LCZero: Progress and Scaling. Relation to CCRL Elo

[Werewolf]

On my XEON 2670 with GTX 1070Ti, LC0 wins the match using the 20 opening positions from John Nunn against stockfish9 strength 15 (1 core for stockfish9).

23-17 (19-13-8; +56)

LC0 is very strong with a fast GPU.

Yes. Titan V anyone..?

[main line]

Come in playchess - Engine room and try in long time control 120 min + 15 sec. But one question:
-I read somewhere that lczero.exe for GPU could to use only in learning matches and for playing is lczero.exe for processors. Is it true or does your lczero combine power of graphic card and processor ?
If you don't have openbook, I will send you unbeatable book for playchess.

[Werewolf]

Come in playchess - Engine room and try in long time control 120 min + 15 sec. But one question:
-I read somewhere that lczero.exe for GPU could to use only in learning matches and for playing is lczero.exe for processors. Is it true or does your lczero combine power of graphic card and processor ?

It can use either. I use GPU as it's faster.

[main line]

If so, what is stronger:
1. lczero.exe + processor 1000$ or
2. lczero.exe + graphics 1000$ ?

[Hai]

If so, what is stronger:
1. lczero.exe + processor 1000$ or
2. lczero.exe + graphics 1000$ ?

2. is ~10 times stronger than 1.

[main line]

All others engines are "processor" version but lczero is "graphcs" player.
Seems it is fight between mathematician (Stockfish) vs painter (Alpha Zero), Albert Einstein against Leonardo da Vinci.

[duncan]

As of now, the progress does not seem to be plateauing. If it continues in the same fashion, and the framework continues its work, by the 20th of April we can expect LCZero to be about 2750 CCRL Elo points.

still on target ?

[Laskos]

As of now, the progress does not seem to be plateauing. If it continues in the same fashion, and the framework continues its work, by the 20th of April we can expect LCZero to be about 2750 CCRL Elo points.

still on target ?

Well, my initial assessments of this engine were complete rubbish, but 2750 CCRL Elo seems already achieved, say on a very strong GPU like Nvidia 1080 and LTC of say 1 minute per move.

As of now, it seems to progress again slowly, by my preliminary results with meager CPU means and short time control of 1s/move, maybe these results are not very representative. I will post my results maybe this night or tomorrow.

[Hai]

As of now, the progress does not seem to be plateauing. If it continues in the same fashion, and the framework continues its work, by the 20th of April we can expect LCZero to be about 2750 CCRL Elo points.

still on target ?

Well, my initial assessments of this engine were complete rubbish, but 2750 CCRL Elo seems already achieved, say on a very strong GPU like Nvidia 1080 and LTC of say 1 minute per move.

As of now, it seems to progress again slowly, by my preliminary results with meager CPU means and short time control of 1s/move, maybe these results are not very representative. I will post my results maybe this night or tomorrow.

And now think a little about how strong it is with 2x Nvidia 1080 ti .
It will win easily in the weakest division of TCEC season 12.
Also because it's long time control and even with increment .

[Werewolf]

As of now, the progress does not seem to be plateauing. If it continues in the same fashion, and the framework continues its work, by the 20th of April we can expect LCZero to be about 2750 CCRL Elo points.

still on target ?

Well, my initial assessments of this engine were complete rubbish, but 2750 CCRL Elo seems already achieved, say on a very strong GPU like Nvidia 1080 and LTC of say 1 minute per move.

As of now, it seems to progress again slowly, by my preliminary results with meager CPU means and short time control of 1s/move, maybe these results are not very representative. I will post my results maybe this night or tomorrow.

I had to finish my match early because I need my PC back.

Sadly it didn't go as well as I expected.

LCZero 127 on Nvidia 1060 @ 15 sec/move
Colossus 2008b @15 sec/move

31.5/86

LCZero was dominating in the first third of the match but then went downhill really badly. Made me think there was either a bug or the hash (or whatever it uses) needs clearing.