Nvidia H200: Leading chip manufacturer Nvidia has released the H200, a top-of-the-line chipset designed for AI applications. With numerous large firms, including Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure, already pledging to utilise the chipset next year, the upgrade to the already well-liked H100 chipset will begin delivering in the second quarter of 2024. In this article, we will tell you how much better it is than its predecessor.
Nvidia H200 vs H100
The AI computing giant said on Monday at the Supercomputing 2023 conference that the H200 GPU will have a memory bandwidth of 4.8 TB/s and 141GB of HBM3e high-bandwidth memory. This represents a significant improvement over the H100’s 80GB of HBM3 and 3.5 TB/s of memory capacity.
When it comes to two widely used large language models (LLMs) and a few high-performance computing tasks, Nvidia is marketing the H200 as a significant improvement over both the H100, which made its debut in 2022, and its predecessor, the A100, which made its debut in 2020.
The H200 is 60% faster than the H100 at running inference on the GPT-3 175B LLM. The GPU receives a 90 per cent improvement for inference on the Llama2 70B LLM.
Nvidia made the decision to pit the H200 against the A100 for HPC, claiming that the new GPU benchmarks two times quicker on average across CP2K, GROMACS, ICON, MILC, Chroma, and Quantum Espresso.
Hopper architecture, the foundation of the H100, also powers the H200. That implies the H200 has the same characteristics as the other models, like the Transformer Engine that accelerates LLMs and other deep learning models built on the transformer architecture, save for the additional memory capacity.
Launch timeline
With a performance bump that is nothing short of remarkable, the H200 is set to appear in Q2 of 2024.
Keep watching our YouTube Channel ‘DNP INDIA’. Also, please subscribe and follow us on FACEBOOK, INSTAGRAM, and TWITTER
