Written by: Martin Dale Bolima, Tech Journalist, AOPG.
Once again, NVIDIA is pushing the proverbial envelope further—and not just in one aspect of technology but in many. That was the overriding theme at NVIDIA’s nearly weeklong GTC 2022 event, where the company announced exciting developments in a range of technologies, including Artificial Intelligence (AI), high-performance computing, networking, robotics, design collaboration and digital twins, automotive and even healthcare. And, in what has become a sort of tradition, NVIDIA gave the media first dibs on said announcements via a virtual pre-brief.
Paresh Kharya, Senior Director of Product Management and Marketing at NVIDIA, got the pre-brief rolling, introducing and talking a bit about the company’s new processors and systems, notably its offerings related to Artificial Intelligence (AI), whose next wave, according to NVIDIA, will be fueled by ‘Transformers’.
“Artificial Intelligence has transformed every industry,” Kharya pointed out. “Recently, a type of AI model called ‘Transformer’ has transformed AI. Originally, it was invented by Google, and it became a dominant building block for the neural networks of today. Transformers solved the challenges of deep learning. They brought unsupervised learning and removed the need for labelled data sets. This dramatically expanded the volume of data that could be used to train AI models.”
NVIDIA Hopper H100: Driving AI Forward
This is where NVIDIA comes in. These Transformers, according to Kharya, require higher performance and scalability as they “grow in size and complexity”—meaning, the computing requirements to train large-scale models must be exponentially large as well to keep pace with this always-growing requirement. But that has not always been the case, so much that training AI with Transformers still takes very long. The challenge, therefore, is to dramatically increase performance at smaller scales and then scale that increased performance over thousands of GPUs.
That is the promise of NVIDIA Hopper H100, the new engine for the world’s AI infrastructure, which Kharya teased in the pre-brief and later on announced officially by Jensen Huang, NVIDIA Founder and CEO, at GTC 2022. Hopper H100, named after the American computer scientist and United States Navy Rear Admiral Grace Hopper, features six breakthrough technologies: the world’s most advanced chip, a new Transformer engine, a 2nd-generation secure multi-instance GPU and a 4th-generation NVIDIA NVLink, along with confidential computing (a world’s first) and DPX Instructions to accelerate dynamic programming.
The NVIDIA H100 GPU, which powers Hopper H100, will “offer a giant leap in performance over its predecessor,” according to Kharya. For AI training, specifically, the H100 GPU will offer four petaflops of performance—as much as six times higher than AI 100. Put simply, the H100 GPU is a breakthrough, and it will drive AI forward even more. At the same time, it will also bring unprecedented performance, scalability and security to every data centre.
Grace Superchips: Ushering in the Data Centre of the Future
Another notable breakthrough Kharya touched on during the pre-brief is NVIDIA’s Grace Superchip, designed specifically for both high-performance computing and AI infrastructure. It provides maximum performance and offers as much as twice the memory bandwidth and energy efficiency as opposed to other leading server chips out in the market.
A high-performance CPU for AI and HPC, the Grace Superchip has 144 Arm cores in a single socket, providing an industry-leading estimated performance of 740 on the SPECrate®2017_int_base. It also boasts an advanced memory subsystem that consists of the world’s first LPDDR5x memory with Error Correction Code, enabling a balance of speed and power consumption since it offers double the bandwidth of traditional DDR5 designs (1 terabyte per second) but consumes dramatically less power—at just 500 watts for the CPU and memory combined.
With these features, expect the Grace Superchip to make light work of even the most demanding applications, whether in AI, HPC, data analytics and even scientific and hyperscale computing. It will run all of NVIDIA’s computing software stacks, and it can be configured into servers either as a standalone CPU-only system or as a GPU-accelerated server for optimised performance.
NVIDIA AI: Democritising AI
Aside from introducing Hopper H100, the pre-brief also covered developments in NVIDIA AI, the company’s cloud-native AI initiative that aims to democratise AI for enterprises worldwide.
“Only NVIDIA provides the full stack of AI, and the foundation of this stack is the AI infrastructure,” noted Erik Pounds, Director of Product Marketing at NVIDIA. “Available in all public clouds, there are systems available from our partners to construct an AI Center of Excellence to enhance existing enterprise data centres for AI. NVIDIA AI software is the operating system of AI. For example, NVIDIA TAO is used for transfer learning or RAPIDS for data science, TensorFlow and PyTorch for AI training, Triton inference Server for AI inferencing and more.”
Pounds also gave some notable use cases of NVIDIA AI, starting with Amazon Search, which uses Triton and TensorFlow to optimise and run models that provide real-time spell checking for users. Microsoft, on the other hand, uses Triton to achieve greater model accuracy for better translation services, while NTT leverages the NVIDIA AI enterprise software suite to have an AI-ready platform for all of its developers. Snapchat, meanwhile, is using NVIDIA Riva to help creators build immersive experiences in their Snap Lens Studio.
