
Large language models and the applications they power enable unprecedented opportunities for organizations to get deeper insights from their data reservoirs and to build entirely new classes of applications.
But with opportunities often come challenges.
Both on premises and in the cloud, applications that are expected to run in real time place significant demands on data center infrastructure to simultaneously deliver high throughput and low latency with one platform investment.
To drive continuous performance improvements and improve the return on infrastructure investments, NVIDIA regularly optimizes the state-of-the-art community models, including Meta's Llama, Google's Gemma, Microsoft's Phi and our own NVLM-D-72B, released just a few weeks ago.
Relentless Improvements Performance improvements let our customers and partners serve more complex models and reduce the needed infrastructure to host them. NVIDIA optimizes performance at every layer of the technology stack, including TensorRT-LLM, a purpose-built library to deliver state-of-the-art performance on the latest LLMs. With improvements to the open-source Llama 70B model, which delivers very high accuracy, we've already improved minimum latency performance by 3.5x in less than a year.
We're constantly improving our platform performance and regularly publish performance updates. Each week, improvements to NVIDIA software libraries are published, allowing customers to get more from the very same GPUs. For example, in just a few months' time, we've improved our low-latency Llama 70B performance by 3.5x.
NVIDIA has increased performance on the Llama 70B model by 3.5x. In the most recent round of MLPerf Inference 4.1, we made our first-ever submission with the Blackwell platform. It delivered 4x more performance than the previous generation.
This submission was also the first-ever MLPerf submission to use FP4 precision. Narrower precision formats, like FP4, reduces memory footprint and memory traffic, and also boost computational throughput. The process takes advantage of Blackwell's second-generation Transformer Engine, and with advanced quantization techniques that are part of TensorRT Model Optimizer, the Blackwell submission met the strict accuracy targets of the MLPerf benchmark.
Blackwell B200 delivers up to 4x more performance versus previous generation on MLPerf Inference v4.1's Llama 2 70B workload. Improvements in Blackwell haven't stopped the continued acceleration of Hopper. In the last year, Hopper performance has increased 3.4x in MLPerf on H100 thanks to regular software advancements. This means that NVIDIA's peak performance today, on Blackwell, is 10x faster than it was just one year ago on Hopper.
These results track progress on the MLPerf Inference Llama 2 70B Offline scenario over the past year. Our ongoing work is incorporated into TensorRT-LLM, a purpose-built library to accelerate LLMs that contain state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT-LLM is built on top of the TensorRT Deep Learning Inference library and leverages much of TensorRT's deep learning optimizations with additional LLM-specific improvements.
Improving Llama in Leaps and Bounds More recently, we've continued optimizing variants of Meta's Llama models, including versions 3.1 and 3.2 as well as model sizes 70B and the biggest model, 405B. These optimizations include custom quantization recipes, as well as efficient use of parallelization techniques to more efficiently split the model across multiple GPUs, leveraging NVIDIA NVLink and NVSwitch interconnect technologies. Cutting-edge LLMs like Llama 3.1 405B are very demanding and require the combined performance of multiple state-of-the-art GPUs for fast responses.
Parallelism techniques require a hardware platform with a robust GPU-to-GPU interconnect fabric to get maximum performance and avoid communication bottlenecks. Each NVIDIA H200 Tensor Core GPU features fourth-generation NVLink, which provides a whopping 900GB/s of GPU-to-GPU bandwidth. Every eight-GPU HGX H200 platform also ships with four NVLink Switches, enabling every H200 GPU to communicate with any other H200 GPU at 900GB/s, simultaneously.
Many LLM deployments use parallelism over choosing to keep the workload on a single GPU, which can have compute bottlenecks. LLMs seek to balance low latency and high throughput, with the optimal parallelization technique depending on application requirements.
For instance, if lowest latency is the priority, tensor parallelism is critical, as the combined compute performance of multiple GPUs can be used to serve tokens to users more quickly. However, for use cases where peak throughput across all users is prioritized, pipeline parallelism can efficiently boost overall server throughput.
The table below shows that tensor parallelism can deliver over 5x more throughput in minimum latency scenarios, whereas pipeline parallelism brings 50% more performance for maximum throughput use cases.
For production deployments that seek to maximize throughput within a given latency budget, a platform needs to provide the ability to effectively combine both techniques like in TensorRT-LLM.
Read the technical blog on boosting Llama 3.1 405B throughput to learn more about these techniques.
Different scenarios have different requirements, and parallelism techniques bring optimal performance for each of these scenarios. The Virtuous Cycle Over the lifecycle of our architectures, we deliver significant performance gains from ongoing software tuning and optimization. These improvements translate into additional value for customers who train and deploy on our platforms. They're able to create more capable models and applications and deploy their existing models using less infrastructure, enhancing th
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...
