BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

Nvidia Unveils Powerful Grace AI Supercomputing CPU And Drive Atlan Secure Autonomous Vehicle Platform

Following
This article is more than 2 years old.

At its 2021 GPU Technology Conference (GTC) today, Nvidia CEO Jensen Huang disclosed a bevy of new products on the company’s roadmap that are intended to accelerate machine learning, AI and High Performance Compute (HPC) workloads in applications from supercomputing to big data analytics and autonomous vehicles. There were a number of new technologies announced during Mr. Huang’s keynote today, some of which are centered on ecosystem support of these technologies, along with industry partner commitments. That said, let’s focus on two major chip-level product announcements here, though I’d encourage you to check out all of Nvidia’s disclosures at the company’s virtual GTC ‘21 site where registration is free. In fact, “Nvidia is now a three-chip company,” noted Huang at the introduction of its new ‘Grace’ CPU, marking the announcement of the company’s first primary host processor for specialized applications in support of GPUs (Graphics Processing Units) and in conjunction with DPUs (Data Processing Units).

Enter Nvidia Grace, An Ultra-High End Processor With A Proud Namesake

Nvidia’s Grace CPU is targeted at massive AI training and HPC workloads, and its name is in honor of Grace Hopper, a computer scientist, programmer and US Navy Admiral who is credited for developing a programming language concept based on natural English language. The name Grace is fitting for Nvidia’s new CPU because it specifically is targeted toward applications like AI training on massive NLP (Natural Language Processing) models “with more than 1 trillion parameters,” as Nvidia classifies it.

Built on licensed next generation Arm core technology, Grace is designed to support high-speed Nvidia NVLink interconnects at up to 900GB/s to GPUs with LPDDR5x memory support, in addition to the HBM2e memory on board the GPU modules. Nvidia claims Grace-powered systems will deliver 10x faster throughput versus its high-end legacy DGX AI supercomputer systems that currently run on X86 CPUs. However, in conference with the company, Nvidia representatives were careful to point out that Grace is not designed to compete head-to-head with Intel Xeon and AMD EPYC data center CPUs. Instead, Grace is “designed to be tightly coupled with an Nvidia GPU to remove bottlenecks for the most complex giant model AI and HPC applications.”

Grace will be built on a bleeding-edge 5nm process node and Nvidia also announced that the CPU’s first deployment at the beginning of 2023 will be in the Swiss National Supercomputing Center’s (CSCS) Alps system. Alps will be built by Hewlett Packard Enterprise (HPE) and also employs Nvidia GPUs, Grace CPUs and Nvidia’s HPC SDK. Nvidia notes, “taking advantage of the tight coupling between Nvidia CPUs and GPUs, Alps is expected to be able to train GPT-3, the world’s largest natural language processing model, in only two days — 7x faster than Nvidia’s 2.8-AI exaflops Selene supercomputer, currently recognized as the world’s leading supercomputer for AI by MLPerf.” CSCS researchers will harness the power of Alps for research in natural language understanding of the vast amounts of data available in scientific research papers, to generating new molecules for drug discovery, for example. There were no definitive timelines set, but the Grace-powered Alps system is expected to come online sometime in 2023 and replace CSCS’s current Piz Daint supercomputer.

Nvidia Says New Grace-Powered DRIVE Atlan Autonomous Vehicle Platform Is A Data Center On Wheels

Also infused with ‘Grace-Next’ core CPU support, Nvidia announced its next-generation DRIVE Atlan autonomous vehicle platform that combines all three processing elements (CPU, DPU and GPU) on a single System-On-A-Chip (Soc). Nvidia claims Atlan delivers over 1000 TOPS (trillion operations per second), which is a 4X lift over its existing DRIVE platform, utilizing Arm core technology in its Grace CPU complex, along with ‘Ampere-Next’ GPU resources, deep learning and machine vision accelerators.

Another major feature of DRIVE Atlan is that it’s the first to employ an Nvidia BlueField DPU (Data Processing Unit) for hardware acceleration of vehicle communication and security processing. Nvidia notes Atlan’s BlueField DPU will offer a “zero-trust” security architecture against cyber-attacks, along with intrusion detection and other features critical for the safety and security of self-driving vehicles.

“To achieve higher levels of autonomy in more conditions, the number of sensors and their resolutions will continue to increase,” Huang said. “AI models will get more sophisticated. There will be more redundancy and safety functionality. We’re going to need all of the computing we can get.”

Though Nvidia Atlan won’t sample until 2023 with production vehicles targeted for 2025, the company notes its software is compatible with its previous DRIVE platforms, like DRIVE Orin, which was announced about year ago in conjunction with Mercedes-Benz with vehicle production starting in 2022.

Key Take-Aways And Final Comments

Though there were several other major announcements at GTC ‘21 today, for me Nvidia Grace and Atlan mark the company’s continued laser-sharp focus on advancing AI, machine learning and machine vision technologies. These announcements also underscore the company’s firm commitment to advancing the Arm core technology portfolio, employing it in emerging fields where the company can make major near-term impact in high profit margin markets, like supercomputing, HPC and autonomous driving. And though its new Grace CPU architecture isn’t intended to directly take on Intel and AMD, it’s clear Nvidia is focused on enabling its own high performance processor technology in market segments that are not only key to its core businesses, but also benefit from a tighter coupling of its various compute resources for optimal performance and platform advancement.

These announcements also speak to some of the areas the company will focus on post-acquisition of Arm, should the deal be approved by regulators. Nvidia will continue to evangelize Arm IP and its benefits moving forward for what could be a completely new set of customers around the world.

Follow me on Twitter or LinkedInCheck out my website or some of my other work here