TalkChess.com

I don't speak Japanese but their architecture seems to be: W1 = 125388 x 256 W2 = 512 x 32 W3 = 32 x 32 W4 = 32 x 1 And I can make some educated guesses: they exploit the fact that W1 doesn't change much to compute the result of that layer incrementally, which is coincidentally the heaviest layer. T...

Another consideration is noise, the 180W thermals from the 10980XE can be cooled down very quietly with a Noctua NH-D15, I have two systems with these running and I can hardly hear that they are turned on. For a TR3-3970X you will start at 280W (probably even more), a 64 core TR3 will be like 500W,...

https://twitter.com/trav_downs/status/1 ... 7962364928

It was already known they were slow, but I didn't see anyone point out before that how slow they get actually depends on the argument.

RubiChess wrote: ↑

Wed Nov 27, 2019 5:46 am

Remember that they still use last official SF10 for testing

They do not.

Jouni wrote: ↑

Mon Nov 25, 2019 5:15 pm

Chessbase don't waste time https://en.chessbase.com/post/optimizing-fat-fritz .

I am sure Albert Silver (who wrote the article) is very proud he wrote the strongest engine in the world.

>Assuming the "steps" they refer to are games

It's training steps of 2048 minibatches. I don't see any mention of games - they run the processes asynchronously. But it seems they generated about 20 billion positions. So, probably quite a bit more than a million games.

https://www.sp-cc.de/lc0-testing.htm

This shows Fat Fritz being 40-50 Elo weaker? Rather incompatible results, unless 49921 is a very weak net or something.

You can use the logit of the probability of winning, and multiply it by a constant so that it looks like centipawns. I do this for my program, and it works very well. I use this, which is the same thing right? https://www.chessprogramming.org/Pawn_Advantage,_Win_Percentage,_and_Elo It's convenient ...

I missed this discussion. So what is agreed up on: info wdl 200 600 200 format for uci, and what for xboard? Thanks. I don't think anything has been agreed upon, but Fritz 17 is now confirmed to ship with a modified lc0 that has the patch pointed out in the original post: https://github.com/DanielU...

GPU: RTX 2070 May network versus elfv2: 10:10 October network versus elfv2: 12:8 So, the "700 Elo points" in self ratings is hugely inflated, and the progress is probably no more than 100 Elo points in 5 months, quite modest for Go. It's hard to say anything from such a short match, no? I think the...