
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
06/10/2025
France T l visions, France's leading broadcaster, has received the 2025 EBU ...
04/09/2025
Monumental Sports & Entertainment (MSE), in collaboration with Dalet, has been a...
30/06/2025
Star Studded Ensemble Cast Are Joined by Richard Rankin as Filming Begins on the Second Season
[June 12, 2025 - Boston, MA]: The Forsytes, Debbie Horsfield...
30/06/2025
The Artemis II Space Launch System core stage is integrated with the solid rocket boosters inside High Bay 3 of the Vehicle Assembly Building at NASAs Kennedy S...
30/06/2025
RALEIGH, N.C. Capitol Broadcasting Co. has named Heather Gray vice president and general manager of WRAL-TV and WRAZ-TV here....
30/06/2025
The Virginia Association of Broadcasters has recognized Bill Sewell, Director of Engineering at WTKR & WGNT in Norfolk, Va. as the recipient of the 2025 J.J. Fr...
30/06/2025
The Society of Broadcast Engineers said its annual member drive resulted in the recruitment of 49 individual members....
30/06/2025
BURLINGTON, Mass. Avid today released its fully integrated news platform, uniting MediaCentral and Wolftech News in a single newsroom solution, and will demonst...
30/06/2025
WASHINGTON The Federal Communication's Enforcement and Media Bureaus have entered into a Consent Decree with Sinclair Broadcast Group to resolve a variety o...
30/06/2025
Eurorack sequencer module reimagined
California-based modular synth innovators Qu-Bit have announced the launch of a new module that offers a fresh new take...
30/06/2025
Berklee at Umbria Jazz Clinics to Host 40th Anniversary Concert The celebration will be held on July 10 in Perugia, Italy.
By
Colette Greenstein
June 30, 202...
30/06/2025
PremiumBeat Tips and Tricks
Brie Clayton June 30, 2025
0 Comments
When editing to impress, you'll need quality music, and if your studio happens t...
30/06/2025
Improved dynamic behaviour, improved audio quality & more
Techivation have announced the release of an upgraded edition of their very first premium plug-in,...
30/06/2025
Back to All News
Bel n Cuesta and Karra Elejalde Star in El ni o, the New Film ...
30/06/2025
Back to All News
A New Dangerous Troll Awakens: Netflix Unleashes Teaser for Troll 2Play Video
Play Video
Entertainment
30 June 2025
GlobalNorwayDenmarkSwe...
30/06/2025
The Focusrite Summer Sale is now on Don't miss unbeatable deals on Scarlett, Vocaster, and more.
Whether you're an artist, a producer, or a podcaste...
30/06/2025
All 8 episodes of Season 1 of 1923 will be available on RT Player from Tuesday ...
30/06/2025
Facebook
Twitter
LinkedIn
52% report AI security spending is displacing tr...
30/06/2025
Facebook
Twitter
LinkedIn
Cannes, June 30th, 2025 - Thales Alenia Space, t...
29/06/2025
Handpan-inspired instrument announced
Roland have announced the launch of the Mood Pan, a unique electronic hand percussion instrument that has been designe...
29/06/2025
Back to All News
A Secret Society, Ritualistic Killings, and a Century-Old Curs...
28/06/2025
Johannesburg, 27 June 2025 - As the nation commemorates Youth Month
2025, the N...
28/06/2025
WASHINGTON In a press conference following the Federal Communications Commission's May Open Meeting, Chair Brendan Carr promised the agency would move rapid...
28/06/2025
STAMFORD, Conn. Charter Communications has awarded $1.1 million in Spectrum Digital Education grants to 55 nonprofit organizations that work to expand access to...
28/06/2025
LAKE FOREST, Calif. June 19, 2025
What's New:
Sonnet Technologies today announced the certification of its Echo 20 Thunderbolt 4 SuperDock as an Engin...
28/06/2025
MASV (massive.io), the fastest and most reliable large file transfer platform for media professionals, has been named an IDC Innovator in the IDC Innovators: Me...
