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
Most recent headlines
04/08/2024
Dalet, a leading technology and service provider for media-rich organizations, is excited to announce Santiago Solanas as its new Chief Executive Officer (CEO)....
03/06/2024
Dalet, a leading technology and service provider for media-rich organizations, a...
08/05/2024
The solutions in-browser stream focuses on low latency, making it suited to low bandwidth scenarios
By Matthew Corrigan
Published: May 8, 2024
The solutio...
08/05/2024
Driven by algorithms, the solution performs a quality analysis informed by language-specific dictionaries
By Matthew Corrigan
Published: May 8, 2024
Drive...
08/05/2024
The increase in fleet size will reduce travel distances, aligning with sustainability goals, said Mobilelinks
By Matthew Corrigan
Published: May 8, 2024
T...
08/05/2024
The company said the new arrivals would drive collaboration and technological transformation across key business units
By Matthew Corrigan
Published: May 8, ...
08/05/2024
Following a successful NAB Las Vegas 2024 and winning a Best of Show Award, Actus Digital, a leading provider of Intelligent Monitoring Platforms, will bring it...
08/05/2024
TAG Video Systems, a leading force in video monitoring solutions, has developed a new Language Detection feature set to transform how operators ensure quality a...
08/05/2024
Pliant Technologies, a leading provider of professional wireless intercom solutions,presents its new SmartBoom LITE Headset at InfoComm 2024 (Booth C5116). The ...
08/05/2024
At the Media Production & Technology Show 2024, Prism Sound will showcase high-quality audio conversion products designed to suit the demands of professional us...
08/05/2024
Soho stalwart Julian Day has joined FooEngine as Business Development Director after 13 years at ZOO Digital. Julian has been at the heart of the London post pr...
08/05/2024
Following its successful launch at NAB 2024, Cleanfeed Cinema - the latest remote recording innovation from Emmy Award-winning Cleanfeed - is already helping au...
08/05/2024
DHD's range of digital broadcast equipment, systems and related software will be promoted on stand D22 at the Media Production & Technology Show by UK-regio...
08/05/2024
Cubbit, the innovator behind Europe's first distributed cloud storage enabler, today announced that CloudReso, a France-based distributor of MSP security so...
08/05/2024
MwareTV, a prominent cloud-based multi-tenant platform provider, has attracted Daniel Conde Coto to join the company as director, sales operations. This is a si...
08/05/2024
In a year set to see record-breaking IP-video adoption, with over 70 elections and global sports events, LiveU heads to Broadcast Asia 2024 with a focus on its ...
08/05/2024
Pioneers in microservices-based production environments, nxtedition, will demonstrate the latest advances in storytelling technology at the Media Production & T...
08/05/2024
Test & measurement innovator, Leader Electronics of Europe, has announced that it will again exhibit at The Media Production & Technology Show (MPTS), which tak...
08/05/2024
Livepeer Studio unveils a revolutionary video streaming platform offering an unprecedented combination of quality and cost-efficiency to content creators, media...
08/05/2024
Ikegami Electronics (Europe) will demonstrate a complete broadcast media production system on stand S1-A15 at CABSAT 2024 in Dubai, Tuesday May 21 through Thurs...
08/05/2024
Experience Commerce (EC), a leading digital agency within the Cheil Network, is pleased to announce that it has won the digital mandate for SAVSOL, the flagship...
08/05/2024
NEEDHAM, Mass. The International Data Corporation is predicting that worldwide spending on telecom services and pay TV services will increase by 1.4% in 2024 to...
08/05/2024
CUPERTINO, Calif. In a notable development for news and live video production, Apple has unveiled a number of significant upgrades to its Final Cut Pro software...
08/05/2024
Pliant Technologies has announced that it will be presenting its new SmartBoom LITE Headset at InfoComm 2024 (Booth C5116). The latest updates include enhanceme...
