rocm

192 GB and 5.3 TB/s per card beat the H100 on both memory numbers at a lower hourly price. Whether that wins in production is a question about achieved bandwidth and ROCm kernel coverage, and there is a short list of numbers to run before you sign the order.

The source translation is a script and takes an afternoon. The assumptions underneath it are the problem: a wavefront is 64 lanes rather than 32, so a warp-level reduction written for 32 ports cleanly and reduces half the data. What breaks silently, what breaks loudly, and the numbers that change every tuning decision.

The weights are portable and almost nothing else is. Four layers have to exist before the model runs at all, the quantization format is the one most likely to be missing, and the honest plan states what will not work in the first quarter rather than promising parity.