Sony Pixel Power calrec Sony

NVIDIA Hopper Leaps Ahead in Generative AI at MLPerf


It's official: NVIDIA delivered the world's fastest platform in industry-standard tests for inference on generative AI.

In the latest MLPerf benchmarks, NVIDIA TensorRT-LLM - software that speeds and simplifies the complex job of inference on large language models - boosted the performance of NVIDIA Hopper architecture GPUs on the GPT-J LLM nearly 3x over their results just six months ago.

The dramatic speedup demonstrates the power of NVIDIA's full-stack platform of chips, systems and software to handle the demanding requirements of running generative AI.

Leading companies are using TensorRT-LLM to optimize their models. And NVIDIA NIM - a set of inference microservices that includes inferencing engines like TensorRT-LLM - makes it easier than ever for businesses to deploy NVIDIA's inference platform.

Raising the Bar in Generative AI TensorRT-LLM running on NVIDIA H200 Tensor Core GPUs - the latest, memory-enhanced Hopper GPUs - delivered the fastest performance running inference in MLPerf's biggest test of generative AI to date.

The new benchmark uses the largest version of Llama 2, a state-of-the-art large language model packing 70 billion parameters. The model is more than 10x larger than the GPT-J LLM first used in the September benchmarks.

The memory-enhanced H200 GPUs, in their MLPerf debut, used TensorRT-LLM to produce up to 31,000 tokens/second, a record on MLPerf's Llama 2 benchmark.

The H200 GPU results include up to 14% gains from a custom thermal solution. It's one example of innovations beyond standard air cooling that systems builders are applying to their NVIDIA MGX designs to take the performance of Hopper GPUs to new heights.

Memory Boost for NVIDIA Hopper GPUs NVIDIA is sampling H200 GPUs to customers today and shipping in the second quarter. They'll be available soon from nearly 20 leading system builders and cloud service providers.

H200 GPUs pack 141GB of HBM3e running at 4.8TB/s. That's 76% more memory flying 43% faster compared to H100 GPUs. These accelerators plug into the same boards and systems and use the same software as H100 GPUs.

With HBM3e memory, a single H200 GPU can run an entire Llama 2 70B model with the highest throughput, simplifying and speeding inference.

GH200 Packs Even More Memory Even more memory - up to 624GB of fast memory, including 144GB of HBM3e - is packed in NVIDIA GH200 Superchips, which combine on one module a Hopper architecture GPU and a power-efficient NVIDIA Grace CPU. NVIDIA accelerators are the first to use HBM3e memory technology.

With nearly 5 TB/second memory bandwidth, GH200 Superchips delivered standout performance, including on memory-intensive MLPerf tests such as recommender systems.

Sweeping Every MLPerf Test On a per-accelerator basis, Hopper GPUs swept every test of AI inference in the latest round of the MLPerf industry benchmarks.

In addition, NVIDIA Jetson Orin remains at the forefront in MLPerf's edge category. In the last two inference rounds, Orin ran the most diverse set of models in the category, including GPT-J and Stable Diffusion XL.

The MLPerf benchmarks cover today's most popular AI workloads and scenarios, including generative AI, recommendation systems, natural language processing, speech and computer vision. NVIDIA was the only company to submit results on every workload in the latest round and every round since MLPerf's data center inference benchmarks began in October 2020.

Continued performance gains translate into lower costs for inference, a large and growing part of the daily work for the millions of NVIDIA GPUs deployed worldwide.

Advancing What's Possible Pushing the boundaries of what's possible, NVIDIA demonstrated three innovative techniques in a special section of the benchmarks called the open division, created for testing advanced AI methods.

NVIDIA engineers used a technique called structured sparsity - a way of reducing calculations, first introduced with NVIDIA A100 Tensor Core GPUs - to deliver up to 33% speedups on inference with Llama 2.

