NVIDIA’s Next Datacenter-Exclusive Architecture Likely Named After Vera Rubin

NVIDIA’s Next Datacenter-Exclusive Architecture Likely Named After Vera Rubin

The tech community’s resident GPU leaker, kopite7kimi, strikes again: while not explicitly mentioned, it looks like the next-gen datacenter GPU will be given a codename of Vera Rubin’s namesake – according to Videocardz. Hence, the chips themselves will likely be given the letter ‘R’ for model naming. For reference, current-gen datacenter GPUs are named after Grace Hopper, which takes the initial of last name to form model names such as ‘H100’, ‘H200’, and more.

The outlet alleges that kopite7kimi has, in another tweet, mentioned her name with a rather unassuming caption: “Miss the Golden Age of Exploration”. For the uninitiated, Vera Rubin (1928-2016) was an astronomer from the United States, who made discoveries on the galaxy’s rotation rates.  NVIDIA’s GPU codename tradition revolves around known figures in the STEM fields, and this is no exception. The leaker has subsequently confirmed Videocardz’s assumption, though things are bound to change, as with the nature of chip development.

Kopite7kimi also mentioned that the architecture is designed for datacenter, meaning it should be the next datacenter-exclusive GPU architecture after Hopper. Today, NVIDIA only has two such architectures in existence: Volta (precedes Turing), and Hopper (alongside Ampere). However, Blackwell will still slot in between Hopper and Rubin, and at the same time expected to be powering the GeForce lineup in the form of RTX 50 series.

It’s also worth noting that nothing has pointed to the possibility of RTX 60 series using Rubin-based chips, and given the launch timeline of Blackwell, which is slated for late 2024 to early 2025, this seems unlikely – as NVIDIA’s roadmap points to R100/GR100 launching sometime in 2025.

Pokdepinion: It’s a nice way of learning about scientists through NVIDIA’s codenames.