March 29, 2024

NVIDIA Solidifies Its Lead in GPUs with New Blackwell Processor

Just a few short years ago, NVIDIA was only known for its graphics cards for Macs and PCs. My how things have changed. Since Generative AI took off, NVIDIA took off and their flagship H100 GPU – and demand was very high. That was before this week. Now they upped the game with their new Blackwell B200 GPU, which was announced at their annual conference in San Jose this week.

The Humble Early Years of NVIDIA

Just a few short years ago, NVIDIA was only known for its graphics cards for Macs and PCs. My how things have changed. Since Generative AI took off, NVIDIA took off and their flagship H100 is still in short supply due to the consumption needs of Large Language Models (LLMs).

What is the NVIDIA Blackwell – aka B200?

In AI circles NVIDIA is the hot engine and this week was the unveiling of their big new GPU Engine – Blackwell. The B200, which was code-named Blackwell, claims to be 30 times faster than its predecessor. Some of the key aspects of the B200:

It has 8,192 CUDA cores,
256 Tensor cores,
64 RT cores,

Combined it can perform up to 312 teraflops of mixed-precision floating-point operations per second. In other words, NVIDIA just extended its lead.

NVIDIA Locks In as a Supplier to Amazon, Google, Microsoft, Oracle

While many of its customers are trying to compete with their own Tensor units, the B200 just separated from the pack. We expect demand for the B200 to be very high – just as with the H100. The question is – how many new standalone competitors will emerge?

The GPU Era is Here – Long Live the CPU

We have reached the tipping point where CPUs cannot keep up with the data-crunching needs for AI. We have entered the GPU era and this will impact data center architectures until Quantum computing kicks in.

Enterprises need to rethink their designs for their own data centers – and negotiate pricing deals for Cloud usage of GPUs. Over time it will always be cheaper to leverage internal data centers – and costs are still not bad. Supply is short but first things first – new Data Center designs are needed to meet the needs of the AI era.

Bottom Line

NVIDIA just set a new standard for GPUs with Blackwell. The rest of the Tech community is on notice. Enterprises need to start now to re-design their approach to AI computing. It is a GPU world now – and that means changing designs – and doing so in a very fast manner.

UPCOMING WEBINAR

Q2 Research Agenda: Planning for AI

The AI Wars are here, prompting enterprises to strategize and incorporate AI into their business frameworks.

On April 4, 2024, join us for a webinar hosted by Jim Lundy, Aragon Research’s CEO and Founder. In this webcast, Jim will provide an overview of Aragon’s research agenda for the 2nd quarter of 2024. Don’t miss out on the opportunity to discover how you can utilize the latest AI insights and technologies to outpace your competitors.

This online presentation will address the following frequently asked questions posed to our analysts, offering comprehensive answers:

What are the key trends in the enterprise?
How can enterprises plan for adding AI to their portfolio?
What is Aragon’s Research Agenda for Q2?

NVIDIA Solidifies Its Lead in GPUs with New Blackwell Processor