FP8 attention: where does the precision break, and how do incoherent processing and per-block scales help?
Three quantization points sit inside FP8 attention, and they fail for different reasons: outlier channels in keys, a probability matrix that lives near zero, and accumulators. The answer simulates each one and says which fix addresses which.
Updated Oct 2026 · Worked answers on inference mechanisms, with calculations, implementation details, and follow-up questions.
Three quantization points sit inside FP8 attention, and they fail for different reasons: outlier channels in keys, a probability matrix that lives near zero, and accumulators. The answer simulates each one and says which fix addresses which.
At batch 1 the step reads every weight once to produce one token per sequence, so arithmetic intensity sits near 1 FLOP per byte and the tensor cores idle.