
Dramatic gains in hardware performance have spawned generative AI, and a rich pipeline of ideas for future speedups will drive machine learning to new heights, Bill Dally, NVIDIA's chief scientist and senior vice president of research, said today in a keynote.
Dally described a basket of techniques in the works - some already showing impressive results - in a talk at Hot Chips, an annual event for processor and systems architects.
The progress in AI has been enormous, it's been enabled by hardware and it's still gated by deep learning hardware, said Dally, one of the world's foremost computer scientists and former chair of Stanford University's computer science department.
He showed, for example, how ChatGPT, the large language model (LLM) used by millions, could suggest an outline for his talk. Such capabilities owe their prescience in large part to gains from GPUs in AI inference performance over the last decade, he said.
Gains in single-GPU performance are just part of a larger story that includes million-x advances in scaling to data-center-sized supercomputers. Research Delivers 100 TOPS/Watt Researchers are readying the next wave of advances. Dally described a test chip that demonstrated nearly 100 tera operations per watt on an LLM.
The experiment showed an energy-efficient way to further accelerate the transformer models used in generative AI. It applied four-bit arithmetic, one of several simplified numeric approaches that promise future gains.
Bill Dally Looking further out, Dally discussed ways to speed calculations and save energy using logarithmic math, an approach NVIDIA detailed in a 2021 patent.
Tailoring Hardware for AI He explored a half dozen other techniques for tailoring hardware to specific AI tasks, often by defining new data types or operations.
Dally described ways to simplify neural networks, pruning synapses and neurons in an approach called structural sparsity, first adopted in NVIDIA A100 Tensor Core GPUs.
We're not done with sparsity, he said. We need to do something with activations and can have greater sparsity in weights as well.
Researchers need to design hardware and software in tandem, making careful decisions on where to spend precious energy, he said. Memory and communications circuits, for instance, need to minimize data movements.
It's a fun time to be a computer engineer because we're enabling this huge revolution in AI, and we haven't even fully realized yet how big a revolution it will be, Dally said.
More Flexible Networks In a separate talk, Kevin Deierling, NVIDIA's vice president of networking, described the unique flexibility of NVIDIA BlueField DPUs and NVIDIA Spectrum networking switches for allocating resources based on changing network traffic or user rules.
The chips' ability to dynamically shift hardware acceleration pipelines in seconds enables load balancing with maximum throughput and gives core networks a new level of adaptability. That's especially useful for defending against cybersecurity threats.
Today with generative AI workloads and cybersecurity, everything is dynamic, things are changing constantly, Deierling said. So we're moving to runtime programmability and resources we can change on the fly,
In addition, NVIDIA and Rice University researchers are developing ways users can take advantage of the runtime flexibility using the popular P4 programming language.
Grace Leads Server CPUs A talk by Arm on its Neoverse V2 cores included an update on the performance of the NVIDIA Grace CPU Superchip, the first processor implementing them.
Tests show that, at the same power, Grace systems deliver up to 2x more throughput than current x86 servers across a variety of CPU workloads. In addition, Arm's SystemReady Program certifies that Grace systems will run existing Arm operating systems, containers and applications with no modification.
Grace gives data center operators a choice to deliver more performance or use less power. Grace uses an ultra-fast fabric to connect 72 Arm Neoverse V2 cores in a single die, then a version of NVLink connects two of those dies in a package, delivering 900 GB/s of bandwidth. It's the first data center CPU to use server-class LPDDR5X memory, delivering 50% more memory bandwidth at similar cost but one-eighth the power of typical server memory.
Hot Chips kicked off Aug. 27 with a full day of tutorials, including talks from NVIDIA experts on AI inference and protocols for chip-to-chip interconnects, and runs through today.
Most recent headlines
14/03/2025
BURLINGTON, Mass. Avid has expanded its executive leadership team (ELT) with the appointment of three experienced executives: Tom Sharma as chief technology off...
14/03/2025
CUPERTINO, Calif. TVU Networks, a pioneer in live video production and a leader in cloud-based media solutions, is marking its 20th anniversary at the 2025 NAB ...
13/03/2025
(L-R) Stephanie Suganami, Tatanka Means, John Malkovich, Ayo Edebiri, and Juliette Lewis attend the premiere of Opus at Eccles Theatre in Park City. (Photo by...
13/03/2025
Spotify took center stage at the London Book Fair this week, reaffirming our commitment to the audiobook market and showcasing our impact on the publishing indu...
13/03/2025
At Spotify, we work every day to lift up new voices, giving creators the opportunity to live off their art. Through Spotify Audiobooks, our in-house publishing ...
13/03/2025
Every time airborne law enforcement (ALE) teams fly, they expect their equipment to perform. Whether airborne units are conducting support for ground teams, bor...
13/03/2025
Sunday February 9 saw the annual return of the US' biggest television event, the Super Bowl LIX. Jamie McCombs, Fox Sports Audio Consultant / Sr. Audio capt...
13/03/2025
On Sunday March 2, stars across the film industry gathered at the Dolby Theatre in Los Angeles for the 97th Academy Awards. Production Sound Mixer Paul Sandweis...
13/03/2025
WUPPERTAL, Germany Riedel Communications will feature its new StageLink family of smart edge devices, Smart Audio and Mixing Engine (SAME) and Virtual Smart Pan...
13/03/2025
TAG Video Systems and Harmonic Partner to Deliver Enhanced Real-Time Monitoring ...
