Center GPU Max 1550 \u9054\u6210\u4e86\u9ad8\u51fa 26% \u7684\u5e0c\u81d8\u503c 10-100k-1260 \u6548\u80fd\uff0c\u7a7a\u9593\u6548\u7387\u4e5f\u63d0\u9ad8\u4e86 4.3 \u500d\u3002<\/p>\n\n\n\n

\u25a0\u5728\u5404\u7a2e HPC \u5de5\u4f5c\u8ca0\u8f09\u4e0a\uff0cIntel\u00ae Data Center GPU Max 1550 \u7684\u6548\u80fd\u6bd4 Nvidia H100 PCIe \u5361\u5e73\u5747\u9ad8\u51fa 36%\uff081.36 \u500d\uff09\u3002<\/p>\n\n\n\n

\u25a0\u82f1\u7279\u723e Data Center GPU Max \u7cfb\u5217\u52a0\u5f37\u4e86\u5c0d AI \u6a21\u578b\u7684\u652f\u63f4\uff0c\u5305\u62ec GPT-J \u548c LLAMA2 \u7b49\u591a\u7a2e\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff08LLM\uff09\u3002<\/p>\n\n\n\n

\u25a0Intel\u00ae Xeon\u00ae CPU Max \u7cfb\u5217\u662f\u552f\u4e00\u5177\u6709\u9ad8\u983b\u5bec\u8a18\u61b6\u9ad4\uff08HBM\uff09\u7684 x86 \u8655\u7406\u5668\uff0c\u8207\u00a0AMD Epyc Genoa \u8655\u7406\u5668\u76f8\u6bd4\uff0c\u6548\u80fd\u5e73\u5747\u63d0\u9ad8 19%\u3002<\/p>\n\n\n\n

\u25a0MLCommons2 \u65bc\u4e0a\u9031\u767c\u5e03<\/a>\u4e86\u7528\u65bc\u8a13\u7df4 AI \u6a21\u578b\u7684\u696d\u754c\u6a19\u6e96 MLPerf training v3.1 \u57fa\u6e96\u6e2c\u8a66\u7d50\u679c\u3002\u900f\u904e\u5728 v3.1 \u8a13\u7df4 GPT-3 \u57fa\u6e96\u6e2c\u8a66\u4e2d\u555f\u7528 FP8 \u529f\u80fd\uff0c\u82f1\u7279\u723e Gaudi2 \u5c55\u793a<\/a> 2 \u500d\u7684\u6548\u80fd\u589e\u76ca\u3002<\/p>\n\n\n\n

o\u00a0\u82f1\u7279\u723e\u5c07\u65bc 2024 \u5e74\u63a8\u51fa\u00a0Intel Gaudi3 AI \u52a0\u901f\u5668\u3002Gaudi3 AI \u52a0\u901f\u5668\u5c07\u57fa\u65bc\u8207\u00a0Gaudi2\u00a0\u76f8\u540c\u7684\u9ad8\u6548\u80fd\u67b6\u69cb\uff0c\u9810\u8a08\u63d0\u4f9b 4 \u500d\u904b\u7b97\u80fd\u529b\uff08BF16\uff09\u3001\u96d9\u500d\u7684\u7db2\u8def\u983b\u5bec\uff0c\u4ee5\u5be6\u73fe\u66f4\u9ad8\u7684\u6a6b\u5411\u64f4\u5c55\u6548\u80fd\uff0c\u4ee5\u53ca 1.5 \u500d\u7684\u5167\u5d4c HBM \u8a18\u61b6\u9ad4\uff0c\u8f15\u9b06\u6eff\u8db3\u5c0d LLM \u9ad8\u6548\u80fd\u3001\u9ad8\u6548\u7387\u904b\u7b97\u4e0d\u65b7\u589e\u9577\u7684\u9700\u6c42\uff0c\u4e14\u540c\u6642\u517c\u9867\u6548\u80fd\u3002<\/p>\n\n\n\n

