Leela Ratio and wattage

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

modolief
Posts: 44
Joined: Tue Apr 30, 2013 6:29 pm

Leela Ratio and wattage

Post by modolief »

Does the Leela Ratio reflect CPU vs GPU watt usage?

https://blog.lczero.org/2018/07/04/leela-ratio/

I.e., if a computer running Stockfish compared to a computer running Leela Zero registers a Leela Ratio of 1.0, and then we measure power usage of each computer (say using a Kill A Watt meter), will that ratio also be 1.0?
Robert Pope
Posts: 558
Joined: Sat Mar 25, 2006 8:27 pm

Re: Leela Ratio and wattage

Post by Robert Pope »

Of course not.

An obvious counter-example: Suppose you have a 1080 and Leela gets 10,000 NPS. Then you upgrade to an 1170, and by a lucky coincidence, you still get 10,000 NPS. Leela ratio is exactly the same, but the new-generation GPU die shrink uses less power.
User avatar
Guenther
Posts: 4606
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: Leela Ratio and wattage

Post by Guenther »

modolief wrote: Tue Jul 17, 2018 4:14 am Does the Leela Ratio reflect CPU vs GPU watt usage?

https://blog.lczero.org/2018/07/04/leela-ratio/

I.e., if a computer running Stockfish compared to a computer running Leela Zero registers a Leela Ratio of 1.0, and then we measure power usage of each computer (say using a Kill A Watt meter), will that ratio also be 1.0?
The so called "Leela-Ratio" is completely arbitrary and meaningless and of course has nothing to do with power consumption.
Someone simply set the reported nps (average??), which no one knows being correct at all BTW, in the AlphaZeroChess vs. SF8
games as ratio 1. That's all...(more than bogus to me)
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica + AlexChess + Eduard + Sylwy