NVIDIA’s “Ampere” Successor Is Reportedly “Hopper,” an MCM-Based GPU Architecture

[Tsing]

NVIDIA's 7 nm "Ampere" products are still behind closed doors, but whispers of its successor can already be heard. Twitter user Kopite7Kimi, who has reported accurate leaks in the past (e.g., RTX SUPER specifications), claims that green team's next-next-generation GPU architecture has been named after computer scientist Grace Hopper.

"Hopper" will supposedly utilize multi-chip modules, which should ring a bell for those who have been following AMD's Zen 2-based Ryzen, Ryzen Threadripper, and EPYC products. Kopite7Kimi says NVIDIA will use these to "build giant cores."

Currently, GPU MCMs are packages that have one GPU die surrounded by memory dies or stacks. This combination of GPU dies could make up "giant cores," at least in the higher end of the performance spectrum. NVIDIA reserves MCMs for only its most expensive Tesla family of compute accelerators, or Quadro professional graphics cards, and seldom offers client-segment GeForce products.

 

[DejaWiz]

Makes sense that future GPUs will consist of scalable cores since it's proven to be a viable solution, thanks to smartphone SoC IC manufacturers and AMD Ryzen/Epyc mobile, desktop, and workstation/server processors.

 

[Maddness]

Interesting, I thought we were still years away from this being viable. I know they had to overcome all the latency issues that don’t seem to affect CPU’s. Am keen to see how they pull this off.