Nvidia boss Jensen Huang opened the tech giant’s conference by announcing that the Hopper graphics processing unit (GPU) will begin shipping in systems from partners including Dell, Hewlett Packard and Cisco Systems next month. Nvidia systems equipped with the GPU will be available in the first quarter of next year, says the tech giant’s CEO.
The Hopper processor, also known as the “H100”, is intended for data center tasks such as artificial intelligence. Nvidia says the H100 can “significantly” reduce the cost of deploying AI programs. For example, the work of earlier 320 high-end GPUs, the A100s, can be done with only 64 H100 chips, which would only require one-fifth of the number of server computers, and reduce power consumption by 3, 5 times. As a reminder, Hopper was first unveiled in March by Nvidia.
Along with Hopper’s availability, Jensen Huang talked about a new cloud-based “as-a-service” offering, Nvidia NeMo LLM, for “large language models”, which will allow customers to easily deploy very large processing models. natural language models such as OpenAI’s GPT-3 and Megatron-Turing NLG 530B, the language model that Nvidia developed with Microsoft.
Nvidia BioNeMo
Available in “early access” from next month, the NeMo service provides GPT-3 and other pre-trained models and can be tuned by a developer using a method Nvidia calls “prompt learning”. adapted from a technique Google scientists introduced last year.
A version of NeMo will be hosted specifically for biomedical uses of large language models, such as drug discovery, and will be called the “Nvidia BioNeMo” service.
Nvidia’s presentation also included a series of computing platforms dedicated to many industries, including healthcare, and the OmniVerse, Nvidia’s software platform for the Metaverse. These platforms include IGX for health applications, IGX for factories, a new version of OVX computing systems dedicated to the development of OmniVerse, and a new version of Nvidia’s DRIVE computing platform for cars, called DRIVE Thor, which will combine GPU Hopper with Nvidia’s upcoming “Grace” CPU.
Source: ZDNet.com