机器学习及人工智能领域开放产业联盟MLCommons披露两项MLPerf基准评测的最新数据,其中英伟达H100芯片组在人工智能算力表现的测试中,刷新了所有组别的纪录,也是唯一一个能够跑完所有测试的硬件平台。
据了解,在投资者比较关注的“大语言模型”训练测试中,英伟达和GPU云算力平台CoreWeave提交的数据为这项测试设定了残酷的业界标准。在896个英特尔至强8462Y+处理器和3584个英伟达H100芯片的齐心协力下,仅仅花了10.94分钟就完成了基于GPT-3的大语言模型训练任务。
值得一提的是,英伟达和英特尔旗下的哈瓦那实验室(Habana Labs)是仅有的两家提交测试结果的芯片公司,其他提交测试结果的公司还包括戴尔(DELL.US)、联想(Lenovo)和超微电脑(SMCI.US)。
在英伟达首席执行官黄仁勋的带领下,英伟达在去年的GTC大会上推出了首款基于Hopper的H100。今年3月,英伟达展示了新款H100NVL。

MLCommons, an open industry alliance in the field of machine learning and artificial intelligence, disclosed the latest data from two MLPerf benchmarks, in which the Nvidia H100 chipset set a new record for all groups in the test of artificial intelligence computing performance, and was the only hardware platform that was able to run all tests.
It is understood that in the "large language model" training test that investors are more concerned about, the data submitted by Nvidia and GPU cloud computing platform CoreWeave has set a cruel industry standard for this test. With 896 Intel Xeon 8462Y+ processors and 3584 Nvidia H100 chips working together, it took just 10.94 minutes to complete the large language model training task based on GPT-3.
Notably, Nvidia and Intel-owned Habana Labs are the only two chip companies to submit test results, along with DELL (DELL.US), Lenovo (LNVY) and Supermicro (SMCI.US).
Led by Nvidia CEO Jen-Hsun Huang, Nvidia introduced its first Hoper-based H100 at last year's GTC conference. In March, Nvidia showed off its new H100NVL.




