Leela vs Stockfish nnue style

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

Moderator: Ras

Sopel
Posts: 391
Joined: Tue Oct 08, 2019 11:39 pm
Full name: Tomasz Sobczyk

Re: Leela vs Stockfish nnue style

Post by Sopel »

I hear that M1 and M2 chips have only enough TLB for 32MiBs of memory, so that tracks. 50% higher for M3, so still pretty bad. Still interesting that it regresses so much compared to Intel/AMD though, because it's still pretty comparable without large pages.
dangi12012 wrote:No one wants to touch anything you have posted. That proves you now have negative reputations since everyone knows already you are a forum troll.

Maybe you copied your stockfish commits from someone else too?
I will look into that.
h1a8
Posts: 518
Joined: Fri Jun 04, 2010 7:23 am

Re: Leela vs Stockfish nnue style

Post by h1a8 »

Sopel wrote: Fri Dec 29, 2023 1:04 am I hear that M1 and M2 chips have only enough TLB for 32MiBs of memory, so that tracks. 50% higher for M3, so still pretty bad. Still interesting that it regresses so much compared to Intel/AMD though, because it's still pretty comparable without large pages.
What is TLB?
Magnum
Posts: 195
Joined: Thu Feb 04, 2021 10:24 pm
Full name: Arnold Magnum

Re: Leela vs Stockfish nnue style

Post by Magnum »

h1a8 wrote: Tue Dec 26, 2023 5:37 pm
Werewolf wrote: Thu Nov 30, 2023 4:08 pm Children, children.

I've benched the M2 Ultra (should be at least as fast as the M3 Max) and it got 17Mn/s

Not great performance.

Excellent performance to power ratio though, but I do think this argument is overplayed.
Yup it's a bug with apple's ram hash function. Using the command
"bench 512 8 2000 default movetime"

M2 Pro, 12‑core CPU, 32GB memory
gets the following result

Total time (ms) : 90548
Nodes searched : 2218332923
Nodes/second : 24498972

This is on par with some i9-13900 machines (from Ipman Chess benchmark listings)
So the M2 Ultra should bench even higher than that result
Much better: Buy 16-inch M3 MAX MacBook Pro.
Sopel
Posts: 391
Joined: Tue Oct 08, 2019 11:39 pm
Full name: Tomasz Sobczyk

Re: Leela vs Stockfish nnue style

Post by Sopel »

h1a8 wrote: Fri Dec 29, 2023 11:17 am
Sopel wrote: Fri Dec 29, 2023 1:04 am I hear that M1 and M2 chips have only enough TLB for 32MiBs of memory, so that tracks. 50% higher for M3, so still pretty bad. Still interesting that it regresses so much compared to Intel/AMD though, because it's still pretty comparable without large pages.
What is TLB?
Cache for virtual to physical address map
dangi12012 wrote:No one wants to touch anything you have posted. That proves you now have negative reputations since everyone knows already you are a forum troll.

Maybe you copied your stockfish commits from someone else too?
I will look into that.
wickedpotus
Posts: 153
Joined: Sun May 16, 2021 5:33 pm
Full name: Aron Rodgriges

Re: Leela vs Stockfish nnue style

Post by wickedpotus »

Magnum wrote: Wed Jan 03, 2024 12:02 am
h1a8 wrote: Tue Dec 26, 2023 5:37 pm Yup it's a bug with apple's ram hash function. Using the command
"bench 512 8 2000 default movetime"

M2 Pro, 12‑core CPU, 32GB memory
gets the following result

Total time (ms) : 90548
Nodes searched : 2218332923
Nodes/second : 24498972

This is on par with some i9-13900 machines (from Ipman Chess benchmark listings)
So the M2 Ultra should bench even higher than that result
Much better: Buy 16-inch M3 MAX MacBook Pro.
It appears that you may have forgotten to enable NNUE in the bench command. To enable NNUE, you must add ‘nnue’ to the bench command as well. Unfortunately, the M2 Pro is not capable of achieving 24Mnps.

If you wish to run the same test with 512MB cache, you should most likely use the following command:

bench 512 8 26 default depth nnue

However, you will likely get slightly better results by using all 12 cores. In that case, you can use the following command:

bench 512 12 26 default depth nnue

Regarding your question, it is not appropriate to generalize a group of people based on their choice of technology. It is important to respect everyone’s preferences and choices
h1a8
Posts: 518
Joined: Fri Jun 04, 2010 7:23 am

Re: Leela vs Stockfish nnue style

Post by h1a8 »

wickedpotus wrote: Thu Jan 11, 2024 5:30 pm
Regarding your question, it is not appropriate to generalize a group of people based on their choice of technology. It is important to respect everyone’s preferences and choices
Generalize a group of people? I don't understand. The question refers to the playing style of two engines (not people).
Hai
Posts: 693
Joined: Sun Aug 04, 2013 1:19 pm

Re: Leela vs Stockfish nnue style

Post by Hai »

h1a8 wrote: Fri Mar 22, 2024 10:57 pm
wickedpotus wrote: Thu Jan 11, 2024 5:30 pm
Regarding your question, it is not appropriate to generalize a group of people based on their choice of technology. It is important to respect everyone’s preferences and choices
Generalize a group of people? I don't understand. The question refers to the playing style of two engines (not people).
LC0 BT4 works fine on MacBook Pro 16-inch M1 MAX or higher.
wickedpotus
Posts: 153
Joined: Sun May 16, 2021 5:33 pm
Full name: Aron Rodgriges

Re: Leela vs Stockfish nnue style

Post by wickedpotus »

Hai wrote: Sat Mar 23, 2024 2:17 pm
h1a8 wrote: Fri Mar 22, 2024 10:57 pm
wickedpotus wrote: Thu Jan 11, 2024 5:30 pm
Regarding your question, it is not appropriate to generalize a group of people based on their choice of technology. It is important to respect everyone’s preferences and choices
Generalize a group of people? I don't understand. The question refers to the playing style of two engines (not people).
LC0 BT4 works fine on MacBook Pro 16-inch M1 MAX or higher.
Define "works fine" compared to a similar priced gaming laptop with Nvidia 2024, running BT4... ... Overpriced & underperforming I would say in so many computing areas.