A second open division test found inference speedups of up to 40% using pruning, a way of simplifying an AI model - in this case, an LLM - to increase inference throughput.

Finally, an optimization called DeepCache reduced the math required for inference with the Stable Diffusion XL model, accelerating performance by a whopping 74%.

All these results were run on NVIDIA H100 Tensor Core GPUs.

A Trusted Source for Users MLPerf's tests are transparent and objective, so users can rely on the results to make informed buying decisions.

NVIDIA's partners participate in MLPerf because they know it's a valuable tool for customers evaluating AI systems and services. Partners submitting results on the NVIDIA AI platform in this round included ASUS, Cisco, Dell Technologies, Fujitsu, GIGABYTE, Google, Hewlett Packard Enterprise, Lenovo, Microsoft Azure, Oracle, QCT, Supermicro, VMware (recently acquired by Broadcom) and Wiwynn.

All the software NVIDIA used in the tests is available in the MLPerf repository. These optimizations are continuously folded into containers available on NGC, NVIDIA's software hub for GPU applications, as well as NVIDIA AI Enterprise - a secure, supported platform that includes NIM inference microservices.

The Next Big Thing The use cases, model sizes and datasets for generative AI continue to expand. That's why MLPerf continues to evolve, adding real-world tests with popular models like Llama 2 70B and Stable Diffusion XL.

Keeping pace with the explosion in LLM model sizes, NVIDIA founder and CEO Jensen Huang announced last week at GTC that the NVIDIA Blackwell architecture GPUs will deliver new levels of performance required for the multitrillion-parameter AI models.

Inference for large language models is difficult, requiring both e
See more stories from nvidia

More from Nvidia


Senua's Story Continues: GeForce NOW Brings Senua's Saga: Hellblade II' to the Cloud

Every week, GFN Thursday brings new games to the cloud, featuring some of the la...


Into the Omniverse: SoftServe and Continental Drive Digitalization With OpenUSD and Generative AI

Editor's note: This post is part of Into the Omniverse, a series focused on ...


Watt a Win: NVIDIA Sweeps New Ranking of World's Most Energy-Efficient Supercomputers

In the latest ranking of the world's most energy-efficient supercomputers, k...


New Performance Optimizations Supercharge NVIDIA RTX AI PCs for Gamers, Creators and Developers

NVIDIA today announced at Microsoft Build new AI performance optimizations and i...


NVIDIA Expands Collaboration With Microsoft to Help Developers Build, Deploy AI Applications Faster

If optimized AI workflows are like a perfectly tuned orchestra - where each comp...


A Superbloom of Updates in the May Studio Driver Gives Fresh Life to Content Creation

Editor's note: This post is part of our In the NVIDIA Studio series, which c...


Every Company to Be an Intelligence Manufacturer,' Declares NVIDIA CEO Jensen Huang at Dell Technologies World

AI heralds a new era of innovation for every business in every industry, NVIDIA ...


Fight for Honor in Men of War II' on GFN Thursday

Whether looking for new adventures, epic storylines or games to play with a friend, GeForce NOW members are covered. Start off with the much-anticipated sequel...


NVIDIA, Teradyne and Siemens Gather in the City of Robotics' to Discuss Autonomous Machines and AI

Senior executives from NVIDIA, Siemens and Teradyne Robotics gathered this week ...


Fire It Up: Mozilla Firefox Adds Support for AI-Powered NVIDIA RTX Video

Editor's note: This post is part of the AI Decoded series, which demystifies AI by making the technology more accessible, and which showcases new hardware, ...


How Basecamp Research Helps Catalog Earth's Biodiversity

Basecamp Research is on a mission to capture the vastness of life on Earth at an unprecedented scale. Phil Lorenz, CTO at Basecamp Research, discusses using AI ...


Needle-Moving AI Research Trains Surgical Robots in Simulation

A collaboration between NVIDIA and academic researchers is prepping robots for surgery. ORBIT-Surgical - developed by researchers from the University of Toront...


