The release for an upcoming beta build of a popular PC monitoring tool, HWiNFO, gave away some exciting updates regarding Nvidia's future GPUs (thanks, Tom's Hardware). The build release notes for HWiNFO version 7.21 Build 4725 adds support for Nvidia's Ada Lovelace GPUs and a pair of GPUs under Nvidia's new Hopper architecture, which was just announced on Tuesday, and even something from codename Blackwell.
The release notes list five Ada Lovelace GPUs that most likely make up the RTX 40-series with the AD102, AD103, AD104, AD106, and AD107. We recently learned that these GPUs would have a significantly higher shader count than the current RTX 30-series. For example, the AD 102 is rumored to have nearly double the CUDA cores than the yet-to-be-released and wildly expensive RTX 3090 Ti.
The other GPUs mentioned are GH100, GH202, GB100, and GB102. We learned a bit of the H100 since Nvidia released information about the super chip during its GTC event. This enterprise GPU performance will skyrocket past the current flagship A100 super chip. This bump in specs is expected to speed up various tasks in numerous fields like disease diagnosis, quantum simulation, graph analytics, and routing optimizations.
We don't know much about the GH202, but if it's part of the Grace Hopper data center architecture, it might be a more powerful tier of an AI and data center-focused GPU. Nvidia plans on releasing the Hopper Superchips next year.
The GB100 and GB102 are likely the Blackwell GPUs, whose codename we only recently learned about in a leak and would be pretty far out in terms of release window. They could be an entirely new line of GPUs for Nvidia or even a successor for Lovelace. We know that whatever it is, HWiNFO will now support it.
Now that HWiNFO will be able to identify these GPUs, it's only a matter of time before we start seeing more details like performance and specs in the next coming months.