\u25a0\u5982\u540c LAMMPS-Copper \u6240\u5c55\u793a\uff0c\u7b2c 5 \u4ee3 Intel\u00ae Xeon\u00ae\u8655\u7406\u5668\u5c07\u5728 HPC \u61c9\u7528\u7a0b\u5f0f\u4e0a\u63d0\u4f9b\u9ad8\u9054 1.4 \u500d\u7684\u6548\u80fd\u63d0\u5347\u3002<\/p>\n\n\n\n

o\u00a0Granite Rapids \u662f\u4e0b\u4e16\u4ee3 Intel Xeon \u8655\u7406\u5668\uff0c\u5c07\u900f\u904e Intel\u00ae Advanced Matrix Extensions\uff08Intel\u00ae AMX\uff09\u63d0\u4f9b\u66f4\u591a\u7684\u6838\u5fc3\u6578\u91cf\u548c\u5167\u5efa\u52a0\u901f\u5668\uff0c\u4e14\u652f\u63f4 MCR DIMMs\u3002Granite Rapids \u7684 DeepMD+LAMMPS AI \u63a8\u8ad6\u6027\u80fd\u5c07\u63d0\u9ad8 2.9 \u500d\u3002MCR \u57fa\u65bc DDR5 \u53ef\u5be6\u73fe 8,800 MT\/s \u7684\u50b3\u8f38\u901f\u5ea6\uff0c\u4e26\u5728\u96d9\u63d2\u69fd\u7cfb\u7d71\u4e2d\u5be6\u73fe\u8d85\u904e1.5 TB\/s \u7684\u8a18\u61b6\u9ad4\u983b\u5bec\uff0c\u9019\u5c0d\u65bc\u6eff\u8db3\u73fe\u4ee3 CPU \u5feb\u901f\u589e\u9577\u7684\u6838\u5fc3\u6578\u91cf\u4e26\u5be6\u73fe\u6548\u7387\u548c\u9748\u6d3b\u6027\u81f3\u95dc\u91cd\u8981\u3002<\/p>\n\n\n\n

<\/p>\n\n\n\n

\u95dc\u65bc<\/strong>oneAPI<\/strong>\u7684\u65b0\u9032\u5c55\uff1a<\/strong>\u82f1\u7279\u723e\u516c\u5e03\u4e86 2024 \u5e74\u8edf\u9ad4\u958b\u767c\u5de5\u5177\u5957\u4ef6\u7684\u529f\u80fd\uff0c\u85c9\u6b64\u63a8\u52d5\u7531 oneAPI \u591a\u67b6\u69cb\u7a0b\u5f0f\u652f\u63f4\u7684\u958b\u653e\u5f0f\u8edf\u9ad4\u958b\u767c\u3002\u65b0\u5de5\u5177\u53ef\u5354\u52a9\u958b\u767c\u4eba\u54e1\u5728\u82f1\u7279\u723e CPU \u548c GPU \u4e0a\u64f4\u5c55\u65b0\u7684 AI \u548c HPC \u529f\u80fd\uff0c\u6db5\u84cb\u7bc4\u570d\u66f4\u5ee3\uff0c\u5305\u62ec\u4f7f\u7528\u6a19\u6e96 Python \u5be6\u73fe\u6578\u4f4d\u8ca0\u8f09\u66f4\u5feb\u7684\u6548\u80fd\u548c\u5e03\u7f72\uff0c\u4ee5\u53ca\u7de8\u8b6f\u5668\u589e\u5f37\u529f\u80fd\uff0c\u4ea4\u4ed8\u8fd1\u4e4e\u5b8c\u6574\u7684 SYCL 2020\uff0c\u63d0\u9ad8\u751f\u7522\u529b\u548c\u7a0b\u5f0f\u78bc\u5378\u8f09\u3002<\/p>\n\n\n\n

\u6b64\u5916\uff0c\u5fb7\u5dde\u5148\u9032\u904b\u7b97\u4e2d\u5fc3\uff08TACC\uff09\u5ba3\u5e03 oneAPI\u5353\u8d8a\u4e2d\u5fc3<\/a>\u5c07\u5c08\u6ce8\u65bc\u958b\u767c\u53ca\u512a\u5316\u5730\u9707\u6210\u50cf\u57fa\u6e96\u7a0b\u5f0f\u78bc\u7684\u8a08\u756b\u3002\u5168\u7403\u5171 32 \u5ea7\u82f1\u7279\u723e oneAPI \u5353\u8d8a\u4e2d\u5fc3\u5171\u540c\u63a8\u52d5\u6b64\u8edf\u9ad4\u53ca\u786c\u9ad4\u5275\u65b0\u8207\u7814\u7a76\u7684\u7522\u696d\u767c\u5c55\u3002<\/p>\n\n\n\n