08/05/2024
SAN FRANCISCO Fox's ad-supported streaming service Tubi has launched Stubios, a fan-fueled studio for aspiring filmmakers and their fans that the company sa...
08/05/2024
SAN ANTONIO, Texas Five more stations have launched NextGen TV service, bringing to nine the number of local TV services on-air with ATSC 3.0....
08/05/2024
TYSONS, Va. Tegna Inc. has announced that Greg Retsinas has been named president and general manager at KGW, Tegna's NBC affiliate serving the Portland area...
08/05/2024
08 05 2024 - Media release Olivia Colman and John Lithgow Star in Sophie Hyde's New Project JIMPA
Olivia Coleman and John Lithgow in JIMPA. Photo credit: ...
08/05/2024
OpenDrives Joins AWS Partner Network
Brie Clayton May 7, 2024
0 Comments
Atlas software-defined platform now available via AWS trusted partner network...
08/05/2024
Animate AI Matte Paintings in After Effects
Graham Quince May 7, 2024
0 Comments
Arguably the best video use for AI generated images is Matte Painting...
08/05/2024
Throw Expression in Adobe After Effects
Andy Ford May 7, 2024
0 Comments
The Throw expression is a time-saver in After Effects. It allows you to mov...
08/05/2024
Join FilmLight in Toronto: Baselight 6.1, Nara and more
Brie Clayton May 7, 2024
0 Comments
Thursday 30th May, from 2:30PM, Hosted by Alter Ego 488 ...
08/05/2024
Gray Television reported a first-quarter profit as advertising revenues rebounded to above pre-COVID-19 levels....
08/05/2024
The Today show shared the performers on its summer concert series, which feature...
08/05/2024
Lewis Dodley, longtime anchor for Spectrum News NY1, will retire at the end of May. He's been on the air at the Charter Communications-owned local news chan...
08/05/2024
Scripps Sports said it plans to air a new studio show when a new season of WNBA Friday Night Spotlight starts May 24....
08/05/2024
Advertisers are looking to reach the right audiences and tell their stories inside compelling content, Karen Kovacs, president of client partnerships at NBCUniv...
08/05/2024
RT has today announced that Doireann Garrihy has decided to leave RT 2FM after five years. Doireann will present her final 2FM Breakfast with Doireann, Donncha...
08/05/2024
Amid an AI revolution sweeping through trillion-dollar industries worldwide, NVIDIA founder and CEO Jensen Huang will deliver a keynote address ahead of COMPUTE...
08/05/2024
AI tools accelerated by NVIDIA RTX have made it easier than ever to edit and wor...
08/05/2024
Professor emeritus John (Jack) Johnson elected to the National Academy of Sciences Johnson's multi-disciplinary research has been instrumental in shaping ou...
08/05/2024
zweiB updates to latest version of R&S CLIPSTER Specialist in deliverables for cinema, events and trade shows
Rohde & Schwarz, a global leader in broadcast...
07/05/2024
NAB Show 2024
Utah Scientific rocked the 2024 NAB Show in Las Vegas, unveiling two new game-changers in video signal management.
Our primary focus this yea...
07/05/2024
7th May 2024 The nations favourite comedy channel Gold has teamed up with the ma...
07/05/2024
There's something special about summer. We're not sure whether it's the warmer weather, the longer days, or something else a bit more magical, but s...
07/05/2024
NITV appoints Michael Rennie as Presenter and Senior Producer for NITV News
Media releases
NITV has announced the appointment of Michael Rennie as Presenter...
07/05/2024
The ability for line-of-site tactical voice and data networks to seamlessly traverse over Satellite Communications is an emerging requirement to assure warfight...
07/05/2024
On April 5, 2024, the L3Harris Avionics Products Repair Station (NN2R037L) in Gr...
07/05/2024
eds3_5_jq(document).ready(function($) { $(#eds_sliderM519).chameleonSlider_2_1({ content_source:......
It's official: NVIDIA delivered the world's fastest platform in industry-standard tests for inference on generative AI.
