4 comments

[ 71.8 ms ] story [ 747 ms ] thread
This is a very interesting read. I wonder if Google has an in-house fabrication facility for all of its TPU's.
They don't; they probably use TSMC.
Their version of fp16 (bfloat as they call it) is very interesting. 7 bits of mantissa is only about 2 decimals of precision!
The TPU 2 was about half the cost of using Nvidia for same work.

https://medium.com/@8fee9a760280/c2bbb6a51e5e

Be interesting how much further ahead Google is now with the 3.

But most impressive is able to offer Wavenet at a competitive cost to the old technique for TTS used by everyone else.

16k through a NN in real time is just hard to believe possible. Nvidia has their work cut out for them.