First Win by Leela Chess Zero against Stockfish dev

[ankan]

2 wins in a match of 50 games:
https://lichess.org/L6Vpgi3w
https://lichess.org/8ulwmKUl

Time control: 5 min + 10 seconds per move.
4GB hash size, No opening book for lc0, Arena\Books\olympiad.abk for Stockfish, no tablebases

lc0-win-20180604-cuda92-cudnn714-experimental running on Titan V (network #374, additional options: --minibatch-size=512 --slowmover=2.5).
against
Recent stockfish dev running on 4 cores of i7 5820k (I know its not the best cpu you can get for stockfish)

Final result:
Lc0-cudnn - Stockfish_18052419_x64_bmi2 : 15.5/50 2-21-27 (=0=0=000==0====0===0=====0=0001=1=00=0=0=0=0==00=0?) 31% -139

[shrapnel]

Awesome Results, Ankan bro !
I envy you your Titan V !
Where in India are you ? Kolkatta ?
Still, I suppose, my measly 1080 Ti will start getting your kinda Results hopefully within a week or so, as the Network IDs steadily improves !
India RULEZ !

[Ozymandias]

I envy you your Titan V !

You'll be happy to hear that the 1180 is supposed to be almost as fast and way cheaper.

Peak FP32 compute performance is expected to be around 13 TFLOPS, depending on how high of a clock speed the graphics card will be able to hit and how often. This puts it slightly ahead of the existing GeForce GTX Titan Xp and GTX 1080 Ti. So you can expect around GTX Titan Xp performance or slightly better, but at significantly lower power consumption.

[Dann Corbit]

And AMD is sampling a 7nm card.

[Werewolf]

And AMD is sampling a 7nm card.

But that won’t be CUDA will it?

[Geonerd]

And AMD is sampling a 7nm card.

But that won’t be CUDA will it?

Does Leela not use OpenCL? Or is CUDA just more efficient for this sort of computing?

[Milos]

And AMD is sampling a 7nm card.

But that won’t be CUDA will it?

Does Leela not use OpenCL? Or is CUDA just more efficient for this sort of computing?

Yes it does. But CUDA version is roughly 10x faster.

[duncan]

But that won’t be CUDA will it?

Does Leela not use OpenCL? Or is CUDA just more efficient for this sort of computing?

Yes it does. But CUDA version is roughly 10x faster.

are you surprised with the win over stockfish ?

[Milos]

Does Leela not use OpenCL? Or is CUDA just more efficient for this sort of computing?

Yes it does. But CUDA version is roughly 10x faster.

are you surprised with the win over stockfish ?

Not at all, when you force SF to start with opening of -1.00 or worse (we don't even know since there is 0 information about actual game beside the moves we don't know what is SF depth, how much time it used, etc, it is so easy to cripple it with some settings) and allow LC0 to freely play the moves, plus huge hardware advantage (more than 10x difference in hardware price) ofc that SF would loose once in a blue moon.
Once LC0 or A0 or any NN program is actually able to beat it in TCEC conditions (while using even dual Titan-V) than we can talk about something significant.

[FWCC]

I would say LCZero eventually will defeat SF with a 1060