According to a new report by VentureBeat, Nvidia is working with ARM chip maker on technology that enables the creation of more powerful supercomputers. This graphics hardware design company will give the ARM ecosystem access to high-performance software and focus on their AI by 2020.
According to Nvidia CEO Jensen Huang, this will help strengthen the industry to achieve exascale computing computing, systems that can perform one billion (or one quintillion) floating point calculations. floating point operations per second (floating point operations per second).
This is not the first time these two companies have worked together – for example, the Nvidia Tegra X1 chip on the Nintendo Switch gaming console is also an ARM CPU – but this move will help ARM become more competitive in the market. school supercomputers.
Currently the market for ultra-high performance computers is dominated by x86 architectures, with only a few exceptions like the HPE Astra supercomputer launched last year. Now with this partnership, manufacturers will be able to choose ARM for their supercomputer without having to worry that their machines do not support Nvidia's technology.
Nvidia is intensifying its focus on high-performance computing applications such as AI, especially when the demand for their GPUs for pre-coding operations has plummeted. Moreover, this partnership also allows Nvidia to access ARM's more energy efficient chip designs, which are very popular on mobile devices.
Nvidia's CEO said that the factor will be an important limitation of the future, so their support for ARM will help overcome that limit. Nvidia's vice president of accelerated computing also highlighted ARM's openness, saying it would provide an open architecture for supercomputers.
The move is Nvidia's latest move to enter the ultra-high-performance computing market. Earlier, the company spent $ 6.9 billion to acquire Mellanox, Israel's chip maker, which focused on server network equipment. This is also the largest acquisition in Nvidia history.
Refer to The Verge