Snapdragon X Elite & Stockfish 16.1 benchmark?

[Lion]

Hello all,

I am interested into a new laptop and was wondering what performance a Snapdragon X Elite would be for chess softwares?

Is there any available Snapdragon X Elite & Stockfish 16.1 benchmark?

Thanks & regards

[Magnum]

Hello all,

I am interested into a new laptop and was wondering what performance a Snapdragon X Elite would be for chess softwares?

Is there any available Snapdragon X Elite & Stockfish 16.1 benchmark?

Thanks & regards

It's much much better to buy a MacBook Pro 16-inch M3 MAX or M4 MAX.

[Lion]

Thanks for your feedback.

There is no M4 Max available right?
Also why do you say its better to go with MacBook Pro M3 MAX, is there any evidence that it is faster using stockfish?

I would also assume that a MacBook Pro M3 is not the same budget than a laptop using Snapdragon X Elite?!

rgds

[Steve Maughan]

Hello all,

I am interested into a new laptop and was wondering what performance a Snapdragon X Elite would be for chess softwares?

Is there any available Snapdragon X Elite & Stockfish 16.1 benchmark?

Thanks & regards

It's much much better to buy a MacBook Pro 16-inch M3 MAX or M4 MAX.

Any link to back this up? From what I've heard Apple's chips haven't been that good for chess and the new Snapdragon chips seem to be better than Intel's. I really doubt the M3 or M4 are "much much" better but I guess we'll find out soon.

Steve

[Viz]

Hello all,

I am interested into a new laptop and was wondering what performance a Snapdragon X Elite would be for chess softwares?

Is there any available Snapdragon X Elite & Stockfish 16.1 benchmark?

Thanks & regards

It's much much better to buy a MacBook Pro 16-inch M3 MAX or M4 MAX.

Again spreading fake news based on benchmarks that don't include a single chess engine.

[smatovic]

Note that it is ARM v8 based and you will need to run x86-64 chess engines via MS Prism:

https://www.anandtech.com/show/21445/qu ... eep-dive/2

In terms of compatibility, the biggest roadblock here is expected to be AVX2 support. Compared to the NEON units on Oryon, the x86 vector instruction set is both wider (256b versus 128b) and the instructions themselves don’t perfectly overlap. As Qualcomm puts it, AVX to NEON translation is a difficult task. Still, we know it can be done – Apple quietly added AVX2 support to their Game Porting Toolkit 2 this week – so it will be interesting to see what happens here in future generations of Oryon CPU cores. Unlike Apple’s ecosystem, x86 isn’t going away in the Windows ecosystem, so the need to translate AVX2 (and eventually AVX-512 and AVX10!) will never go away either.

Maybe team SF will make a special MS ARM v8 NEON compile in future, Idk.

--
Srdja

[adnoh]

I have some numbers I can share.

Snapdragon X Elite X1E78100 3.42 GHz 32.0 GB RAM (ASUS Vivobook S 15) running the Stockfish 16.1 popcnt build from stockfish.org gets the following on the bench command. The Snapdragon cannot run the AVX2 or BMI2 build. It just exits with no message whatsoever.

popcnt
===========================
Total time (ms) : 1696
Nodes searched : 1303971
Nodes/second : 768850

On my Intel i7-12700K 3.60 GHz (Turbo Boost to 4.6GHz) 32GB I get the following
popcnt
===========================
Total time (ms) : 1221
Nodes searched : 1303971
Nodes/second : 1067953

For comparison, on my machine running the BMI2 build I get this.
===========================
Total time (ms) : 1017
Nodes searched : 1303971
Nodes/second : 1282174

So depending on how you want to look at it you are losing 30-40%. Would a proper ARM build be competitive if you lose the advantages of the AVX2 or BMI2 instructions? I do not know.

ps. The snapdragon machine is not mine but I did get my nephew to run the benchmark for me.

[Lion]

That is strangely low as with and AMD 8840HS I hit around 9000 kn/s

[smatovic]

That is strangely low as with and AMD 8840HS I hit around 9000 kn/s

Not strange, consider x86-64 SSE to ARM NEON translation via MS Prism, consider that AVX2 is 256b wide and NEON 128b wide.

A native Stockfish MS ARM NEON build will be for sure faster than one running via Prism.

--
Srdja

[Lion]

Thank you.

Isn’t it still a lot to be around 10x slower?

Rgds