But on the other side of the coin naive GPU acceleration ends up with poor latency despite ‘not dropping a frame’ 5ms input latency is a dubious claim, what re

But on the other side of the coin naive GPU acceleration ends up with poor latency despite ‘not dropping a frame’ 5ms input latency is a dubious claim, what really matters is input to photon / end-to-end latency