What is Nvidia’s most powerful chip to date?

Latest update time：2024-02-12

Reads：

????If you want to see each other often, please mark it as a star???? and add it to your collection~

Source : Content compiled from techspot by Semiconductor Industry Observation (ID: i c b ank), thank you.

Nvidia's Grace Hopper GH200 is Nvidia's answer to the advanced AI and HPC processing needs of enterprises and data centers. When Grace Hopper was announced last year, Nvidia CEO Jensen Huang said the tech industry was suffering from traditional architectures, which was why it was increasingly turning to GPUs and accelerated computing to solve complex computing tasks.

Nvidia's DGX GH200 supercomputing platform contains 256 Grace Hopper GH200 superchips, each of which combines a 72-core Arm-based Grace CPU and an H100 Tensor Core GPU. The chip also supports up to 480GB LPDDR5 memory and 96GB HBM3 or 144GB HBM3e memory.

Regardless, the specs are impressive, and now Linux-focused publication Phoronix got its hands on a $40,000-plus GH200-based system to run some benchmarks to see if the GH200 holds up.

This is Nvidia’s most powerful AI chip to date, designed for ultra-large-scale AI and high-performance computing applications. The company claims it delivers up to ten times better performance for applications running terabytes of data, an increase in power that enables scientists and researchers to create unprecedented solutions to complex problems.

From an enterprise perspective, it could serve as an alternative to x86 CPUs in the server space, allowing Nvidia to offer products that are competitive with AMD and Intel products. Nvidia said that at the same power, Grace CPU is twice as powerful as Intel Sapphire Rapids and AMD Genoa CPU, and 3.5 times more efficient than AMD's previous generation Epyc Milan CPU.

Phoronix conducted benchmarks on Linux that proved the Superchip was competitive. For the HPCG benchmark, the GH200's performance was just ahead of the Xeon Platinum 8380 2P and just shy of the Epyc 9654 Genoa 2P, which Phoronix claims is "not bad at all for initial performance." It also found that the GH200 was the fastest compared to 2P Genoa(X) and Intel Emerald Rapids uniprocessor configurations. The GH200's performance is nearly twice that of the Ampere Altra Max 128-core Arm processor.

The GH200 is side-by-side with an AMD Epyc 9684X Genoa-X 1P processor and tested using the LavaMD test case for the Rodinia HPC benchmark. For AMG, the Super Chip system's CPU performance is almost on par with the Xeon Platinum 8380 2P.

Phoronix calls the processor performance of NWChem's computational chemistry software "very impressive." In the leading 2P configuration, a single GH200 is nearly tied with the AMD EPYC Genoa.

Original link

https://www.phoronix.com/review/nvidia-gh200-gptshop-benchmark

Click here to follow and lock in more original content

END

*Disclaimer: This article is original by the author. The content of the article is the personal opinion of the author. The reprinting by Semiconductor Industry Watch is only to convey a different point of view. It does not mean that Semiconductor Industry Watch agrees or supports the view. If you have any objections, please contact Semiconductor Industry Watch.

Today is the 3673rd issue of "Semiconductor Industry Observation" shared with you. Welcome to pay attention.