Gemma, Meet NIM: NVIDIA Teams Up With Google DeepMind to Drive Large Language Model Innovation

Large language models that power generative AI are seeing intense innovation - m...


Drug Discovery, STAT! NVIDIA, Recursion Speed Pharma R&D With AI Supercomputer

Described as the largest system in the pharmaceutical industry, BioHive-2 at the Salt Lake City headquarters of Recursion debuts today at No. 35, up more than 1...


Drug Discovery, STAT! NVIDIA, Recursion Speed Pharma R&D With AI Supercomputer

Described as the largest system in the pharmaceutical industry, BioHive-2 at the...


Dial It In: Data Centers Need New Metric for Energy Efficiency

Data centers need an upgraded dashboard to guide their journey to greater energy efficiency, one that shows progress running real-world applications. The formu...


Generating Science: NVIDIA AI Accelerates HPC Research

Generative AI is taking root at national and corporate labs, accelerating high-performance computing for business and science. Researchers at Sandia National L...


NVIDIA Blackwell Platform Pushes the Boundaries of Scientific Computing

Quantum computing. Drug discovery. Fusion energy. Scientific computing and physics-based simulations are poised to make giant steps across domains that benefit ...


Through the Wormhole: Media.Monks' Vision for Enhancing Media and Marketing With AI

Meet Media.Monks' Wormhole, an alien-like, conversational robot with a quirk...


Honkai: Star Rail' Blasts Off on GeForce NOW

Gear up, Trailblazers - Honkai: Star Rail lands on GeForce NOW this week, along with an in-game reward for members to celebrate the title's launch in the cl...


Get On the Train' NVIDIA CEO Says at ServiceNow's Knowledge 2024

Now's the time to hop aboard AI, NVIDIA founder and CEO Jensen Huang declared Wednesday as ServiceNow unveiled a demo of futuristic AI avatars together with...


‘Get On the Train,’ NVIDIA CEO Says at ServiceNow's Knowledge 2024

Now's the time to hop aboard AI, NVIDIA founder and CEO Jensen Huang declare...


NVIDIA CEO Jensen Huang to Deliver Keynote Ahead of COMPUTEX 2024

Amid an AI revolution sweeping through trillion-dollar industries worldwide, NVIDIA founder and CEO Jensen Huang will deliver a keynote address ahead of COMPUTE...


AI Decoded: New DaVinci Resolve Tools Bring RTX-Accelerated Renaissance to Editors

AI tools accelerated by NVIDIA RTX have made it easier than ever to edit and wor...


NVIDIA DGX SuperPOD to Power US Government Generative AI

In support of President Biden's executive order on AI, the U.S. government will use an NVIDIA DGX SuperPOD to produce generative AI advances in climate scie...


NVIDIA and Alphabet's Intrinsic Put Next-Gen Robotics Within Grasp

Intrinsic, a software and AI robotics company at Alphabet, has integrated NVIDIA AI and Isaac platform technologies to advance the complex field of autonomous r...


A Mighty Meeting: Generative AI, Cybersecurity Connect at RSA

Cybersecurity experts at the RSA Conference this week will be on the hunt for ways to secure their operations in the era of generative AI. They'll find man...


GeForce NOW Delivers 24 A-May-zing Games This Month

GeForce NOW brings 24 new games for members this month. Ninja Theory's highly anticipated Senua's Saga: Hellblade II will be coming to the cloud soon -...


NVIDIA AI Microservices for Drug Discovery, Digital Health Now Integrated With AWS

Harnessing optimized AI models for healthcare is easier than ever as NVIDIA NIM,...


Explainable AI: Insights from Arthur's Adam Wenchel enhances the performance of AI systems across various metrics like accuracy, explainability and fairness. In this episode of the NVIDIA AI Podcast, re...


AI Takes a Bow: Interactive GLaDOS Robot Among 9 Winners in Challenge

YouTube robotics influencer Dave Niewinski has developed robots for everything f...


