Nvidia Launches World’s Largest GPU To Assist You Construct Your Private Supercomputer

Google+ Pinterest LinkedIn Tumblr +

Yesterday wasn’t nearly Apple Schooling Occasion that launched a shiny, new iPad for training markets. Nvidia CEO Jensen Huang additionally delivered his keynote at Nvidia’s Graphics Expertise Conference and made some impressive announcements.

The most important announcement from the occasion got here within the type of Nvidia’s largest graphics card until date — DGX-2.

This new providing from the GPU big delivers a staggering efficiency with 2 petaflops processing energy, 512GB RAM, and 10 kilowatts energy enter. It weighs about 350 kilos–it goes with out saying that no human can raise it. In comparison with its predecessor, DGX-2 is 10x quicker.

Only for some perspective, you may additionally need to know concerning the current limits of computing. At the moment, the world’s quickest laptop, which runs Linux, can crunch knowledge at 125 petaflops.

What’s extra? It has 200 occasions extra bandwidth of the very best community controller interface (NIC). You additionally get 30 teraflops storage onboard and a couple of CPUs for quickest potential knowledge processing.

From the state, Huang mentioned that you should use this GPU to construct your private supercomputer. Sadly, you’ll be able to’t play your high-end video games on DGX-2. Additionally, it prices $400,000. So, that’s one other draw back.

The corporate additionally confirmed off the brand new Quadro GV100. It’s a high-end professional-grade GPU. Once more, you’ll be able to’t play video games on it but it could excel at deep studying, CGI, knowledge processing, and many others.

It’s the primary workstation GPU primarily based on Nvidia’s Volta structure. With its assist, one can render photos at 1/fifth price, 1/seventh house, and 1/17th energy. It’s anticipated worth is round $7,000.

Learn different fascinating bulletins apart from world’s largest GPU from Nvidia’s occasion here.



Source link

Comments
Share.

Leave A Reply