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 ...
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...
01/06/2026
January 6 2026, 05:30 (PST) Dolby Sets the New Standard for Premium Entertainment at CES 2026
Throughout the week, Dolby brings to life the latest innovatio...
31/05/2026
Olivia P rez-Collellmir to Premiere Original Work at Gaud Centennial in Barcelona The Berklee graduate and faculty member will debut her choral symphony with...
30/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
30/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
30/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
30/05/2026
Zero in on one that says yes (and no)
Andy Marken May 29, 2026
0 Comments
Hero image courtesy of Deposit Photos
For content creators the most difficu...
29/05/2026
With InfoComm 2026 just weeks away, NDI is giving attendees plenty of reasons to...
29/05/2026
Reaffirming a partnership that has defined Canadian sports broadcasting since 19...
29/05/2026
Mobile/tablet is No. 2 device for watching TV, suggesting that the sports-production industry needs to take another look at the format
Ring Digital's Sprin...
29/05/2026
Berliner Ensemble, one of Berlin's five major theater companies, has expande...
29/05/2026
Solid State Logic will showcase its new compact, fly-away TCA Tour audio product...
29/05/2026
Gerald (Jerry) Pierce, a pioneering technologist who helped shape the digital transformation of the motion picture industry, passed away last month on April 12 ...
29/05/2026
Paramount+ will be the English-language U.S. home for Barclays Women's Super...
29/05/2026
Further strengthening its virtualisation strategy to fully support broadcasters ...
29/05/2026
Swiss broadcaster Canal Alpha has deployed Harmonic's award-winning, software-based XOS Advanced Media Processor to modernize playout operations across cant...
29/05/2026
PTZOptics will showcase a new generation of intelligent video workflows at InfoComm 2026, June 17-19, Las Vegas. Visitors to booth N8227 will see how PTZOptics ...
29/05/2026
Arizona's Family has launched the Arizona's Family Sports (AZFS) streaming app, a new direct-to-consumer destination for live, local sports. The app is ...
29/05/2026
Starting in 2027, DAZN will be the exclusive home of The Canadian Football Leagu...
29/05/2026
Comcast Business has detailed the advanced network infrastructure it has deploye...
29/05/2026
In two-day event, leaders from academia and industry explored solutions to chall...
29/05/2026
The Basketball Tournament (TBT), now entering their 13th year of competition, ha...
29/05/2026
Roku has launched FOX One as a Premium Subscription on The Roku Channel in the U.S. Roku customers can now subscribe to FOX One using their Roku account for liv...
29/05/2026
In its sixth year, the broadcaster's coverage has become a global brand and ...
29/05/2026
Ratings Roundup is a rundown of recent ratings news and is derived from press re...
29/05/2026
The days are getting longer, the temperatures are rising, and playlists are filling up for the season. With summer around the corner, Spotify's global edito...
29/05/2026
New retro-inspired MPC announced
There are few devices that have gained the status held by Akai Pro's MPC range, and in recent years, the company have s...
29/05/2026
Save up to 30 on acclaimed titles
Following a successful launch at Superbooth 2026, Bjooks have revealed that they will be continuing the Kickstarter campa...
29/05/2026
Binaural monitoring application improved
Genelec have just released an update that brings some powerful new features to their HRTF-based binaural headphone ...
29/05/2026
6 June 2026 at SAE Institute, London, UK
IMSTA FESTA 2026 is almost upon us, with some of the biggest names in pro-audio set to descend upon SAE Institute i...
29/05/2026
Gerald (Jerry) Pierce, a pioneering technologist who helped shape the digital transformation of the motion picture industry, died April 12, 2026, at his home in...
29/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
29/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
29/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
29/05/2026
Comparing 5 AI Video Enhancers for Restoring Old Video Quality
Kate Luvis May 29, 2026
0 Comments
Digitizing VHS, MiniDV, and other legacy formats doe...
29/05/2026
Studio Hamburg Builds New Post Pipeline with DaVinci Resolve Studio
Brie Clayton May 29, 2026
0 Comments
Workflow replaces a patchwork of legacy tools...
29/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
29/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
29/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
29/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
29/05/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
29/05/2026
At the Intersection of Music and Dance, an Epic Collaboration Boston Conservatory musicians and dancers found creative parallels in their recent performance o...
29/05/2026
Back to All News
Latino Film Institute Fellows Screen Their Short Films at the Egyptian Theatre
The Inclusion Fellowship and Spark Animation Grant uplifts 15 ...
29/05/2026
GLOOKAST LAUNCHES CINNAFILM TACHYON PLUGIN FOR MEDIA PRODUCER AND MEDIA SERVICES
This release first appeared here.
Visit our Tachyon product page or contact...
29/05/2026
May 29 2026, 09:00 (PDT) Dolby and rednote Bring More Immersive Storytelling to...
29/05/2026
Something fundamental has shifted in how people consume media. Audiences aren't abandoning television or radio content; they're just expanding how, wher...
29/05/2026
Youtube exclusive special drops today
Watch now
UKTV today announces another e...
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.
