Nvidia, one of the world’s leading developers of semiconductor chips, revealed its latest chip on Aug. 7, designed to power high-level artificial intelligence (AI) systems.
The company said its next-generation GH200 Grace Hopper Superchip is one of the first to be equipped with an HBM3e processor, and is designed to process “ the world’s most complex generative AI workloads, spanning large language models, recommender systems and vector databases.”
Jensen Huang, the CEO of Nvidia, commented in a keynote that it is giving the processor a “boost” and that:
While the GH200 has the same general processing unit as the H100 — the company’s most high-end chip and one of the top in the world — it comes with 141 gigabytes of advanced memory and a 72-core ARM central processor, which is at least three times more powerful than the previous chip.
The latest chip from Nvidia is designed for inference, one of the two primary components of working with AI models after training them. Inference is when the model is used to generate content, make predictions and is constantly running.
Huang said “pretty much any” large language model (LLM) can be run through this chip, and it will “inference like crazy.”
The GH200 becomes available in the second quarter of 2024, according to Huang, and should be available for sampling by the end of 2023.
Related: OpenAI CEO highlights South Korean chips sector for AI growth, investment
This development comes as Nvidia’s market dominance is currently being challenged by the emergence of new semiconductor chips from rival companies racing to create the most powerful products.
Presently, it has over an 80% market share for AI chips and briefly topped $1 trillion in market value.
On May 28, Nvidia
Read more on cointelegraph.com