Jen-Hsun Huang, president and chief government officer of Nvidia Corp., announces the EGX Edge Supercomputing Platform in the course of the company’s celebration at Mobile World Congress Americas in Los Angeles, California, Oct. 21, 2019.
Patrick T. Fallon | Bloomberg | Getty Visuals
Nvidia’s most-innovative graphics playing cards are marketing for more than $40,000 on eBay, as desire soars for chips needed to coach and deploy synthetic intelligence software program.
The rates for Nvidia’s H100 processors had been noted by 3D gaming pioneer and previous Meta consulting engineering chief John Carmack on Twitter. On Friday, at minimum 8 H100s were mentioned on eBay at rates ranging from $39,995 to just beneath $46,000. Some vendors have available it in the earlier for about $36,000.
The H100, introduced previous calendar year, is Nvidia’s most current flagship AI chip, succeeding the A100, a approximately $10,000 chip that’s been referred to as the “workhorse” for AI apps.
Builders are utilizing the H100 to construct so-called huge language styles (LLMs), which are at the coronary heart of AI apps like OpenAI’s ChatGPT. Managing those people systems is high-priced and requires strong computers to churn by means of terabytes of facts for days or weeks at a time. They also depend on hefty computing electricity so the AI design can make textual content, images or predictions.
Coaching AI products, in particular big ones like GPT, involves hundreds of high-close Nvidia GPUs functioning alongside one another.
Microsoft used hundreds of millions of pounds on tens of countless numbers of Nvidia A100 chips to assistance construct ChatGPT. Nvidia controls the vast majority of the marketplace for AI chips.
Nvidia also presents a supercomputer with 8 GPUs working with each other termed the DGX. Previously this 12 months, the firm declared new solutions that would enable businesses to lease obtain to DGX computer systems for $37,000 for each thirty day period. At that price, the technique would use Nvidia’s more mature A100 processors.
Nvidia claims the H100 is the initial chip to be optimized for the particular AI architecture underpinning a lot of of the recent advancements in AI, identified as transformers. Sector specialists say the a lot more strong chips will be necessary to establish even even bigger and a lot more data-hungry models than people that are at present obtainable.
Enjoy: Google chip production normally takes goal at Nvidia’s generative A.I.
