CC Blog Product News Tech News

NVIDIA Grace Hopper Superchip Sweeps MLPerf Inference Benchmarks

Nvidia Grace Hopper superchip
Written by Kirsten Campbell

NVIDIA GH200, H100 and L4 GPUs and Jetson Orin modules show exceptional performance running AI in production from the cloud to the network’s edge.

In its debut on the MLPerf industry benchmarks, the NVIDIA GH200 Grace Hopper Superchip ran all data center inference tests, extending the leading performance of NVIDIA H100 Tensor Core GPUs. The overall results showed the exceptional performance and versatility of the NVIDIA AI platform from the cloud to the network’s edge. Better hop to it!

GH200 Superchips Shine in MLPerf

The GH200 links a Hopper GPU with a Grace CPU in one superchip. The combination provides more memory, bandwidth and the ability to automatically shift power between the CPU and GPU to optimize performance.

Separately, NVIDIA HGX H100 systems that pack eight H100 GPUs delivered the highest throughput on every MLPerf Inference test in this round.

Grace Hopper Superchips and H100 GPUs led across all MLPerf’s data center tests, including inference for computer vision, speech recognition and medical imaging, in addition to the more demanding use cases of recommendation systems and the large language models (LLMs) used in generative AI. Overall, the results continue NVIDIA’s record of demonstrating performance leadership in AI training and inference in every round since the launch of the MLPerf benchmarks in 2018. The latest MLPerf round included an updated test of recommendation systems, as well as the first inference benchmark on GPT-J, an LLM with six billion parameters, a rough measure of an AI model’s size.

L4 Boosts Interference on Mainstream Servers

In the latest MLPerf benchmarks, NVIDIA L4 GPUs ran the full range of workloads and delivered great performance across the board. For example, L4 GPUs running in compact, 72W PCIe accelerators delivered up to 6x more performance than CPUs rated for nearly 5x higher power consumption. In addition, L4 GPUs feature dedicated media engines that, in combination with CUDA software, provide up to 120x speedups for computer vision in NVIDIA’s tests.

The MLPerf benchmarks are transparent and objective, so users can rely on their results to make informed buying decisions. They also cover a wide range of use cases and scenarios, so users know they can get performance that’s both dependable and flexible to deploy.

Nvidia |

Keep up-to-date with our FREE Weekly Newsletter!

Don't miss out on upcoming issues of Circuit Cellar.

Note: We’ve made the Dec 2022 issue of Circuit Cellar available as a free sample issue. In it, you’ll find a rich variety of the kinds of articles and information that exemplify a typical issue of the current magazine.

Would you like to write for Circuit Cellar? We are always accepting articles/posts from the technical community. Get in touch with us and let's discuss your ideas.

Product Editor | + posts

Kirsten Campbell is a Marketing Tornado and junk robot of information. Analytical and creative, she has been in marketing and communications since 2008 and worked with everyone from small businesses to your favorite household names. 


Ask her about the time she made a numismatics blog interesting (yes, really) or wrote an obit for a family she never met.


An ardent admirer of corporate snark played out online, Kirsten loves Reese’s peanut butter cups and still isn't over the Mars Rover.

Supporting Companies

Upcoming Events

Copyright © KCK Media Corp.
All Rights Reserved

Copyright © 2024 KCK Media Corp.

NVIDIA Grace Hopper Superchip Sweeps MLPerf Inference Benchmarks…

by Kirsten Campbell time to read: 2 min