Say It Again: ChatRTX Adds New AI Models, Features in Latest Update

Editor's note: This post is part of the AI Decoded series, which demystifies AI by making the technology more accessible, and which showcases new hardware, ...


SEA.AI Navigates the Future With AI at the Helm

Talk about commitment. When startup SEA.AI, an NVIDIA Metropolis partner, set out to create a system that would use AI to scan the seas to enhance maritime safe...


AI Drives Future of Transportation at Asia's Largest Automotive Show

The latest trends and technologies in the automotive industry are in the spotlight at the Beijing International Automotive Exhibition, aka Auto China, which ope...


Into the Omniverse: Unlocking the Future of Manufacturing With OpenUSD on Siemens Teamcenter X

Editor's note: This post is part of Into the Omniverse, a series focused on ...


Blast From the Past: Stream StarCraft' and Diablo' on GeForce NOW

Support for on GeForce NOW expands this GFN Thursday, as titles from the iconic StarCraft and Diablo series come to the cloud. StarCraft Remastered,...


Rays Up: Decoding AI-Powered DLSS 3.5 Ray Reconstruction

Editor's note: This post is part of the AI Decoded series, which demystifies AI by making the technology more accessible, and which showcases new hardware, ...


Forecasting the Future: AI2's Christopher Bretherton Discusses Using Machine Learning for Climate Modeling

Can machine learning help predict extreme weather events and climate change? Chr...


NVIDIA to Acquire GPU Orchestration Software Provider Run:ai

To help customers make more efficient use of their AI computing resources, NVIDIA today announced it has entered into a definitive agreement to acquire Run:ai, ...


How Virtual Factories Are Making Industrial Digitalization a Reality

To address the shift to electric vehicles, increased semiconductor demand, manufacturing onshoring, and ambitions for greater sustainability, manufacturers are ...


Small and Mighty: NVIDIA Accelerates Microsoft's Open Phi-3 Mini Language Models

NVIDIA announced today its acceleration of Microsoft's new Phi-3 Mini open l...


Climate Tech Startups Integrate NVIDIA AI for Sustainability Applications

Whether they're monitoring miniscule insects or delivering insights from satellites in space, NVIDIA-accelerated startups are making every day Earth Day. S...


Wide Open: NVIDIA Accelerates Inference on Meta Llama 3

NVIDIA today announced optimizations across all its platforms to accelerate Meta Llama 3, the latest generation of the large language model (LLM). The open mod...


Up to No Good: No Rest for the Wicked' Early Access Launches on GeForce NOW

It's time to get a little wicked. Members can now stream No Rest for the Wicked from the cloud. It leads six new games joining the GeForce NOW library of m...


NVIDIA Honors Partners of the Year in Europe, Middle East, Africa

NVIDIA today recognized 18 partners in Europe, the Middle East and Africa for their achievements and commitment to driving AI adoption. The recipients were hon...


Seeing Beyond: Living Optics CEO Robin Wang on Democratizing Hyperspectral Imaging

Step into the realm of the unseen with Robin Wang, CEO of Living Optics. The sta...


Moving Pictures: Transform Images Into 3D Scenes With NVIDIA Instant NeRF

Editor's note: This post is part of the AI Decoded series, which demystifies AI by making the technology more accessible, and which showcases new hardware, ...


New NVIDIA RTX A400 and A1000 GPUs Enhance AI-Powered Design and Productivity Workflows

AI integration across design and productivity applications is becoming the new s...


To Cut a Long Story Short: Video Editors Benefit From DaVinci Resolve's New AI Features Powered by RTX

Editor's note: This post is part of our In the NVIDIA Studio series, which c...


AI Is Tech's Greatest Contribution to Social Elevation,' NVIDIA CEO Tells Oregon State Students

AI promises to bring the full benefits of the digital revolution to billions acr...