
The full-stack NVIDIA accelerated computing platform has once again demonstrated exceptional performance in the latest MLPerf Training v4.0 benchmarks.
NVIDIA more than tripled the performance on the large language model (LLM) benchmark, based on GPT-3 175B, compared to the record-setting NVIDIA submission made last year. Using an AI supercomputer featuring 11,616 NVIDIA H100 Tensor Core GPUs connected with NVIDIA Quantum-2 InfiniBand networking, NVIDIA achieved this remarkable feat through larger scale - more than triple that of the 3,584 H100 GPU submission a year ago - and extensive full-stack engineering.
Thanks to the scalability of the NVIDIA AI platform, Eos can now train massive AI models like GPT-3 175B even faster, and this great AI performance translates into significant business opportunities. For example, in NVIDIA's recent earnings call, we described how LLM service providers can turn a single dollar invested into seven dollars in just four years running the Llama 3 70B model on NVIDIA HGX H200 servers. This return assumes an LLM service provider serving Llama 3 70B at $0.60/M tokens, with an HGX H200 server throughput of 24,000 tokens/second.
NVIDIA H200 GPU Supercharges Generative AI and HPC The NVIDIA H200 Tensor GPU builds upon the strength of the Hopper architecture, with 141GB of HBM3 memory and over 40% more memory bandwidth compared to the H100 GPU. Pushing the boundaries of what's possible in AI training, the NVIDIA H200 Tensor Core GPU extended the H100's performance by up to 47% in its MLPerf Training debut.
NVIDIA Software Drives Unmatched Performance Gains Additionally, our submissions using a 512 H100 GPU configuration are now up to 27% faster compared to just one year ago due to numerous optimizations to the NVIDIA software stack. This improvement highlights how continuous software enhancements can significantly boost performance, even with the same hardware.
This work also delivered nearly perfect scaling. As the number of GPUs increased by 3.2x - going from 3,584 H100 GPUs last year to 11,616 H100 GPUs with this submission - so did the delivered performance.
Learn more about these optimizations on the NVIDIA Technical Blog.
Excelling at LLM Fine-Tuning As enterprises seek to customize pretrained large language models, LLM fine-tuning is becoming a key industry workload. MLPerf introduced a new LLM fine-tuning benchmark this round, based on the popular low-rank adaptation (LoRA) technique applied to Meta Llama 2 70B.
The NVIDIA platform excelled at this task, scaling from eight to 1,024 GPUs, with the largest-scale NVIDIA submission completing the benchmark in a record 1.5 minutes.
Accelerating Stable Diffusion and GNN Training NVIDIA also accelerated Stable Diffusion v2 training performance by up to 80% at the same system scales submitted last round. These advances reflect numerous enhancements to the NVIDIA software stack, showcasing how software and hardware improvements go hand-in-hand to deliver top-tier performance.
On the new graph neural network (GNN) test based on R-GAT, the NVIDIA platform with H100 GPUs excelled at both small and large scales. The H200 delivered a 47% boost on single-node GNN training compared to the H100. This showcases the powerful performance and high efficiency of NVIDIA GPUs, which make them ideal for a wide range of AI applications.
Broad Ecosystem Support Reflecting the breadth of the NVIDIA AI ecosystem, 10 NVIDIA partners submitted results, including ASUS, Dell Technologies, Fujitsu, GIGABYTE, Hewlett Packard Enterprise, Lenovo, Oracle, Quanta Cloud Technology, Supermicro and Sustainable Metal Cloud. This broad participation, and their own impressive benchmark results, underscores the widespread adoption and trust in NVIDIA's AI platform across the industry.
MLCommons' ongoing work to bring benchmarking best practices to AI computing is vital. By enabling peer-reviewed comparisons of AI and HPC platforms, and keeping pace with the rapid changes that characterize AI computing, MLCommons provides companies everywhere with crucial data that can help guide important purchasing decisions.
And with the NVIDIA Blackwell platform, next-level AI performance on trillion-parameter generative AI models for both training and inference is coming soon.
Most recent headlines
05/01/2027
Worlds first 802.15.4ab-UWB chip verified by Calterah and Rohde & Schwarz to be ...
06/09/2026
June 9 2026, 23:00 (PDT) Dolby and MagentaTV Bring Fans Closer to the FIFA Worl...
04/08/2026
Dalet, a leading technology and service provider for media-rich organizations, t...
04/07/2026
April 7 2026, 19:00 (PDT) Detective Conan: Fallen Angel of the Highway Opens in...
22/06/2026
Behind The Mic provides a roundup of recent news regarding on-air talent, includ...
22/06/2026
Cosm has announced the appointment of David Ho as Chief Legal Officer, a newly created executive role reporting to President and CEO Jeb Terry. Ho will oversee ...
22/06/2026
Warner Bros. Discovery and Amazon Web Services (AWS) have announced the developm...
22/06/2026
Daktronics has completed an audio control system upgrade at Petco Park in San Di...
22/06/2026
Accelerate Media has named John Willi as President and announced the launch of the Accelerate Sports Network (ASN), a prep sports media and streaming platform c...
