NVIDIA's new Eos supercomputer uses more than 10,000 H100 Tensor Core GPUs to train a 175 billion-parameter GPT-3 model in under four minutes.
Depending on the hardware you're using, training a large language model of any significant size can take weeks, months, even years to complete. That's no way to do business — nobody has the electricity and time to be waiting that long. On Wednesday, NVIDIA unveiled the newest iteration of its Eos supercomputer, one powered by more than 10,000 H100 Tensor Core GPUs and capable of training a 175 billion-parameter GPT-3 model on 1 billion tokens in under four minutes.
NVIDIA was quick to note that the 175 billion parameter version of GPT-3 used in the benchmarking is not the full-sized iteration of the model . The larger GPT-3 offers around 3.7 trillion parameters and is just flat out too big and unwieldy for use as a benchmarking test. For example, it'd take 18 months to train it on the older A100 system with 512 GPUs — though, Eos needs just eight days.
"Scaling is a wonderful thing," Salvator said."But with scaling, you're talking about more infrastructure, which can also mean things like more cost. An efficiently scaled increase means users are "making the best use of your of your infrastructure so that you can basically just get your work done as fast and get the most value out of the investment that your organization has made.
NVIDIA plans to apply these expanded compute abilities to a variety of tasks, including the company's ongoing work in foundational model development, AI-assisted GPU design, neural rendering, multimodal generative AI and autonomous driving systems.
Australia Latest News, Australia Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Nvidia’s RTX 4070 / 4080 Super cards rumored for CES 2024 launchNvidia might be ready to update its RTX 40 series cards in early 2024 after skipping the Super for the 30 series.
Read more »
Nvidia’s next GPU could enter Apple naming purgatoryA new rumor hints at Nvidia releasing an RTX 4070 Ti Super, which would be one of the most confusing names we've ever seen out of the company.
Read more »
Exclusive-Baidu placed AI chip order from Huawei in shift away from Nvidia -sourcesExclusive-Baidu placed AI chip order from Huawei in shift away from Nvidia -sources
Read more »
Exclusive: Baidu placed AI chip order from Huawei in shift away from Nvidia -sourcesBaidu (9888.HK) ordered artificial intelligence chips from Huawei (HWT.UL) this year, two people familiar with the matter said, adding to signs that U.S. pressure is prompting Chinese acceptance of the firm's products as an alternative to Nvidia's.
Read more »
Nvidia’s stock rallies toward longest win streak in 8 monthsTomi Kilgore is MarketWatch's deputy investing and corporate news editor and is based in New York. You can follow him on Twitter TomiKilgore.
Read more »
New MLPerf Benchmarks Show Why NVIDIA Reworked Its Product RoadmapI love to learn and share the amazing hardware and services being built to enable Artificial Intelligence, the next big thing in technology.
Read more »