Nvidia will support ARM-based CPUs for new, energy-efficient supercomputers

Boosting the community to exascale.

Photo by Sam Byford / The Verge

Nvidia is working with chip designer ARM on technology that’s designed to allow for more energy-efficient supercomputers. VentureBeat reports that the graphics company will give the ARM ecosystem access to all its high-performance and AI-focused software by 2020, in a move which Nvidia’s CEO Jensen Huang said will help to “boost” the industry to achieve so-called exascale computing, systems capable of a billion billion (or a quintillion) floating point operations a second. The announcement means that Nvidia now supports all major CPU architectures.

This isn’t the first time the two companies have worked together — the Nvidia Tegra X1 chip that powers the Nintendo Switch has an ARM CPU for example — but this move makes ARM more competitive in the supercomputer market. High-performance computers have tended to be dominated by x86-based technologies, with a few key exceptions like last year’s HPE Astra supercomputer. Now manufacturers will be able to choose without having to worry that an ARM machine won’t be able to support Nvidia’s technology.

INTEL’S X86 PROCESSORS DOMINATE THE HIGH-PERFORMANCE COMPUTING MARKET

Nvidia is increasingly focused on high-performance computing for applications such as AI, especially as demand for its GPUs for cryptocurrency mining has declined. The benefit to Nvidia of this partnership will be access to ARM’s more energy-efficient chip designs, which most commonly power mobile devices. Nvidia’s CEO says that power will be a key limit in the future, which its support for ARM will help to overcome. Nvidia’s vice president of accelerated computing also praised ARM’s openness, saying it provided an open architecture for supercomputing.

The move is just the latest push into the high-performance computing market for Nvidia, which recently paid $6.9 billion for Mellanox, an Israeli chipmaker focused on server networking. The acquisition was the largest in Nvidia’s history.

Source: Jon Porter|@JonPorty (The Verge).
~Best Feeds ™...