M1 Apple Silicon for Chess?

[AlexChess]

@Acepoint.de

I'm really worried about the soldered SSD stressed by 6-man tablebases and on countinous usage: Checked now: I'm a 86% after less than 3 monshs.
I hope that external Thunderbolt 3 SSD will be bootable by Mac mini M1, I suspend my chess engines tornament for a while...

https://9to5mac.com/2021/02/23/m1-mac-u ... -lifespan/

[Ras]

I'm really worried about the soldered SSD stressed by 6-man tablebases and on countinous usage: Checked now: I'm a 86% after less than 3 monshs.

Tablebases should be read-only usage. I don't see why this would cause any SSD wear at all - unless your filesystem is misconfigured. It should be mounted with noatime or something equivalent to avoid storing the last read access time.

[AlexChess]

I'm really worried about the soldered SSD stressed by 6-man tablebases and on countinous usage: Checked now: I'm a 86% after less than 3 months.

Tablebases should be read-only usage. I don't see why this would cause any SSD wear at all - unless your filesystem is misconfigured. It should be mounted with noatime or something equivalent to avoid storing the last read access time.

Thank you for the explanation!
But also the system until now has been on day and night to play my Arena engine tournament... Do you think that Mac M1 will boot from an external Thunderbold 3 SSD? (I have always user USB 3.0 SSD until now, but there are only 2 ports on new M1 computers

[towforce]

@Acepoint.de

I'm really worried about the soldered SSD stressed by 6-man tablebases and on countinous usage: Checked now: I'm a 86% after less than 3 monshs.
I hope that external Thunderbolt 3 SSD will be bootable by Mac mini M1, I suspend my chess engines tornament for a while...

https://9to5mac.com/2021/02/23/m1-mac-u ... -lifespan/

Speculation on my part, but unless they explicitly told you when you bought it that it should not be used for applications that use the SSD a lot, then under British law it might not be "fit for purpose".

[Ckappe]

I haven't read the whole thread - is there a conclusion? Does a light laptop with an Apple M1 SOC get roughly the same amount of analysis on one charge as a laptop with a heavy battery and big power-hungry CPU? Or is it just too difficult to compare?

Edit: the post that appeared while I was writing this (see above) seems to indicate that in this particular game, a power-hungry AMD CPU with a heavy battery would be the winner.

Conclusion is that the higher-end laptop with bigger battery wins hands down. running the same SF13 code it does approx 2x nodes analyzed before battery runs out..and due to the bigger battery stamina is only about 30% less at full throttle, despite much faster Ryzen 9 CPU )

And due to M1s relatively weak GPU analyzing with GPU engines like LC0, Ceres, Allie etc. the difference becomes abysmal for those use-cases..

Thanks. This makes sense: the M1 SOC is, IMO, an absolutely STUNNING piece of technology, even by today's standards, and Apple deserve kudos for creating it. However, it's not really surprising that a low powered lightweight laptop created from it is not optimised for this particular comparison, and it's also not surprising that it can't win it.

Apologies if this is actually already happening, but I hope the manufacturers of computers that use other operating systems (especially Windows and Chrome OS) take note and start using SOCs in place of CPUs (though obviously they can't buy an SOC as good as the Apple M1 right now).

I have nothing against M1 as such if it was open and not tied to Apple's locked proprietary DRM-control, locked bios, AIO "appliances". it would be a more valid option. This said the 39W TDP and performance/power ratios do not appear to be significantly ahead of other more low-powered mobile 5nm ARMs like Snapdragon, Kirin, Exynos, etc. in the perf/power ratio aspect (not even that much ahead of AMD's 7nm CPU it seems in real-life testing), hence their massive marketing bandwagon pushes out endless of comparisons against only older-gen Intel x86 CPUs, online.

The main problem with M1 today is the performance/price ratio, external I/O, and the lack of GPU/AI processing matching the price-tag if I see this from a chess-perspective.

I also think more laptop/phone/tablet makers of both ARM CPUs and x86-CPUs will follow suit and include on-chip RAM. I would also assume Apple as well as others will need much more than the limiting 16GB on-chip as long as external/extended RAM is not accessible from CPU.

I am not that fond of combining too much on the SOCs, I think RAM and GPU make sense but not really at the expense of the ability to access RAM and GPU externally as well. I am even more against soldering wear-and-tear parts like SSDs etc to motherboards in laptops. In essence, this leads to throw-away products with an OS-controlled planned obsolesce scheme that I am not too fond of. And when people report that the OS overuses the soldered SSDs the lifespan of these "netbooks" looks questionable, indeed.

[AlexChess]

Ok,
I'll constanly monitor my SSD, stopping my Arena Tournament for 1 month and then watching effects.
Luckily I have verified that Big Sur M1 can be booted also from USB 3 or Thunderbolt 3 external SSD, but Apple crazy policy to solder vital components deserves a class action like the batterygate

[Ras]

But also the system until now has been on day and night to play my Arena engine tournament...

Good consumer SSDs with 500 GB allow about 300 TBW (terabyte written). If you have used up 14% of that, that's 42 TB. With 90 days of usage, that's more than 400 GB written per day, which is huge. There's no way that this could be due to Arena. The general opinion on the issue seems to point to the OS itself wildly swapping RAM to disk, which would hit particularly the M1 models with only 8 GB RAM.

Do you think that Mac M1 will boot from an external Thunderbold 3 SSD?

I have no idea. Then again, I wouldn't recommend a laptop that has the flash storage soldered instead of using a proper M.2 socket. The main reason why Apple doesn't use M.2 is so that they can shut out market competition and demand insane prices. For changing from 512 GB to 2 TB flash, Apple charges 690 EUR. A Samsung Evo 970 Evo Plus with 2 TB (M.2 NVMe fast SSD) is technically on a par, but costs only 290 EUR.

[Ckappe]

@Acepoint.de

I'm really worried about the soldered SSD stressed by 6-man tablebases and on countinous usage: Checked now: I'm a 86% after less than 3 monshs.
I hope that external Thunderbolt 3 SSD will be bootable by Mac mini M1, I suspend my chess engines tornament for a while...

https://9to5mac.com/2021/02/23/m1-mac-u ... -lifespan/

86% after less than 3 monshs sounds really bad, but I guess Apple could fix the overuse by issuing updates to MacOs..Do you have the 8GB M1 ?

[AlexChess]

Arena was running on Windows 10 ARM64 under Parallels M1 on mac internal HD, but the virtual image is on an external USB 3.0 512 GB SSD (together with Sygyzy 6-man Tablebases). The only way for me to know if is it the cause, is to stop continuous usage of both. I really hope that is an Apple software bug, solved with next builds. It's an M1 8GB 256GB SSD.

[Ckappe]

Arena was running on Windows 10 ARM64 under Parallels M1, but the virtual image is on an external USB 3.0 512 GB SSD (together with Sygyzy 6-man Tablebases). The only way for me to know if is it the cause, is to stop continuous usage of both. I really hope that is an Apple software bug, solved with next builds.

I've seen VMware, Virtualbox and other virtual machines do lots of extra disk-writes to manage machine-states etc..And of course requires lots of Mem,,, Maybe this is a parrallels issue as well?