Additionally, Pounds spoke about the enhancements to the NVIDIA® AI Accelerated program, which Huang formally launched later in the week. The improvements will bolster the program’s performance and improve the reliability of the various AI applications developed by NVIDIA’s software and solution partners—among them, Adobe, Red Hat and VMware. NVIDIA’s AI Accelerated program also increases the visibility of the company’s AI-accelerated applications, helping enterprise customers deploy on the NVIDIA AI platform with much confidence.
Driving Towards a More Advanced Future With NVIDIA DRIVE
Underscoring NVIDIA’s encompassing reach is its presence even in the automotive industry via NVIDIA DRIVE.
“NVIDIA is a platform company, and DRIVE is our platform for autonomous vehicles,” noted Danny Shapiro, Vice President of Automotive at NVIDIA. “We’ve been innovating in the transport industry for well over a decade. Inside the car, we are developing an AI public chauffeur and a concierge, making mobility safer and more convenient. This unique platform, which starts in the data centre is full-stack and end-to-end. It’s open for developers to use in whole or in parts.”
DRIVE, according to Shapiro, is as complete as it gets, with technologies that deliver “a wide range of capabilities.” This is why it is being adopted by the entire transport industry. In fact, DRIVE has a leadership position in all segments, with 20 out of 30 passenger electric vehicle makers leveraging DRIVE, 7 out of 10 trucking companies relying on it, 8 out of 10 Robotaxi brands powered by it and 28 out of 30 autonomous vehicle data centres running it.
“An enormous number of companies rely on NVIDIA to develop their vehicles,” Shapiro added, unveiling a slide outlining some of these companies, including global brands such as Mercedes, Jaguar, Volvo and Land Rover, as well as Asian powerhouses Hyundai and Li Auto. “This is just a snapshot of some of those companies [referring to the list presented]. Many others will be unveiled soon.”
The DRIVE ecosystem, in short, is growing exponentially, comprising an automotive pipeline that has already exceeded USD $11 Billion—a nearly 33% increase from last year’s USD $8 Billion. This underscores NVIDIA’s now primordial role in the world of automotive, and it will only grow in the years ahead.
A Fitting Prelude to a Week Full of Exciting Announcements
The virtual media pre-brief was a fitting appetiser to the main course that was the GTC 2022, where Huang and other NVIDIA executives formally announced many of the innovations and breakthroughs—including NVIDIA’s very own Omniverse—that the company will soon be offering to the world. To find out more about these announcements, check out the official GTC link here.
Archive
- October 2024(44)
- September 2024(94)
- August 2024(100)
- July 2024(99)
- June 2024(126)
- May 2024(155)
- April 2024(123)
- March 2024(112)
- February 2024(109)
- January 2024(95)
- December 2023(56)
- November 2023(86)
- October 2023(97)
- September 2023(89)
- August 2023(101)
- July 2023(104)
- June 2023(113)
- May 2023(103)
- April 2023(93)
- March 2023(129)
- February 2023(77)
- January 2023(91)
- December 2022(90)
- November 2022(125)
- October 2022(117)
- September 2022(137)
- August 2022(119)
- July 2022(99)
- June 2022(128)
- May 2022(112)
- April 2022(108)
- March 2022(121)
- February 2022(93)
- January 2022(110)
- December 2021(92)
- November 2021(107)
- October 2021(101)
- September 2021(81)
- August 2021(74)
- July 2021(78)
- June 2021(92)
- May 2021(67)
- April 2021(79)
- March 2021(79)
- February 2021(58)
- January 2021(55)
- December 2020(56)
- November 2020(59)
- October 2020(78)
- September 2020(72)
- August 2020(64)
- July 2020(71)
- June 2020(74)
- May 2020(50)
- April 2020(71)
- March 2020(71)
- February 2020(58)
- January 2020(62)
- December 2019(57)
- November 2019(64)
- October 2019(25)
- September 2019(24)
- August 2019(14)
- July 2019(23)
- June 2019(54)
- May 2019(82)
- April 2019(76)
- March 2019(71)
- February 2019(67)
- January 2019(75)
- December 2018(44)
- November 2018(47)
- October 2018(74)
- September 2018(54)
- August 2018(61)
- July 2018(72)
- June 2018(62)
- May 2018(62)
- April 2018(73)
- March 2018(76)
- February 2018(8)
- January 2018(7)
- December 2017(6)
- November 2017(8)
- October 2017(3)
- September 2017(4)
- August 2017(4)
- July 2017(2)
- June 2017(5)
- May 2017(6)
- April 2017(11)
- March 2017(8)
- February 2017(16)
- January 2017(10)
- December 2016(12)
- November 2016(20)
- October 2016(7)
- September 2016(102)
- August 2016(168)
- July 2016(141)
- June 2016(149)
- May 2016(117)
- April 2016(59)
- March 2016(85)
- February 2016(153)
- December 2015(150)