Nvidia finally showed off working GP100 devices at Siggraph, sort of. While there were no actual running boxes at the show, SemiAccurate thinks there really are working ones in the wild now.
You may recall SemiAccurate’s exclusive news about why the ‘launch’ of GP100 didn’t have any working chips at the GTC shindig. None on stage for Jen-Hsun to show off, none on the show floor, and none to NDA wielding friends in back rooms. They couldn’t even show off a demo recorded on GP100 systems. Something was wrong and that is what we explained.
A little before Computex in early June, SemiAccurate’s moles brought word of their burrows being warmed by the heat from a few GP100 samples that actually powered on non-destructively. While this wasn’t the same thing as functional devices in our hands, it was a sign that things were heading in the right direction. Siggraph moved the needle quite a bit farther with this demo.
Monsters University on 8x GP100s
The demo above was ostensibly running live on 8x GP100s in a dual Xeon system. It was raytracing a scene from Monsters University from Pixar. The demo was interactive and the raytracing was decently fast for a large, complex scene. The reason for our hesitancy is that the box running the demo was back at Nvidia HQ, the demo system was simply a remote desktop head unit. Conspiracy theories aside, we don’t think it was a faked demo this time but there are some things you can infer from the lack of local devices.
Note: The following is for professional and student level subscribers.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Latest posts by Charlie Demerjian (see all)
- Intel shows off 10nm 112Gbps SerDes - Mar 12, 2019
- Intel releases Compute Express Link spec - Mar 11, 2019
- Qualcomm rolls out a second gen 5G modem called X55 - Feb 19, 2019
- What is Intel’s Foveros tech and what isn’t it? - Feb 11, 2019
- Why SemiAccurate called 10nm wrong - Jan 25, 2019