Top of the page

NVIDIA A100 still a premier punt for A.I. and all things GPT

During NVIDIAs annual GPU conference (GTC), the GPU industry titan gives a lot of detail and pays attention to GPU market trends and this year was no exception.

ChatGPT has become a craze - even something you might call an obsession - for many. H100 NVL, combines two NVIDIA H100 GPUs together to work on language models like ChatGPT. H100 isn’t a new product for Nvidia - but it applies it to accelerate this specific use case.

The new NVL has 96GB of memory - very impressive - especially for language models. The H100 is an upgrade from the A100 and, NVIDIA recently told the public that A100s have helped to train ChatGPT.

The model uses NVLink - using NVLink - you can deploy up to 256 H100s.

We are seeing many requests come in for GPT based GPU use cases and as a premier NVIDIA partner we are ready to help you on your GPU journey.

General Enquiry

NVIDIA A100 still a premier punt for A.I. and all things GPT

Even with the H100 advancements, the NVIDIA A100 is still in the race as a GPU of choice for A.I use cases. The latest craze of ChatGPT, and Bing AI NLP is making a dent in NVIDIAs stockpiles. The new H100, no doubt, is the leading GPU solution but its worth bearing in mind that the A100 is more readily available.

The original use case for A100 was not A.I or GPT but, like all markets, especially technological ones, things change fast.

With large language models being so talked about in 2023, and with large language models needing hundreds of GPUs to function properly, either the A100 or H100 is desirable. Large language models require many petabytes of data to be processed to look at patterns and give desired outputs. While one model of GPU may be optimal for this, it is a race to get as many functioning GPUs as possible.

NVIDIA really is taking the lead for A.I in 2023 and so their share price is reflective. The share price has nearly doubled already since the start of the year.

As a premier NVIDIA partner, we can help you put together your GPU systems

General Enquiry