{"id":11736,"date":"2024-03-19T12:39:16","date_gmt":"2024-03-19T10:39:16","guid":{"rendered":"https:\/\/forklog.com\/en\/nvidia-unveils-new-technologies-at-gtc-2024\/"},"modified":"2024-03-19T12:39:16","modified_gmt":"2024-03-19T10:39:16","slug":"nvidia-unveils-new-technologies-at-gtc-2024","status":"publish","type":"post","link":"https:\/\/forklog.com\/en\/nvidia-unveils-new-technologies-at-gtc-2024\/","title":{"rendered":"Nvidia Unveils New Technologies at GTC 2024"},"content":{"rendered":"

Nvidia has announced a new generation of AI chips and software designed for large language models. The presentation<\/a> took place at the GTC 2024 conference.<\/p>\n

<\/p>\n

According to the company’s CEO, Jensen Huang, Nvidia updates its GPU<\/span> architecture every two years to enhance performance. Many previously released AI models have been trained on the Hopper architecture, which is used in chips like the H100, announced in 2022.<\/p>\n

\n\u201cHopper is fantastic, but we need more powerful GPUs,\u201d stated the company’s CEO.<\/cite><\/p><\/blockquote>\n

The new generation of AI graphics processors is named Blackwell. The first model, GB200, will be released later this year.<\/p>\n

\"image3-62\"
Blackwell chip. Source: Nvidia.<\/figcaption><\/figure>\n

Nvidia claims the new processors offer a significant performance boost for companies working with neural networks\u201420 petaflops with the new chips compared to 4 petaflops with the H100. According to Huang, the additional computational power will enable them to train larger and more complex models.<\/p>\n

For instance, training GPT-4 in 90 days required 8,000 old chips and 15 MW of power. Training on the new graphics card requires only 2,000 processors and 4 MW.<\/p>\n

It was also revealed that the GB200 will have 192 GB of memory with a bandwidth of 8 TB\/s.<\/p>\n

\n\u201cBlackwell is not just a chip, it’s a whole platform,\u201d noted Huang.<\/cite><\/p><\/blockquote>\n

The graphics processor combines two separately manufactured dies into a single chip, produced by TSMC. Together, they contain 208 billion transistors.<\/p>\n

\"image2-90\"
Nvidia Grace Blackwell Superchip. Source: Nvidia.<\/figcaption><\/figure>\n

The model will be available as a server called GB200 NVLink 2, consisting of 72 Blackwell graphics processors. Its total RAM will be 30 TB.<\/p>\n

\"image1-138\"
GB200 NVLink 2 server. Source: Nvidia.<\/figcaption><\/figure>\n

The company will sell access to the GB200 through cloud services. Nvidia announced that Amazon Web Services will build a server cluster consisting of 20,000 of these models.<\/p>\n

Huang also announced the addition of a new product called Nvidia Inference Microservice (NIM) to the enterprise software subscription. It simplifies the use of older graphics processors for computations.<\/p>\n

Back in January 2024, Mark Zuckerberg promised<\/a> to purchase Nvidia chips for Meta, as the \u201cfuture roadmap\u201d in artificial intelligence requires a \u201cvast computational infrastructure.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"

Nvidia has announced a new generation of AI chips and software designed for large language models. The presentation took place at the GTC 2024 conference.<\/p>\n","protected":false},"author":1,"featured_media":11735,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"","news_style_id":"","cryptorium_level":"","_short_excerpt_text":"","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[438,1295,1294],"class_list":["post-11736","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence","tag-chips","tag-nvidia"],"aioseo_notices":[],"amp_enabled":true,"views":"38","promo_type":"","layout_type":"","short_excerpt":"","is_update":"","_links":{"self":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/11736","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/comments?post=11736"}],"version-history":[{"count":0,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/11736\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media\/11735"}],"wp:attachment":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media?parent=11736"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/categories?post=11736"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/tags?post=11736"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}