13/03/2025
DigitalGlue Invites NAB Visitors to Experience New Features in Managed Storage P...
13/03/2025
FOR-A America Theme - Connecting the Present, Building the Future - Comes to Lif...
13/03/2025
CTIA, the wireless industry association, has named former FCC Chairman Aji Pai as its President and Chief Executive Officer, effective April 1. He replaces Mere...
13/03/2025
he National Association of Broadcasters is urging the FCC to end its investigation of that controversial CBS interview with Kamala Harris stating there is no ...
13/03/2025
MIAMI CBS News & Stations continues to expand its use of augmented and virtual reality technologies with the planned launch of an augmented reality/virtual real...
13/03/2025
Srividhya Srinivasan, co-founder and chief customer success & innovation Officer at Amagi, tells TVBEurope how staying ahead of the latest trends is essential f...
13/03/2025
WASHINGTON FCC Chairman Brendan Carr announced that the agency has launched a massive, new deregulatory initiative that could potentially subject virtually all ...
13/03/2025
SUNNYVALE, Calif. SDVI has announced that it has integrated its Rally media supply chain management platform with the Spectra Vail multi-cloud data management s...
13/03/2025
ATLANTA Gray Media has announced that the Federal Communications Commission (FCC) has granted a waiver of its local ownership rules to permit Gray Media to acqu...
13/03/2025
TV Tech: What do you anticipate will be the most significant technology trends at the 2025 NAB Show?...
13/03/2025
Watch Student Naomi Soleils Folk Pop Performance on The Voice The songwriting major sang Stars by Grace Potter and the Nocturnals during the blind auditions.
...
13/03/2025
Roku Debuts NWSL Zone' Within Roku Sports to Spotlight Live Games, Content Feature marks Rokus first women's league-branded zone within service By Bra...
13/03/2025
Pixels, Images, Video, AI - and Us: A Perspective on AI from NDI Inventor Andrew...
13/03/2025
SVG New Sponsor Spotlight: Lighting Design Group's Steve Brill, Dennis Size ...
13/03/2025
IOC, Comcast NBCU Ink $3B Media-Rights Extension for Olympic Games Through 2036;...
13/03/2025
NCAA March Madness Live Returns with Expanded Availability of MultiView Vertical...
13/03/2025
SVG Sit-Down: Cosm's Devin Poolman on What It Takes To Deliver the Immersive...
13/03/2025
Sky Business, Sky's B2B division, providing connectivity and content to busi...
13/03/2025
Rohde & Schwarz launches R&S NRP140TWG(N) thermal power sensor: A new benchmark ...
13/03/2025
We asked the questions, analysed the answers and now we're excited to share the results of the fifth edition of the ICG Marketing Survey.
Our 2025 survey p...
13/03/2025
SAN JOSE, Calif. - March 13, 2025 Harmonic (NASDAQ: HLIT) today announced a breakthrough in video streaming innovation with the launch of new origin capabilitie...
13/03/2025
Haivision Showcases Mission-Critical Video Solutions at SOF Week 2025
Haivision's video wall systems, ISR technology, and DoDIN APL-certified video distri...
13/03/2025
Company Leads the Way with New Software-Based Production Platform, 12G Switcher,...
13/03/2025
A new Apple Immersive concert experience, Metallica, is coming to Apple Vision Pro this Friday, March 14. Filmed in Mexico City during the sold-out second-year ...
13/03/2025
Here is your host, Patrick Kielty!
Shake your Shamrocks, Patrick Kielty will be...
13/03/2025
This St. Patrick's weekend, RT invites you to celebrate all things Irish, showcasing the very best of Irish sport, entertainment and live coverage from the...
13/03/2025
What's next in AI is at GTC 2025. Not only the technology, but the people and ideas that are pushing AI forward - creating new opportunities, novel solution...
13/03/2025
Facebook
Twitter
LinkedIn
By combining T-Mobile's robust network, Thal...
13/03/2025
Bundle up - GeForce NOW is bringing a flurry of Blizzard titles to its ever-expanding library.
Prepare to weather epic gameplay in the cloud, tackling the genr...
13/03/2025
AI is leveling up the world's most beloved games, as the latest advancements...
13/03/2025
PC game modding is massive, with over 5 billion mods downloaded annually. Mods p...
12/03/2025
O Spotify acaba de lan ar o relat rio Loud & Clear deste ano, uma vis o transpar...
12/03/2025
Spotify acaba de presentar el informe Loud & Clear de este a o, una mirada trans...
12/03/2025
Ramadan, a period of profound spiritual significance for Muslims worldwide, is a time for fasting, prayer, reflection, and community. Enrich your experience thi...
12/03/2025
Spotify has just unveiled this year's Loud & Clear report, a transparent loo...
12/03/2025
More than 70% of FAST programming has been produced since 2010, according to new Gracenote report
NEW YORK March 12, 2025 Gracenote, the content data busin...
12/03/2025
GEONA, Nev. Independent digital and linear advertising rep firm Viamedia will adopt cloud-based ShowSeeker Pilot as its primary ad campaign and order management...
12/03/2025
BRUSSELS Mediagenix has announced that Wael Yasin has joined the company as sales director Central Europe....
12/03/2025
LONDON A new study highlights opportunities for shoppable TV and the massive impact online consumer spending is having on the economy, with Omdia predicting tha...
12/03/2025
CUPERTINO, Calif. Interra Systems has announced that Comcast Technology Solutions has integrated recent updates to BATON Version 9 into its operations....