Scoping out the Tenstorrent Wormhole

The Tenstorrent Wormhole n300s PCIe accelerator board is available for purchase, featuring 672 RISC-V cores driving 466 TFLOP/s of FP8 matmul. There is an open source software stack, including kernel driver, user-mode driver and SDK. It'll be a challenge compressing it down to just 20 minutes, but I'll give a description of what the hardware can (and cannot) do, and try to convey how it differs from a conventional GPU. Expect to hear about network-on-chip, ethernet, RISC-V ISA extensions, memory-mapped I/O, and more.