What computers to buy for Chess? [Solved]

[keos]

Hello,

I assume this is a problem that can only be solved with the appropriate hardware and has nothing to do with the software category in computers. Neither the Intel Core i5-8265U Inspiron 3583 nor the other Yoga 11e will work for this chess issue.

In another forum I was told, without further comment something that I do not understand well but that seems to indicate that the issue has no solution with these PCs:

"Without GPU, Lc0 is very slow as CPU is not very well suited to evaluate neural networks.

On CPU only, Lc0 is indeed weaker than other top chess engines, even though it's still much stronger than any of human players."

--Lc0 is a chess program.

I need please that some of you professionals here recommend me which model or type of computer I should buy so that the chess programs can work correctly.

If you could indicate me some specific ones selling by Amazon it would be the most convenient for my case, thanks for any information in this regard.

Greetings

[j2mcgreg]

Lc0 is not available in a prebuilt package for MX or any other distro, but there are instructions available to build it from source code. See here:

https://lczero.org/play/download/

As this is an MX Linux forum, I think that your question is really out of our purview. It’s the sort of question that you should be putting to Tom's Hardware, or Linus Tech Tips et al.

[keos]

The lc0 is already installed and it runn well in mxlinux, the problem is not with mxlinux, i think, but with my PC since the program play weak against other programs (for the reason he explain ...)

-- there should be a PC more convinient for chess programs.

[MadMax]

Well, you kinda gave the answer yourself already.

You will need a PC with a dedicated GPU. I'm not familiar with chess programs, but when it comes to neural networks you'll probably want something as beefy as your budget allows. Does Lc0 have some sort of benchmarking function?

A quick google search recommends a 3080 which is in the area I'd have expected (around 2080/3070+): https://www.google.com/search?q=lc0+recommended+gpu

[j2mcgreg]

The lc0 is already installed and it runn well in mxlinux, the problem is not with mxlinux, i think, but with my PC since the program play weak against other programs (for the reason he explain ...)

-- there should be a PC more convinient for chess programs.

Where did you get it?

[keos]

@ MadMax,

All what he said was:

"Without GPU, Lc0 is very slow as the CPU is not very suitable for evaluating neural networks."

No more details, and I don't know about computers, only about chess. I only make deductions, from what he said it seems that "it is not very suitable for evaluating neural networks" so it seems that this PC I have is not suitable for the most modern chess programs of today that come with a 'nnue' (something related/belongs to the field of computer science, nothing to do with chess).

I will see what to do with the information that you give in the link, I must be sure, I have to continue investigating more in the matter. Thank you.

@ j2mcgreg,

From github the program, https://github.com/LeelaChessZero/lc0/b ... E.md#linux
and the 'nnue' from: https://lczero.org/play/networks/bestnets/

Note:

Computer chess has taken a huge leap forward, now it seems that more efficient machines are needed if you want to get the most out of the programs; and if we are not an expert in computers ...

[MadMax]

Well, ok, if you come from a non-computer-savvy standpoint, I can clarify this a little.

Every computer has a CPU (central processor unit) which is the "multi-purpose" chip of the system. It does most of the calculations needed to run the system and is incredibly versatile, but mainly does sequential calculations.

Then there are GPUs (graphics processor units) which are chips that are specialized for parallel workloads, which historically were graphics calculations, hence their usage in graphic cards. In recent years a lot more types of calculations were optimized for GPUs, like rendering, en/decoding, mining (which makes heavy use of hash algorithms), AI, neural networks, you name it. So it's not a surprise to me that a chess engine works a lot faster on a GPU.

In typical circumstances we are talking about a mid-level GPU being 100x faster than the fastest CPU on the market, so the difference can be huge.

[j2mcgreg]

So you built it from source code, good on you. Taking into account what MadMax says (above), I think that your best bet would be to assemble it yourself. I say this because, in order to get the gpu you need, you would have to buy a gaming rig. Most prebuilt gaming rigs use Nvidia graphics and given the problems the newest Nvidia cards can have with MX, I think that you are going to have a better end result if you use a high-end AMD card instead.

[keos]

I think it is now quite clear to me, especially with what you say MadMax, that the matter is beyond my possibilities ...

-- I have heard that the great masters nowadays have computer consulting by specialists, now I understand the reason, the programs demand more and more every day and only a specialized technician can manage to put together such sophisticated gadgets ... and in the near future a kind of Cyborg-Frankenstein will play chess.

Well, it's not so depressing, at least I understand now what I didn't understand before.

Thanks to both of you, best regards!

[keos]

@ MadMax,

Back to the point.

What he said:

"Without GPU, Lc0 is very slow as the CPU is not very suitable for evaluating neural networks."

I found this in wikipedia:

"NNUE runs efficiently on central processing units without a requirement for a graphics processing unit (GPU)."

https://en.wikipedia.org/wiki/Efficient ... al_network

I don't think I understand anything at all now.