NVIDIA has announced H200, the company’s newest high-end chip for training AI models.
Described by the company as ‘the world’s most powerful GPU for supercharging AI and HPC workloads,’ the H200 Tensor Core GPU will aim to supercharge generative AI and high-performance computing (HPC) workloads with game-changing performance and memory capabilities.
This news comes in the midst of NVIDIA working increasingly to enhance industry developments with AI, which includes business partnerships to unlock greater enterprise potential. The company has also recently introduced a new range of AI chips for the Chinese market, following US export restrictions on certain AI and gaming chips.
Fuelling greater global generative AI adoption
The H200 will reportedly have larger and faster memory capabilities to help fuel the company’s acceleration of generative AI and large language models (LLMs) while advancing scientific computing for HPC workloads.
As reported by CNBC, the H200 includes 141GB of next-generation “HBM3” memory that will help it to generate text, images or predictions using AI models.
NVIDIA has experienced huge AI and generative AI breakthroughs in 2023, having revealed very rapid sales growth and expecting to become the world’s first trillion-dollar semiconductor stock. The company is also very well known for its ‘AI is a reinvention of computing’ mindset, which was confirmed by company founder and CEO, Jensen Huang.
The H200 announcement comes in the midst of a massive chip shortage that NVIDIA is working to capitalise on, as businesses are seeking to get their hands on new developments to train their AI models. The company’s new GPU is already an upgrade from the H100 – the chip that OpenAI used to train GPT-4.
Achieving enterprise-ready AI
NVIDIA AI Enterprise, together with NVIDIA H200, is aiming to simplify the building of an AI-ready platform, accelerating AI development and deployment of production-ready generative AI, computer vision and speech AI.
The goal is that they will deliver enterprise-grade security, manageability, stability and support to gather actionable insights and value for businesses faster.
The company has also stated that the H200 is already compatible with the H100, which is convenient for AI companies as they will not have to change their existing server systems or software.
NVIDIA continues to introduce new generative AI platforms for a wide range of industries worldwide, including chips. Through these developments, the company is working to support businesses in adopting and customising AI to boost productivity.
The H200 is expected to be available from the second quarter of 2024.
Author: Amber Jackson
Source: https://technologymagazine.com/articles/h200-nvidia-release-high-end-chip-for-training-ai-models
