Twin 2080Tis and large NN

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
shrapnel
Posts: 1221
Joined: Fri Nov 02, 2012 8:43 am
Location: New Delhi, India

Twin 2080Tis and large NN

Post by shrapnel » Fri Oct 25, 2019 4:11 pm

Large NNs (384x30) really do improve the performance of lc0( whether Laskos Sir fully agrees or not ). Of course, twin 2080 Tis help :) , but on the other hand using a large NN with a weak GPU makes no sense and would simply backfire.
Following is a Game against a guy using 16 Cores and Stockfish..



Against a slightly weaker opponent, lc0 won, even though it was a 3 minute Game, supposedly not lc0's forte....

i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis

shrapnel
Posts: 1221
Joined: Fri Nov 02, 2012 8:43 am
Location: New Delhi, India

Re: Twin 2080Tis and large NN

Post by shrapnel » Fri Oct 25, 2019 4:51 pm

A Game against the popular, much-vaunted Eman Engine, a Stockfish Clone I think....

i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis

shrapnel
Posts: 1221
Joined: Fri Nov 02, 2012 8:43 am
Location: New Delhi, India

Re: Twin 2080Tis and large NN

Post by shrapnel » Fri Oct 25, 2019 5:04 pm

i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis

mwyoung
Posts: 1641
Joined: Wed May 12, 2010 8:00 pm

Re: Twin 2080Tis and large NN

Post by mwyoung » Fri Oct 25, 2019 5:20 pm

shrapnel wrote:
Fri Oct 25, 2019 4:11 pm
Large NNs (384x30) really do improve the performance of lc0( whether Laskos Sir fully agrees or not ). Of course, twin 2080 Tis help :) , but on the other hand using a large NN with a weak GPU makes no sense and would simply backfire.
Following is a Game against a guy using 16 Cores and Stockfish..



Against a slightly weaker opponent, lc0 won, even though it was a 3 minute Game, supposedly not lc0's forte....

Thanks for the games. It looks like the best CPU you played was a older 8 core 2700x CPU, and this game was a draw. I have not seen any data that suggest using multi GPUs with Lc0 is worth the investment at this point. But this could change.... Right now SF is just smoking Lc0 and the other NN in my testing.

Stockfish and a good CPU provides the best chess value and strength.
Professing themselves to be wise, they became fools,
Take on me. foes 0

shrapnel
Posts: 1221
Joined: Fri Nov 02, 2012 8:43 am
Location: New Delhi, India

Re: Twin 2080Tis and large NN

Post by shrapnel » Fri Oct 25, 2019 5:28 pm

A Loss on Time with Black to the mighty IpMan running a 44 Core CPU :shock: . For much of the Game, lc0 was on equal Terms and lost only because of Time pressure I think....No doubt lc0's Time Management needs improvement
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis

mwyoung
Posts: 1641
Joined: Wed May 12, 2010 8:00 pm

Re: Twin 2080Tis and large NN

Post by mwyoung » Fri Oct 25, 2019 5:30 pm

shrapnel wrote:
Fri Oct 25, 2019 5:04 pm
I guess you were still posting games... This is the 16 core game.
Professing themselves to be wise, they became fools,
Take on me. foes 0

shrapnel
Posts: 1221
Joined: Fri Nov 02, 2012 8:43 am
Location: New Delhi, India

Re: Twin 2080Tis and large NN

Post by shrapnel » Fri Oct 25, 2019 5:33 pm

mwyoung wrote:
Fri Oct 25, 2019 5:20 pm
for the games. It looks like the best CPU you played was a older 8 core 2700x CPU, and this game was a draw. I have not seen any data that suggest using multi GPUs with Lc0 is worth the investment at this point. But this could change.... Right now SF is just smoking Lc0 and the other NN in my testing.

Stockfish and a good CPU provides the best chess value and strength.
I play with the Players I get ; I can't help what CPU they use.
Its early days yet, I hope to prove you wrong as Sergio's NNs can only get better....
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis

mwyoung
Posts: 1641
Joined: Wed May 12, 2010 8:00 pm

Re: Twin 2080Tis and large NN

Post by mwyoung » Fri Oct 25, 2019 5:35 pm

shrapnel wrote:
Fri Oct 25, 2019 5:33 pm
mwyoung wrote:
Fri Oct 25, 2019 5:20 pm
for the games. It looks like the best CPU you played was a older 8 core 2700x CPU, and this game was a draw. I have not seen any data that suggest using multi GPUs with Lc0 is worth the investment at this point. But this could change.... Right now SF is just smoking Lc0 and the other NN in my testing.

Stockfish and a good CPU provides the best chess value and strength.
I play with the Players I get ; I can't help what CPU they use.
Its early days yet, I hope to prove you wrong as Sergio's NNs can only get better....
I hope you prove me wrong also. Stockfish right now is a tough nut to crack for Lc0. Keep posting results....
Professing themselves to be wise, they became fools,
Take on me. foes 0

shrapnel
Posts: 1221
Joined: Fri Nov 02, 2012 8:43 am
Location: New Delhi, India

Re: Twin 2080Tis and large NN

Post by shrapnel » Fri Oct 25, 2019 5:56 pm

OK ! A Draw against the mighty 44 Core IpMan ! No question of lco "getting smoked" provided Two or more 2080 Tis are used. You are only using one I think. I only hope I don't "smoke" my two 2080 Tis, the heat being produced is stunning. I'm running all Case/System fans , a Table Fan and the Air-Conditioning on full blast.... OK, enough for today, have to go to Work early tomorrow morning.

i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis

corres
Posts: 1681
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Twin 2080Tis and large NN

Post by corres » Fri Oct 25, 2019 7:32 pm

I made some short matches between 42872 (256x20) net and 384x30-t40-970 (big) net.
GPUs 2 x RTX 2060 OC, its chess power more than 1 x RTX 2080Ti OC
1. match
Time = 1 min + 2 sec / move
Result
42872net : 384x30-t40-970net = 9 : 1 (20 draw) 2 x 15 = 30 games
2. match
Time = 1 min + 20 sec / move
Result
42872net : 384x30-t40-970net = 3 : 0 (27 draw) 2 x 15 = 30 games
Opening positions was the same in both matches.
Leela parameters were also the same in the case of the two net.
So enhancing the move time (or the power of GPUs) the result tend to reach equality.
It would be interesting to make a match with Time = 1 min + 200 sec / move (~ FIDE tempo)

Post Reply