Grace was the first-ever datacenter processor from NVIDIA. At the time, the company only shared a few tidbits of information about the chip, noting that it would utilize its NVLink technology to provide data transfer speeds of up to 900 GB/s between components. The first day of the GPU Technology Conference was on Tuesday. At the event, CEO Jensen Huang showed off the GraceCPU Superchip, a first in the Grace lineup.
The Grace CPUs are connected via the company's aforementioned NVLink technology. It consumes 500 watt of power and integrates a staggering 144 ARM cores into a single sockets. bandwidth speeds can be as high as 1 Terabyte per second with ultra-fast memory built into the chip.
Apple's recently announced M1 Ultra is a good example of how different the chips are. The M1 Ultra is made up of two M1 Max chips that are connected via UltraFusion technology.
The Department of Energy will be the first to get the Grace Superchip when it arrives in the first half of 2023. The company says its new chip is twice as fast as traditional server. It is estimated that it will achieve a score of approximately 740 points in SPECrate, putting it in the upper tier data center processors.
The Hopper architecture was announced by NVIDIA. It is the successor to the company's current Ampere architecture, and it is named after Grace Hopper, a computer scientist. Before you get excited, be aware that no mainstream GPUs were announced at GTC. We got to see the H 100. It is an 80 billion transistor behemoth built using TSMC's cutting-edge 4nm process. The H 100 is powered by the new Transformer Engine from NVIDIA, which the company claims allows it to offer unparalleled performance when it needs to compute transformer models. Over the past few years, transformer models have become popular with scientists working with systems like AlphaFold. The H 100 can cut the time it takes to train large models in half. The H 100 will be available later this year.
All Rights Reserved © 2024