22/06/2026
All Women's Sports Network (AWSN) and 3XBA (3 3 Basketball Association) have announced live television coverage of the annual 3XBA tournament on Friday, Jun...
22/06/2026
OWL AI has announced the appointment of Jay Prasad as Chief Executive Officer and member of the Board of Directors. Prasad succeeds Josh Gwyther, who has served...
22/06/2026
CP Communications delivered RF video and audio support for TNT's Inside the NBA at the 2026 NBA Finals, providing main show coverage in San Antonio and ea...
22/06/2026
Polymarket has announced a partnership with GRID, an official esports data platf...
22/06/2026
As sports venues continue to evolve into more video-centric, fan-engagement-driv...
22/06/2026
As the regional sports production scene shifts toward streaming, this Texan helps lead the engineering behind Victory+'s growing live platform...
22/06/2026
By Kristin Feeley, Director, Documentary Film & Artist Programs
the memories of your elders [are] a scaffolding for you to build your identity on - and t...
22/06/2026
New hyper-resolution analyser EQ revealed
CEDAR Audio's all-new Icons plug-in series has just gained its newest member, Blade. Described by the compan...
22/06/2026
Turn any live input into a cinematic soundscape
Designed for use in the studio and on stage, Sampleson's latest creation is capable of taking any audio ...
22/06/2026
Adds guitar strings to Eurorack rigs
ADDAC System are renowned for their weird and wonderful synth designs, and their line-up includes plenty of gear that...
22/06/2026
FIFA World Cup 2026 fever grows, as more than one third of Australians tune in ...
22/06/2026
In our latest blog, Tim Pearson explores NAGRA Venturi, the new streaming security solution for the AI era from NAGRAVISION. Designed to aggregate and analyze ...
22/06/2026
Expanded integrations give advertisers access to distinct contextual signals acr...
22/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
22/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
22/06/2026
Xilica today announced the release of Dynamic Voice Lift, a new feature in Xilica Designer v4.12 that brings adaptive speech reinforcement to large meeting spac...
22/06/2026
Monday 22 June 2026
Official trailer released for Katie Price: Nothing to Hide,...
22/06/2026
The next era of AI will not be defined by compute alone. Its growth will be dete...
22/06/2026
Mission, Vision and Veritas - new Los Alamos National Laboratory (LANL) supercom...
22/06/2026
At the ISC conference running in Hamburg this week, NVIDIA is introducing new so...
22/06/2026
For the past two years, the U.S. National Science Foundation's National Arti...
22/06/2026
JUPITER, Europe's first exascale supercomputer at Germany's Forschungszentrum J lich, runs on NVIDIA Grace Hopper Superchips and NVIDIA Quantum-X800 Inf...
21/06/2026
To call the 2026 FIFA World Cup a big undertaking would be a big understatement....
21/06/2026
New series now live on Udemy
Regular SOS contributor and Cubase workshop columnist John Walden has just released a new Cubase video course that is now avail...
21/06/2026
Hot tubs sit at about 38 to 40 degrees Celsius, warm enough that most people can only soak for about 15 minutes. NVIDIA's newest AI servers can run their co...
21/06/2026
Sunday 21 June 2026
Sky announces immersive documentary series The Wargame
The Wargame first looks
ZIP (2MB)
Sky today confirms the commission of The Wargam...
20/06/2026
New add-on creates doubles & vocal stacks
IK Multimedia's latest ReSing add-on kits the innovative software out with the ability to automatically genera...
20/06/2026
What exactly is Apogee Control V3?
Control V3 is a new mixer application that controls Apogee interfaces. The new hit feature is that V3 finally allows for...
19/06/2026
Split compound eases operational challenges at Shinnecock Hills Golf Club...
19/06/2026
North Carolina, Oklahoma meet in the best-of-three Finals as ESPN leans into spe...
19/06/2026
Company launch comprehensive mix-comparison tool
The Him DSP are a plug-in company founded by The Him, an EDM DJ and producer who has amassed over half a bi...
19/06/2026
Major Sampler upgrades introduced
The latest version of Bitwig's DAW software has just entered public beta testing, and is available now for all users w...
19/06/2026
Four times the power of their predecessors
Akai Pro have just introduced upgraded versions of two of their popular standalone MPC systems, kitting them out ...
19/06/2026
Data from May shows seasonal outdoor trends triggers lower viewing
Warsaw, Pola...
19/06/2026
Buttons is best control system in the rAVe Best of Infocomm Awards 2026...
19/06/2026
Mavis Studio Makes iPad Production More Powerful
Brie Clayton June 19, 2026
0 Comments
InfoComm update brings new NDI Preview, PTZ control, USB audio ...
19/06/2026
Immersive Studio Metaverse Stage Tackles Post with Blackmagic Design
Brie Clayton June 19, 2026
0 Comments
New narrative projects rely on DaVinci Reso...
19/06/2026
How to Run the Original 1993 After Effects
Graham Quince June 19, 2026
0 Comments
How to the original After Effects v1 in an emulator, and you don'...
19/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
19/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
19/06/2026
SMPTE , the home of media professionals, technologists and engineers, has announced that its entire Standards catalog is now freely available to the global medi...