28/06/2025
Grass Valley today announced that TV SKYLINE GmbH, one of Europe's top mobile production providers, has expanded its camera inventory with 30 LDX 135 UHD/HD...
28/06/2025
AgileTV, a European leader in TV and video technology solutions, signed an agreement with Austrian telco LIWEST to develop and implement its TV service in Austr...
28/06/2025
Music theory plug-in updated
Three months on from the release of the latest version of their renowned music theory plug in, Scaler Music have launched an up...
28/06/2025
The 48th Annual Indian National Finals Rodeo Shot with Blackmagic PYXIS 6K
Brie Clayton June 27, 2025
0 Comments
Filmmaker Cameron Mackey relied on Bl...
28/06/2025
Social, Streaming Don't Compete, They Compliment
Andy Marken June 27, 2025
0 Comments
I think we've all arrived at a very special place. Spir...
28/06/2025
Blackmagic Design Captures Filipino Rock Band Drama Singtala
Brie Clayton June 27, 2025
0 Comments
Blackmagic URSA Mini Pro 12K and DaVinci Resolve St...
28/06/2025
Enhance Videos Faster with Aiarty Video Enhancer - Offline, Sharp, and Natural
Brie Clayton June 27, 2025
0 Comments
If you've used AI video tools...
27/06/2025
By Jessica Herndon
One of the most exciting things about the Sundance Film Fest...
27/06/2025
K-Pop remains one of the biggest genres globally, and many fans just can't get enough of it. That's why Spotify has launched a new series of K-Pop perf...
27/06/2025
In our latest blog post, Rafael Rivera highlights the rising threat of online scams, and the important role cybersecurity plays in protecting families across ge...
27/06/2025
WASHINGTON The Federal Communications Commission has set deadlines for comments to a notice of proposed rulemaking (NPRM) to codify certain foreign ownership re...
27/06/2025
WASHINGTON In a press conference following the Federal Communications Commission's May Open Meeting, Chair Brendan Carr promised the agency would move rapid...
27/06/2025
From grounded realism to bending, impossible geometries
Klevgrand have announced the release of a new algorithmic reverb plug-in which they say deconstruct...
27/06/2025
Learn to use REW for room analysis
Acoustic treatment is one of the most important factors in any studio, and with the extensive range of products available...
27/06/2025
SAN JOSE, Calif. The HDMI Forum has released Version 2.2. of the HDMI Specification with 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provid...
27/06/2025
MONTREAL Grass Valley has announced that TV Skyline GmbH, one of Europe's top mobile production providers, has expanded its camera inventory with the acquis...
27/06/2025
NEW YORK & CHICAGO FuboTV Inc. and Weigel Broadcasting Co. have announced a multi-year agreement for distribution of seven networks including MeTV, H&I, Movies!...
27/06/2025
NEW YORK A national survey of U.S. consumers shows 66% of us watch TV all or most of the time and also multitask while doing it....
27/06/2025
MIAMI Sunbeam Television has reached a multiyear agreement with Findal Media & Technology Group to broadcast the new ABC Miami beginning Aug. 4....
27/06/2025
STATE COLLEGE, Pa. AccuWeather has announced a deal with Perplexity, a AI-powered search and answer engine, that will bring AccuWeathers weather data and severe...
27/06/2025
WASHINGTON After being sworn in on June 23 as the Federal Communications Commission's newest Commissioner, Olivia Trusty has hit the ground running with the...
27/06/2025
WASHINGTON The Federal Communications Commissions has set deadlines for comments to a Notice of Proposed Rulemaking (NPRM) to codify certain foreign ownership r...
27/06/2025
TAG Video Systems, the leader in software-based IP media probing, monitoring, visualization, and analytics, has announced a new collaboration with Gencom Techno...
27/06/2025
Pixel Power (A Rohde & Schwarz Company) has recently been working with France T l visions, the French national public TV broadcaster, on a number of projects fo...