Skip to content

Comment on The Nvidia DGX-1 Deep Learning Supercomputer in a Box

Comments

Wait, how many chips did they cram in there that they're getting 170 TFlops. Even at a very generous 10 TFLOP per chip that would be 17 chips.

NVIDIA Tesla P100 has 21 TeraFLOPS of FP16 performance by their words. So they got 8 chips there.

Yep, they showed a diagram of how it fits together: http://i.imgur.com/xk1daFG.jpg

I wish that made that information more accessible. I wasn't able to find it on the site and it was all I really cared about.

Ah, half-floats. That explains it. Still pretty high but realistic at least.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.