GTC 2024 has already been huge for Nvidia, revealing the Blackwell chips. These have dominated the tech sphere since they were announced, and now even the smallest amount of information is making headlines. But, the most recent news is one of the biggest yet; the specs for the Blackwell B200 are here.
The specs, revealed by NVIDIA SVP & GPU Architect Jonah Albe and NVIDIA VP of Hyperscale & HPC Ian Buck and reported by Wccftech, presented some key factors about the Blackwell B200. Specifically, the new microarchitecture is vastly superior to the previously used Hopper. Blackwell Architecture employs the 2nd Gen of Transformer Engine technology, adding in FP4 and FP6 compute formats. This is what allows the B200 to be the fastest AI chip on the market.
Impressive specs for the Blackwell B200 are revealed
Also talked about was that the GB200 super chip, the flagship of the Blackwell chip line, has a maximum TDP of 1200W. This makes sense, considering that the chip is a dual-GPU that has the highest computing abilities that Nvidia has ever released. There are two of the B200s in the GB200 super chip, allowing for incredible computing power like never before.
Even the lower-spec variants are nothing to gloss over, though. The Blackwell B200 chip offers 90% of the computing power of the GB200 while having a TDP of 1000W. The Blackwell B100 is the smallest one, with 80% of the computing power, and has a TDP of 700W. Even though the B100 is the smallest chip, it’s nothing to ignore.
These impressive chips will already begin shipping out later this year, to the many eager customers. Unfortunately, the computing power of the B200 chip is meant for AI and enterprises, leaving gamers out of the equation. However, it’s possible this technology could make its way over to future Nvidia GPUs, as technology crosses over. Even if the Blackwell B200 isn’t for gamers, there’s plenty to be excited about.
