The core of deep learning is a mathematical operation repeated billions of times, a matrix multiple & add — imagine being able to do an 8-bit multiply & add in a single transistor! That is the power of analog computing that cannot be replicated in any digital system. It’s closer to how our brain computes, with incredible power efficiency.
Revealed today: Mythic is the first and only company that have been able to implement a deep learning model like ResNet 50 in a non-digital architecture: > 50 layers, 1000 fps, 3W total, 9->2ms latency, 8 TOPS/W in 40nm silicon. 10x cost advantage over digital chips.
And keep in mind that a 65x transistor density advantage can still be had by moving from 40nm to a modern 5nm fab.
From the video summary by CEO Mike Henry.
FD: I was their first investor and board member.
Co-dependency of 
Leave a Reply