Nvidia details Grace CPU Superchip, an Arm-based server processor

Missed a session at the Data Summit? View on demand here.

Early last year, Nvidia unveiled the Grace processor, the company’s inaugural data center ARM-based CPU for AI and high-performance computing applications. But at the time, Nvidia declined to reveal some key details about Grace, including the number of transistors in the chip and where it might be used. That changed today at the company’s March 2022 GTC event, where Nvidia got the Grace CPU Superchip, the first discrete chip in the Grace portfolio, off the ground.

Nvidia’s Grace CPU Superchip packs 144 Arm cores into a single socket. It consists of two CPU chips connected via Nvidia’s NVLink-C2C – a new chip-to-chip interconnection – and is designed to complement the Grace Hopper Superchip, which pairs a Grace CPU with Nvidia’s newly announced op Hopper architecture. based GPUs.

“A new type of data center has emerged: AI factories that process and refine mountains of data to produce intelligence,” Nvidia co-founder and CEO Jensen Huang said in a press release. “The Grace CPU Superchip offers the highest performance, memory bandwidth and Nvidia software platforms in one chip and will shine as the CPU of the world’s AI infrastructure.”

Arm-based architecture

The Grace CPU Superchip is based on Arm’s Neoverse family. Neoverse was first introduced in 2018 and includes the technologies of the Arm ecosystem that span from the edge to the cloud, especially those that focus on the data center.

The specific architecture is Arm’s Armv9, which made its debut in May 2021. The Grace CPU Superchip — which consumes about 500 watts of power according to Nvidia — has a memory subsystem made up of LPDDR5x memory that provides bandwidth speeds of up to 1 terabyte per second.

Nvidia’s Grace CPU Superchip.

Combined with Nvidia’s ConnectX-7 NICs, the Grace CPU Superchip can be configured in servers as standalone CPU-only systems or as GPU-accelerated servers with one, two, four or eight Hopper-based GPUs. Nvidia says it is working with customers in high-performance compute, supercomputing and cloud on applications for the Grace CPU Superchip, which is expected to be available — along with the Grace Hopper Superchip — in the first half of 2023.

Upcoming market

While the data center market is largely dominated by x86 processors from Intel and AMD — Intel’s server market share is estimated to be over 90% — the power efficiency of ARM chips has made them an increasingly attractive alternative to enterprise customers. Amazon Web Services uses Arm-based chips like Graviton3 in its data centers, while supercomputers like Fujitsu’s Fugaku use Arm processors in conjunction with GPUs.

Nvidia previously announced that Grace will be installed for the first time in 2023 in supercomputers designed by Hewlett Packard Enterprise HPE for the Swiss National Supercomputing Center and the US Department of Energy’s Los Alamos National Laboratory. And in 2025, Nvidia plans to introduce the next generation of Grace architecture, codenamed Grace Next, ahead of a new GPU family.

As analyst Roger Entner noted in a 2021 piece for Fierce Electronics, several companies — including Marvell, Ampere and Huawei — have attempted to make an impact with Arm chips in the server market. Internally, Amazon, Microsoft, Baidu, Alibaba, Facebook and others have designed custom Arm-based chips tailored to their own data centers and needs. But not all of these efforts were successful. Huawei spent about $3 billion to fund a development platform and ecosystem around its Arm server chips, while Qualcomm was forced to leave the space after attempting to do the same.

“The Grace chip is designed with a common Arm-cores design that is available to everyone [Arm] licensee and not specially adapted to work with Nvidia GPUs,” noted Entner. “If Nvidia is allowed to buy Arm for $40 billion, the combined company would have many reasons to build specialized CPUs and an integrated architecture that will be significantly more efficient than standard ARM cores, giving the larger Nvidia a greater edge over the server. market.”

Of course, Nvidia ultimately declined to go ahead with the planned acquisition of Arm, citing regulatory pressure. Still, Nvidia’s recent moves, along with its BlueField DPUs, which also tap into Arm-based chips for processing, show that it’s serious about making waves in a server processor market that could be worth $17.89 billion by 2026.

“Leading AI and data science are pushing the boundaries of today’s computing architecture by processing unimaginable amounts of data,” Huang said at the unveiling of the Grace portfolio last year. “With the aid of licensed ARM[intellectualproperty[heeftNvidiaGraceontworpenalseenCPUdiespecifiekisvoorgrootschaligeAIenHPCIncombinatiemetdeGPUenDPUgeeftGraceonsdederdefundamenteletechnologievoorcomputergebruikendemogelijkheidomhetdatacenteropnieuwteontwerpenomAIvooruittehelpenNvidiaisnueenbedrijfmetdriechips”[intellectualproperty[NvidiahasdesignedGraceasaCPUspecificallyforgiant-scaleAIandHPCCoupledwiththeGPUandDPUGracegivesusthethirdfoundationaltechnologyforcomputingandtheabilitytore-architectthedatacentertoadvanceAINvidiaisnowathree-chipcompany”[intellectueeleigendom[heeftNvidiaGraceontworpenalseenCPUdiespecifiekisvoorgrootschaligeAIenHPCIncombinatiemetdeGPUenDPUgeeftGraceonsdederdefundamenteletechnologievoorcomputergebruikendemogelijkheidomhetdatacenteropnieuwteontwerpenomAIvooruittehelpenNvidiaisnueenbedrijfmetdriechips”[intellectualproperty[NvidiahasdesignedGraceasaCPUspecificallyforgiant-scaleAIandHPCCoupledwiththeGPUandDPUGracegivesusthethirdfoundationaltechnologyforcomputingandtheabilitytore-architectthedatacentertoadvanceAINvidiaisnowathree-chipcompany”

VentureBeat’s mission is to be a digital city square for tech decision makers to learn about transformative business technology and transactions. Learn more

This post Nvidia details Grace CPU Superchip, an Arm-based server processor

was original published at “https://venturebeat.com/2022/03/22/nvidia-details-grace-cpu-superchip-an-arm-based-server-processor/”