I saw very few matches between this 2 engines.
Conditions like :
3min+2sec
200 games
book : short lines (where Lc0 was the best over A/B engines)
RTX 2080 Ti for Lc0
16 cores for SF-NNUE (latest exe + latest NN file)
Could someone run a match like this ?
Thanks,
Vincent
How does perform SF-NNUE vs Lc0 ?
Moderators: hgm, Rebel, chrisw
-
- Posts: 5228
- Joined: Thu Mar 09, 2006 9:40 am
- Full name: Vincent Lejeune
-
- Posts: 3293
- Joined: Wed Mar 08, 2006 8:15 pm
-
- Posts: 28
- Joined: Fri Jul 26, 2013 6:24 am
Re: How does perform SF-NNUE vs Lc0 ?
SF NNUE is just obliterating Lc0. Think with me. If SF was already beating Lc0 as demonstrated at TCEC...imagine now that it was improved in about one weak +110 ELO!!!
-
- Posts: 5228
- Joined: Thu Mar 09, 2006 9:40 am
- Full name: Vincent Lejeune
-
- Posts: 5228
- Joined: Thu Mar 09, 2006 9:40 am
- Full name: Vincent Lejeune
Re: How does perform SF-NNUE vs Lc0 ?
For the last match, a book with long lines was used. This reduces the Lc0 power.
I prefer to see results than use my imagination
-
- Posts: 2041
- Joined: Wed Mar 08, 2006 8:30 pm
-
- Posts: 2444
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
Re: How does perform SF-NNUE vs Lc0 ?
I am running a 300 games match between Lc0 0.26.1 Leelenstein 15 vs. SF 200807 9931db908a9b with Noomen lowdraw-openings (selected by J.Noomen out of his TCEC-openings (you can not play a match like this with short opening-lines, that would lead to draw-rates around 90%)) with long time: 5min+3sec for Lc0 (and +50% more time for SF nnue, to get a Leela-ratio of 1.0 (measured with SF 8 and Lc0 with old 11260 net, like Leela-ratio should be calculated), because my CPU is not so fast, compared to my RTX 2060).Vinvin wrote: ↑Wed Aug 12, 2020 12:09 am I saw very few matches between this 2 engines.
Conditions like :
3min+2sec
200 games
book : short lines (where Lc0 was the best over A/B engines)
RTX 2080 Ti for Lc0
16 cores for SF-NNUE (latest exe + latest NN file)
Could someone run a match like this ?
Thanks,
Vincent
Result in 3-4 days.
At the moment, it looks like a clear win for SF nnue.
-
- Posts: 3553
- Joined: Thu Jun 07, 2012 11:02 pm
-
- Posts: 2727
- Joined: Wed May 12, 2010 10:00 pm
Re: How does perform SF-NNUE vs Lc0 ?
What kind of electric bill do you think you will have? My power is cheap, if I run at full power 24/7 it will cost me less then 30 bucks a month. With a 2950x and a 2080 ti. It cost me more to run my pool pump. That explains CCRL testing.Modern Times wrote: ↑Thu Aug 13, 2020 9:33 pmI was never tempted to spend hundreds of pound on a high-end GPU and new power supply to run it. Not to mention the electricity bills.
"The worst thing that can happen to a forum is a running wild attacking moderator(HGM) who is not corrected by the community." - Ed Schröder
But my words like silent raindrops fell. And echoed in the wells of silence.
But my words like silent raindrops fell. And echoed in the wells of silence.
-
- Posts: 2727
- Joined: Wed May 12, 2010 10:00 pm
Re: How does perform SF-NNUE vs Lc0 ?
I can run this test. I have the exact hardware. This will be my next test. So do you want it to play the offical Stockfish NNUE with NNUE enabled? If no answer this is what will be played.Vinvin wrote: ↑Wed Aug 12, 2020 12:09 am I saw very few matches between this 2 engines.
Conditions like :
3min+2sec
200 games
book : short lines (where Lc0 was the best over A/B engines)
RTX 2080 Ti for Lc0
16 cores for SF-NNUE (latest exe + latest NN file)
Could someone run a match like this ?
Thanks,
Vincent
"The worst thing that can happen to a forum is a running wild attacking moderator(HGM) who is not corrected by the community." - Ed Schröder
But my words like silent raindrops fell. And echoed in the wells of silence.
But my words like silent raindrops fell. And echoed in the wells of silence.