In this deep-dive video, NVIDIA’s lead architect walks through the Blackwell GPU die, explaining the new fifth-generation Tensor Cores, the NVLink interconnect, and the 192GB HBM3e memory stack that makes it the most powerful AI training chip ever built.
Key timestamps: 0:00 Intro, 2:30 Die overview, 8:45 Tensor Cores explained, 15:00 NVLink, 22:10 Real-world benchmarks.

Leave a Reply