Part 1 got a certain amount of discussion on HN about 3 months back [0]. As well as a different set of workloads, one significant point of interest to this one is that it appears the high idle package power they saw back then was due to the board/system, not the chip. There'd been a bunch of speculation about what changes in Milan might have caused that, but it just being the system all along means the value is even better. The extra power can then go to cores too resulting in a notable 7-13% performance improvement in some workloads. Interesting to me anyway and surprising if only in that I'd have expected AMD to want to put their launch in the best light.

----

0: https://news.ycombinator.com/item?id=26466933