07/07/2026
eds3_5_jq(document).ready(function($) { $(#eds_sliderM519).chameleonSlider_2_1({ content_source:......
07/07/2026
New Kessler 10 data now available in Nielsen CMI gives brands a deeper view of h...
07/07/2026
New Nielsen Ad Intel and CMV data reveals rising competition in Australia's ...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Riedel Communications and ARRI played a key role in one of the most technically ambitious Eurovision Song Contest productions to date. For the first time in the...
07/07/2026
Berklee Valencia Honors Eric Whitacre at the 2026 Commencement Ceremony The Grammy-winning composer was celebrated alongside this years cohort of masters degr...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
The appointment will drive growth and innovation across the Media Business Unit
Fincons Group, a leading IT business consultancy and systems integrator company...
07/07/2026
International broadcast and streaming expert Big Blue Marble has announced a multi-month Proof of Concept (PoC) for Media over QUIC (MoQ) during a major interna...
07/07/2026
Test & measurement innovator, Leader Electronics of Europe, is proud to announce that two of its leading products, the PHABRIX Qx rasterizer and LeaderPhabrix L...
07/07/2026
Clear-Com announced that Ross Production Services (RPS), a leading provider of mobile production solutions for live events and broadcasts, has standardized on ...
07/07/2026
LiveU redefined live newsgathering with the invention of cellular bonding. At IBC2026, Stand 7.C24, the company will show how that foundation is evolving into t...
07/07/2026
LucidLink, a leader in cloud-based file services for content creators, today announced an enhanced workflow integration between LucidLink and Adobe Frame.io tha...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
07/07/2026
New appointments to Screen Australia board 7 July 2026
The Albanese Labor Government has made new appointments to the Screen Australia board.
Dr Tania Chamber...
07/07/2026
The right team. The right time.' How The Killings at Parrish Station stuck t...
07/07/2026
Berklee College of Music Alum and Faculty Perform at Boston Public Library's...
07/07/2026
How MXL Connects More Than Media In this series, we explore the technologies, architectures and operational realities shaping modern media operations. Along the...
07/07/2026
To Whom It Concerns RT today confirmed that Patrick Kielty is to extend his contract as host of The Late Late Show.
Last season, the world's longest-runn...
07/07/2026
Max single-threaded CPUs at scale are a new category of CPUs built for the agentic AI era.
Across the creation and deployment of an agentic system, the CPU is...
07/07/2026
Scripps Research scientists train the immune system to make antibodies against numerous HIV strains Preclinical findings mark progress toward a long-sought vacc...
06/07/2026
Long-developed relationships and informal access, three production models, and c...
06/07/2026
The deal follows Lumine's recent acquisition of Synamedia's video-networ...
06/07/2026
This former SportsCenter producer is shaping local-team studio coverage to surro...
06/07/2026
IMS Productions and FOX Sports deployed more than 100 cameras, 32 onboard feeds,...
06/07/2026
Lumine Group Inc. has announced the completion of its acquisition of the Video Network Business from Synamedia. The business will operate as a standalone compan...
06/07/2026
NASCAR has announced a content and commercial partnership with Motorsport Network, a motorsports and automotive media platform, to distribute NASCAR original co...
06/07/2026
Versant Media Group has announced a definitive agreement to acquire Full Swing, a sports technology company, from Bruin Capital and a group of minority investor...
06/07/2026
DAZN has announced global broadcast rights to 3ICE, a professional hockey format featuring three skaters and a goalkeeper per side, with no power plays or offsi...
06/07/2026
Bundle gains 100 new synths & more
Synth Anthology 5 the largest and most feature-packed edition of UVI's hardware synth emulation bundle, bringing toge...
06/07/2026
Available now via free firmware update
MOTU have just announced that three of their popular audio interfaces now boast official Milan Certification. The 16A...
06/07/2026
Rohde & Schwarz completes Hunter Preliminary Design Review for the Integrated Co...
06/07/2026
The R&S IMAGER from Rohde & Schwarz uses millimeter wave imaging to inspect the ...
06/07/2026
The National Film and Video Foundation (NFVF), an agency of the Department of Sp...
06/07/2026
Digital Domain Appoints Industry Veteran Patrick Davenport as Executive Vice Pre...
06/07/2026
Gravity Media USA appoints Brittney Boston as Head of Business Development
Brie Clayton July 6, 2026
0 Comments
The strategic new appointment supports...
06/07/2026
Remote Editing Without Cloud Storage in Resolve
Michael Cioni July 6, 2026
0 Comments
This is a step-by-step guide that demonstrates how to setup your...
06/07/2026
Automatic Caption Translation in Adobe Premiere Pro
Colin Smith July 6, 2026
0 Comments
Here I showcase the innovative automatic caption translation f...
06/07/2026
Flavor's First Ever Managing Director Neal Cohen Retires; Longstanding Execu...