\u4e0b\u4e00\u6b65\uff1a<\/strong>\u82f1\u7279\u723e\u5f37\u8abf\u5176\u5c0d AI \u548c HPC \u7684\u6295\u5165\uff0c\u51f8\u986f\u5e02\u5834\u7684\u5f37\u52c1\u52d5\u80fd\u3002\u76ee\u524d\u63a1\u7528\u82f1\u7279\u723e Max \u7cfb\u5217 GPU \u548c CPU \u6280\u8853\u7684\u65b0\u8d85\u7d1a\u96fb\u8166\u5e03\u7f72\u5305\u62ec Aurora\u3001Dawn Phase 1\u3001SuperMUC-NG Phase 2\u3001Clementina XX1 \u7b49\u7cfb\u7d71\uff0c\u4ee5\u53ca\u4e00\u53f0\u4ee5 Stability AI \u70ba\u4e3b\u8981\u5ba2\u6236\uff0c\u63a1\u7528\u82f1\u7279\u723e Gaudi2 \u52a0\u901f\u5668\u65b0\u7cfb\u7d71\u7684\u5927\u578b AI \u8d85\u7d1a\u96fb\u8166\u3002<\/p>\n\n\n\n

\u9019\u4e9b\u52d5\u80fd\u5c07\u5960\u5b9a\u82f1\u7279\u723e\u5c0d\u65bc AI \u548c HPC \u7684\u6b21\u4e16\u4ee3 GPU\uff0dFalcon Shores \u7684\u7814\u767c\u57fa\u790e\u3002Falcon Shores \u5c07\u63a1\u7528 oneAPI \u958b\u653e\u6a19\u6e96\u5efa\u69cb\u7684\u55ae\u4e00 GPU \u7a0b\u5f0f\u8a2d\u8a08\u4ecb\u9762\uff0c\u61c9\u7528\u5728 Intel Gaudi \u548c Intel Xe<\/sup> \u7684 IP\u3002\u8b93\u76ee\u524d\u5efa\u69cb\u5728 Intel Gaudi AI \u52a0\u901f\u5668\u4ee5\u53ca Intel Max \u7cfb\u5217 GPU \u7684\u61c9\u7528\u7a0b\u5f0f\uff0c\u672a\u4f86\u4e5f\u80fd\u8f15\u9b06\u8f49\u79fb\u5230 Falcon Shores\u3002<\/p>\n\n\n\n

<\/p>\n\n\n\n

\ud83d\udfe6\u73fe\u5728\u5c31\u52a0\u5165 ioioTIMES \u81c9\u66f8\u7c89\u7d72\u5718<\/a> \u66f4\u591a\u4e92\u52d5\u3001\u66f4\u591a\u597d\u5eb7\u650f\u62b5\u52a0!!<\/strong>
\ud83d\udfe6\u6211\u5011\u6709LINE TODAY<\/a>\u983b\u9053\u4e86\uff0c\u5feb\u4f86\u8ffd\u8e2a\u6211\u5011\u5427!!–\u6700\u65b0\u79d1\u6280\u65b0\u805e \u76e1\u5728\u4f60\u624b<\/strong><\/h4>\n","protected":false},"excerpt":{"rendered":"

\u82f1\u7279\u723e\u5728\u7f8e\u570b\u4e39\u4f5b Super Com<\/p>\n","protected":false},"author":3,"featured_media":50175,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"rank_math_lock_modified_date":false,"footnotes":""},"categories":[13],"tags":[35,580,3083,7808,7794,126],"class_list":["post-50164","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","tag-ai","tag-focus","tag-hpc","tag-sc23","tag-super-computing","tag-126"],"_links":{"self":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/50164"}],"collection":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=50164"}],"version-history":[{"count":5,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/50164\/revisions"}],"predecessor-version":[{"id":50177,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/50164\/revisions\/50177"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/media\/50175"}],"wp:attachment":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=50164"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=50164"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=50164"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}