M1 Apple Silicon for Chess?

[AlexChess]

Tomorrow at WDCC21 Apple will show new Silicon Mx devices 2-3 times faster, but still slims, silents and freshes. AMD like INTEL and Microsoft with Windows 11 are moving to ARM, that is the future

So after yesterday's lack of announcement, it seems Apple silicon may be stuck with the 4-core (4 fully powered ones) M1 they have today. And will not get any closer to the already available Ryzens at least before autumn.

I have the Macbook Pro M1 and it’s great.
I will buy the 16 inch too.
But I’m waiting for a MacBook Pro 16 inch (or better 18 inch) and 16 CPU performance cores or more to use Stockfish.

To use LC0 I would need a MacBook Pro with 1032 or 2064 GPU cores.

See discussions here
https://github.com/official-stockfish/S ... ssues/3529
https://forums.macrumors.com/threads/ap ... 62/page-12
= It’s a problem of Apple M1 hardware, no it’s a problem of Stockfish code, no it’s…

I'm also very happy of my Mac mini M1. I have solved all issues related to the first Big Sur's releases and now it is stable and performant. I like very much that with it I can use not only iPad and iPhone apps, but also Windows 10 and Linux | Raspberry Pi4 ARM64 applications! No hurry to update it with M1x, M2 news Silicon socks. I'll buy another Mac mini | Macbook when M3-M4 will be out. If Apple has decided to postpone new computers, could be dued to hardware shortage but also simply because it is selling well M1 devices...

[wickedpotus]

So after yesterday's lack of announcement, it seems Apple silicon may be stuck with the 4-core (4 fully powered ones) M1 they have today. And will not get any closer to the already available Ryzens at least before autumn.

I'm also very happy of my Mac mini M1. I have solved all issues related to the first Big Sur's releases and now it is stable and performant. I like very much that with it I can use not only iPad and iPhone apps, but also Windows 10 and Linux | Raspberry Pi4 ARM64 applications! No hurry to update it with M1x, M2 news Silicon socks. I'll buy another Mac mini | Macbook when M3-M4 will be out. If Apple has decided to postpone new computers, could be dued to hardware shortage but also simply because it is selling well M1 devices...

I am sure there are plenty of happy Mac users out there for various reasons. Many of them being very vocal about this as well . But let's keep the discussion to the subject of M1 Apple Silicion's usefulness (or lack thereof - due to its poor performance compared to current alternative 7nm computers) for Chess.

[Ras]

If Apple has decided to postpone new computers, could be dued to hardware shortage but also simply because it is selling well M1 devices...

Or because it would have been pretty unusual for them to present such an upgrade within less than a year. Whereever that rumor even came from, it was clearly not based on Apple's past actions.

[AlexChess]

New open source updates availables:

Fire 8.NN.MC.3
Berserk 4.5.1
Clover 2.4
Seer 2.20
Cfish 290721
Corchess 170621
Sugar AI 2.30
Minic 2.10
Koivisto 6.0

Best regards, Alex

[AlexChess]

New native M! engine: Zahak 6.2 https://github.com/amanjpro/zahak/releases/tag/6.2
Multi CPU support (up to 8) approximate 3000 ELO

Thanks Amanj

[amanjpro]

New native M! engine: Zahak 6.2 https://github.com/amanjpro/zahak/releases/tag/6.2
Multi CPU support (up to 8) approximate 3000 ELO

Thanks Amanj

Thanks for sharing, I believe it is more like 2800 elo as per CCRL, and it can actually support as many threads as there are CPUs on the device

[AlexChess]

New native M! engine: Zahak 6.2 https://github.com/amanjpro/zahak/releases/tag/6.2
Multi CPU support (up to 8) approximate 3000 ELO

Thanks Amanj

Thanks for sharing, I believe it is more like 2800 elo as per CCRL, and it can actually support as many threads as there are CPUs on the device

I'm testing it with 4 CPUs, others are using only 1 CPU so I think that is 100 ELO stronger on my hardware. The win build it's a quantum leap from 5.0 . It has drawn also with Marvin 5.1.0 (3093 ELO) and Rofschade 2.3 (3293 ELO) . If you add NNUE and tablebases in 7.0 you could make another jump of 100 ELO. And I haven't still tested it on Mac M1 native where it's 2-3 times faster. I'm grateful to you for supporting latest Macs, too.
http://www.computerchess.org.uk/ccrl/40 ... +opponents

[Ras]

How do you prevent engines getting moved off to the efficiency cores? Reading this here:
https://arstechnica.com/civis/viewtopic ... #p39560471
suggests that there is no good way to keep a long-running computational task on the performance cores, in particular when minimised. It seems that using QOS_CLASS_USER_INITIATED could work at least in engine tournaments with moderate thinking times because the engine would have made its move before getting shoved off to a small core.

However, the next problem would be engines without special Apple support because I see the risk that these would be moved very quickly to efficiency cores if the tournament manager is minimised, completely invalidating the match results. This here:
https://eclecticlight.co/2021/05/14/cor ... do-better/
suggests a performance factor of 7 between the big and the little cores. That would be between 150 and 200 Elo just because of where the engine is running.

[AlexChess]

How do you prevent engines getting moved off to the efficiency cores? Reading this here:
https://arstechnica.com/civis/viewtopic ... #p39560471
suggests that there is no good way to keep a long-running computational task on the performance cores, in particular when minimised. It seems that using QOS_CLASS_USER_INITIATED could work at least in engine tournaments with moderate thinking times because the engine would have made its move before getting shoved off to a small core.

However, the next problem would be engines without special Apple support because I see the risk that these would be moved very quickly to efficiency cores if the tournament manager is minimised, completely invalidating the match results. This here:
https://eclecticlight.co/2021/05/14/cor ... do-better/
suggests a performance factor of 7 between the big and the little cores. That would be between 150 and 200 Elo just because of where the engine is running.

It's interesting, thank you!
I would really like to learn how to compile open source github engines by myself on M1, but also on Linux and Windows ARM64. If you know tutorials that esplain the procedures step by step (and all tools needed) and share them I would be very grateful. Until now I have managed only to compile Stockfish and LC0 with homebrew and Fat Fritz 2.0 github for Ubuntu ARM64 and Mac M1.

Best regards, Alex

[Ras]

If you know tutorials that esplain the procedures step by step (and all tools needed) and share them I would be very grateful.

That depends on the engine, programming language, and operating system. You can have makefiles, build scripts, or even MS Visual Studio projects (mostly for Windows-only engines). Sometimes, the Windows version is not even built under Windows, but cross-compiled e.g. from Linux. The specific instructions and requirements should be in the engine documentation.