Lc0 wins TCEC Cup 11 Final

[Graham Banks]

Use MacBook Pro M1 MAX 32 GPU cores. Or soon 64 GPU cores.

NO.

[Jouni]

This Lc0 is number 3 engine and beaten by Dragon at CCC:
1
Dragon (3832)
157.5/290
2
Lc0 (3822)
132.5/290

[Werewolf]

GPU hardware is improving at a much faster rate than CPU it seems. So the future of Lc0 is looking good, apart from the huge power consumption of the high-end cards.

Use MacBook Pro M1 MAX 32 GPU cores. Or soon 64 GPU cores.

Using the recommended best big net, T78, my MacBook Air M2 gets a whopping 300 nps. My PC with my 3090 card gets 20000 nps. The 4090 would be around 35000 nps.

Apple just sucks for chess.

[Kanizsa]

1 In the CCC league Leela played on a 15*512 network.
In TCUP Leela played on a 15*768 network
2 In the CCC league ratio GPU/CPU is much more biased in favor of CPU.
3 With 30 minutes as in TCUP instead of 5 minutes as in CCC, neural network scales better than A/B programs.
4 Leela won against Stockfish with a newer version, while Leela lost against Komodo running on a older version

[smatovic]

GPU hardware is improving at a much faster rate than CPU it seems. So the future of Lc0 is looking good, apart from the huge power consumption of the high-end cards.

Use MacBook Pro M1 MAX 32 GPU cores. Or soon 64 GPU cores.

Using the recommended best big net, T78, my MacBook Air M2 gets a whopping 300 nps. My PC with my 3090 card gets 20000 nps. The 4090 would be around 35000 nps.

Apple just sucks for chess.

Basically Apple scaled up the A14 Bionic mobile SoC to use in laptops++, now it seems they reached a limit, to scale up for the M-series Mac Pro, the 4-way-slice M2 Extreme chip, IMHO they should have gone the other route, scale the server-class ARM Neoverse down to mobiles... nevertheless, I am not aware of any chess engine that makes use of the dedicated neural engines (TPUs) present in mobile SoCs and M-series.

--
Srdja

[Werewolf]

GPU hardware is improving at a much faster rate than CPU it seems. So the future of Lc0 is looking good, apart from the huge power consumption of the high-end cards.

Use MacBook Pro M1 MAX 32 GPU cores. Or soon 64 GPU cores.

Using the recommended best big net, T78, my MacBook Air M2 gets a whopping 300 nps. My PC with my 3090 card gets 20000 nps. The 4090 would be around 35000 nps.

Apple just sucks for chess.

Basically Apple scaled up the A14 Bionic mobile SoC to use in laptops++, now it seems they reached a limit, to scale up for the M-series Mac Pro, the 4-way-slice M2 Extreme chip, IMHO they should have gone the other route, scale the server-class ARM Neoverse down to mobiles... nevertheless, I am not aware of any chess engine that makes use of the dedicated neural engines (TPUs) present in mobile SoCs and M-series.

--
Srdja

Yes the top-down route seems better to me too. I heard the M2 Extreme (2x M2 Ultras stitched together) was possible, but too expensive. So instead of a 48 CPU core Mac Pro it'll be a more modest 24 core. While I'm sure the CPU part will be faster than the 28 core Cascade Lake (Skylake, cough cough) dinosaur it replaces, it's hardly earth-shattering.

[Werewolf]

I cannot find this version of Leela anywhere ? Is it private?

No. It is located here (build 130) https://ci.appveyor.com/project/Etcaqab/lc0/history

CUDA build https://ci.appveyor.com/api/buildjobs/c ... %2Flc0.exe
cudNN build https://ci.appveyor.com/api/buildjobs/1 ... %2Flc0.exe

BT2-768x15-swa-3250000.pb.gz https://drive.google.com/file/d/1ERjfPE ... qZwTx/view

So BT2 is a smaller net? How come it was used? Was Lc0 playing with or without a GPU?

[Magnum]

GPU hardware is improving at a much faster rate than CPU it seems. So the future of Lc0 is looking good, apart from the huge power consumption of the high-end cards.

Use MacBook Pro M1 MAX 32 GPU cores. Or soon 64 GPU cores.

Using the recommended best big net, T78, my MacBook Air M2 gets a whopping 300 nps. My PC with my 3090 card gets 20000 nps. The 4090 would be around 35000 nps.

Apple just sucks for chess.

It was your mistake to buy an Air.
Which net size?

Are you talking about this: https://github.com/LeelaChessZero/lc0/issues/1562
LC0 is reaching 0.3K now.
5x faster = 1.5K with 8 GPU cores
But the Apple M1X/ will use 32 GPU cores.
1.5K x 4 = 6K !!

Have you seen the other results:
https://github.com/LeelaChessZero/lc0/issues/1795
Big LC0 806286 net
GPU + 2 CPU cores:
Total time (ms) : 350806
Nodes searched : 498436
Nodes/second : 1421
GPU + 1 CPU core:
Total time (ms) : 350752
Nodes searched : 498331
Nodes/second : 1421
That's without the 5x faster improvements.
There was a match some months ago where this net lost only 4 and won 1 game against newest Stockfish. The rest were a draw.
With the new LC0 nets, which are much better than T80 nets, I would expect a tie.
Yes you are right, Apple just sucks for chess

It reaches on my MacBook 1.5K.
If the LC0 developers would start to improve LC0 on Apple devices, then LC0 would reach 7.5K.
With the newest nets, LC0 should win a match against Stockfish.
Don't compare RTX watt consumption to Apple GPU if you have an Nvidia RTX, because it's safer for your life.

[Kanizsa]

You can find most recent developments here

https://ci.appveyor.com/project/Etcaqab/lc0/history

[wickedpotus]

GPU hardware is improving at a much faster rate than CPU it seems. So the future of Lc0 is looking good, apart from the huge power consumption of the high-end cards.

Use MacBook Pro M1 MAX 32 GPU cores. Or soon 64 GPU cores.

Using the recommended best big net, T78, my MacBook Air M2 gets a whopping 300 nps. My PC with my 3090 card gets 20000 nps. The 4090 would be around 35000 nps.

Apple just sucks for chess.

It was your mistake to buy an Air.
Which net size?

Are you talking about this: https://github.com/LeelaChessZero/lc0/issues/1562
LC0 is reaching 0.3K now.
5x faster = 1.5K with 8 GPU cores
But the Apple M1X/ will use 32 GPU cores.
1.5K x 4 = 6K !!

Have you seen the other results:
https://github.com/LeelaChessZero/lc0/issues/1795
Big LC0 806286 net
GPU + 2 CPU cores:
Total time (ms) : 350806
Nodes searched : 498436
Nodes/second : 1421
GPU + 1 CPU core:
Total time (ms) : 350752
Nodes searched : 498331
Nodes/second : 1421
That's without the 5x faster improvements.
There was a match some months ago where this net lost only 4 and won 1 game against newest Stockfish. The rest were a draw.
With the new LC0 nets, which are much better than T80 nets, I would expect a tie.
Yes you are right, Apple just sucks for chess

It reaches on my MacBook 1.5K.
If the LC0 developers would start to improve LC0 on Apple devices, then LC0 would reach 7.5K.
With the newest nets, LC0 should win a match against Stockfish.
Don't compare RTX watt consumption to Apple GPU if you have an Nvidia RTX, because it's safer for your life.

Not sure what points you are trying to make..

You kind of comparing Apples $5000 machine with any run to the mill $900 gaming PC with RTX and Apple looses out badly wit less than half the nps or even less.. How on earth can you even try to make that a "positive" ??? fanboy much???