High end PC can't beat lousy Galaxy A7 (2017)

[Guenther]

Heck.. I am not even sure the Smartphone is even 64bit LOL!

The A7 uses the Exynos 7880 Octa SOC (link) which has a 64 bit bus.

Even if this PC/phone comparison is not exactly fair, it seems to raise the serious issue that chess is beginning to close in on a draw even on low-cost low-powered hardware now.

You also did not read the link given... you cannot expect real games with manually transferring moves by Humans especially with ponder on.
This completely wrecks any given (pseudo) tc.

I read the link, but I don't have an AMD CPU and It's not what this thread is about.
Please explain what is wrong with manually playing the moves myself. If a certain machine is significantly stronger it should win regardless.

I cannot hold your hand further, if you cannot read and understand above...I expected some basic understanding which is missing,
thus I close this thread for me. Strong hardware in the hands of certain users just means nothing, it cannot magically overcome
all user errors and believe me, over 90% of similar problems boil down to user errors.

How should we e.g. know, if you just play smartphone 8cores vs. pc 1core? and so on...there are bazillion of things users can wreck.
Also we had users saying similar things like - 'why are all games draw?' - x posts later it was revealed

Code: Select all

user:all = 2 games!

ahem

I directly linked to those quotes below, which have zero to do with 'AMD', but with general setup problems.
(and BTW data is game+comments = pgn with eval/depth/time, but with manual making moves in two different GUIs you cannot deliver useful data anyway)

You cannot be serious, you are using ponder, so while you are manually entering the moves from one platform to the other engines are actually thinking??? That can skew things a lot coz you are certainly taking more time to enter the moves than what is your time control.[/code]

OP dont even know how to set up engine tournment properly.

OP should do the first step properly then discussion about technical details.

[MikeGL]

Hi,
My PC specs:
CPU: intel i9 10900K (10 cores @4.9ghz)
RAM: Gskill TridentZ 32gb (2x16gb) 3600mhz cl16
SSD: 1TB Samsung 970 EVO plus (nvme)
Motherboard: Gigabyte Z490 Aorus Master

I do 10 minutes blitz games between the PC and the A7 (Droidfish - Stockfish + NNUE).
on the PC I use fully updated Fritz17 GUI running differnt engines: Stockfish 12 (BMI2/ AVX2 + NNUE), Stockfish 10 (no NNUE), Komodo 14, Dragon (NNUE + AVX2)...I give them at least 8gb of Hash. I use Fritz17 built in opening book. I also tried using syzygy but it doesn't change much.

Just to put things into perspective, The Galaxy A7 has only 8 (small) cores running at 1.9ghz max (I'm not even sure if it even reaches this speed the entire time), and I don't think the IPC even comes close to intel desktop CPU. Also the RAM is probably much slower on the smartphone and less quantity of it. I know the Droidfish app uses only 64mb of hash, and has very limited opening book (maybe 3 or 4 moves before starts to think).
Heck.. I am not even sure the Smartphone is even 64bit LOL!

The PC version is not only 64bit but uses BMI2 and AVX2 versions for all the engines!

So with all those hardware benefits, it still can't beat the little A7 Smartphone. It always ends up in a Draw! not even a hint of an edge (well..maybe sometimes, but it disappears very fast).

What is the explanation? Doesn't the hardware worth anything?

according to this thread,
http://talkchess.com/forum3/viewtopic.p ... &start=220


Even its lowly Samsung Galaxy A5 brother is no pushover, clocking @ 25% or 1/4th of the fastest on the table.

Code: Select all

Samsung Galaxy A5 2016 (Exynos 7580 8 cores 1.6 Ghz)    1.610.000

So your discovery is not so surprising, I guess.

EDIT:
sorry, the ratio is wrong because I saw new updates on the thread with i7 on the table.

Code: Select all

PC with i7-6700k				       12.120.000

So it is not actually 25%, but actually it is lower.

[CRoberson]

There are lots of ways for a user to produce erroneous test results.
The first thing I see is your use of the stock Fritz opening book. There are lots of tuning parameters for the book use and the defaults have historically been bad for testing engine vs engine.