It has been said for quite some time that following Ampere, the next GPU architecture coming from Nvidia would be called Hopper, but according to the latest rumours, the company may have redesigned its roadmap.
The rumour comes from @WildCracks (via VideoCardz), in reply to another tweet from @kopite7kimi about Hopper's delay. Named after the British mathematician Ada Lovelace, this new architecture is expected to use a monolithic design and be based on the 5nm process node.
Apparently, there's some confusion around the next 5nm architecture because Nvidia has been internally using the “ADxxx” codename when referring the new architecture, while externally it's being called Lovelace.
5nm is right
— kopite7kimi (@kopite7kimi) December 21, 2020
The Hopper GPU architecture is expected to be based on the 5nm process node and released after Lovelace. This architecture is rumoured to use a multi-chip design (MCM), but it isn't clear if we will see it reaching the mainstream segment. If it doesn't get to it, Hopper would be an HPC market-only product.
The name “Lovelace” has been previously mentioned in Nvidia's campaign “Company of Heroes”, back at GTC 2018. Back then, there were eight other names of great scientific figures besides Lovelace, including Dijkstra, Knuth, Hamilton, Turing, Sutherland, Boole, Hopper, and Von Neumann. As you know, Turing has already been used, but the rest are still possible names for future Nvidia GPU architectures.
KitGuru says: It could be a while until we get official confirmation on what Nvidia is working on next beyond Ampere, but for now, it looks like we can count on two new GPU architectures, with Lovelace seemingly jumping the queue ahead of Hopper.