4 comments

[ 2.3 ms ] story [ 20.2 ms ] thread
So mainly a little more SRAM (320K -> 512K) and faster machine learning workloads?
Which might still make a big difference and avoid switching to another chip.