Eman 9.30 vs Stockfish pre release

[Sedat Canbaz]

Hello sedat, I have used with my architecture two cpu xeon 2697v2 20 cores, 40 threads with 224 gb of ECC ram, the configuration of 256 hash, one core per engine

Hello dear Jordi,

1st of all,
You are as good expert in our chess hobby!
E.g you are one of best in our opening book making and
I'm very glad that we are chess friends for almost 20 years!

That's true...as you stated...
I have some experience in eng testings, but still I learn...
Like you, as you know...this is as never ending process )

And since 2020, how NNUE appeared on scene, e.g in 2023:
Almost all games ending as draw..and in the past we often
Used to play Rapid, Slow TC...but nowadays is meaningless!
Sure in case of running 3700+ and under optimal conditions..

On other hand, and as I stated before:
Bullet or Blitz is not so bad way to test NN 3700+ engines!
E.g in your case (if you run on 2697v2 20 cores machine),
And if the target is to produce many games in short time.. then
You can use and test under Cutechess GUI's current methods:
Bullet 1m+1s: 1 Core, Ponder OFF, 64 mb hash, Concurrent: 16
Blitz 2m+1s: 2 Cores, Ponder OFF, 128 mb hash, Concurrent: 8
Blitz 3m+1s: 4 Cores, Ponder OFF, 512 mb hash, Concurrent: 4

Btw, let's say the goal is to run a serious Blitz MP tour, then
I would prefer running the engines as Ponder ON, in your case:
Blitz 3m+2s: 2 Cores, 512 mb hash, Concurrent: 4
Blitz 5m+2s: 4 Cores, 1024 mb hash, Concurrent: 2

And as you may noticed or not,
Just in case of under the above conditions..your CPUs will be safe...
E.g not so good idea..if running long-tours with 100% CPU usage

Actually if Ponder OFF,
You can run even at 20 Cores (per side) but ıt's not recommended..
Especially in summer time...sure there are some way to reduce temps..
But is it worth?..this is other question of course..

Note also that
The above systems are for Hyper-Threading as Disabled
One thing more,
You may know or not..according to my overall chess benchmarks:
Mostly chess engines run better: when Hyper Threading as Disabled!

By the way, as alternative...
You can try to use my latest Unique Opening Suite, for more details:
https://sites.google.com/site/computers ... test-suite

And if you are interested.. missed... I ran a new Hash test:
viewtopic.php?f=4&t=3600&start=30

Greetings

[botunnet]

You should try Onpaste for your pictures, look :

Games ( 1000)
https://pixeldrain.com/u/JDDuE84u

Regards!

[Desvelemosafrica]

For to make screenshots?

[Desvelemosafrica]

Hello sedat, I have used with my architecture two cpu xeon 2697v2 20 cores, 40 threads with 224 gb of ECC ram, the configuration of 256 hash, one core per engine

Hello dear Jordi,

1st of all,
You are as good expert in our chess hobby!
E.g you are one of best in our opening book making and
I'm very glad that we are chess friends for almost 20 years!

That's true...as you stated...
I have some experience in eng testings, but still I learn...
Like you, as you know...this is as never ending process )

And since 2020, how NNUE appeared on scene, e.g in 2023:
Almost all games ending as draw..and in the past we often
Used to play Rapid, Slow TC...but nowadays is meaningless!
Sure in case of running 3700+ and under optimal conditions..

On other hand, and as I stated before:
Bullet or Blitz is not so bad way to test NN 3700+ engines!
E.g in your case (if you run on 2697v2 20 cores machine),
And if the target is to produce many games in short time.. then
You can use and test under Cutechess GUI's current methods:
Bullet 1m+1s: 1 Core, Ponder OFF, 64 mb hash, Concurrent: 16
Blitz 2m+1s: 2 Cores, Ponder OFF, 128 mb hash, Concurrent: 8
Blitz 3m+1s: 4 Cores, Ponder OFF, 512 mb hash, Concurrent: 4

Btw, let's say the goal is to run a serious Blitz MP tour, then
I would prefer running the engines as Ponder ON, in your case:
Blitz 3m+2s: 2 Cores, 512 mb hash, Concurrent: 4
Blitz 5m+2s: 4 Cores, 1024 mb hash, Concurrent: 2

And as you may noticed or not,
Just in case of under the above conditions..your CPUs will be safe...
E.g not so good idea..if running long-tours with 100% CPU usage

Actually if Ponder OFF,
You can run even at 20 Cores (per side) but ıt's not recommended..
Especially in summer time...sure there are some way to reduce temps..
But is it worth?..this is other question of course..

Note also that
The above systems are for Hyper-Threading as Disabled
One thing more,
You may know or not..according to my overall chess benchmarks:
Mostly chess engines run better: when Hyper Threading as Disabled!

By the way, as alternative...
You can try to use my latest Unique Opening Suite, for more details:
https://sites.google.com/site/computers ... test-suite

And if you are interested.. missed... I ran a new Hash test:
viewtopic.php?f=4&t=3600&start=30

Greetings

I had read the hypertreading issue regarding the Komodo 3.0 versions; 3.1 said by the author of the engine on his website. I have disabled the turbo option (it is something similar to overclocking, but the bios itself does it) on the recommendation of the author of Catropolis, and it makes sense. More voltage, more temperature, therefore equal to possible hardware damage or errors. In fact, I myself have noticed that I have no reboot problems since I have not used this option. I take note of the rest, because as I am going to use cutechess and if hyperthreading is not recommended for komodo, because it will not be applied to the rest of the engines. Thanks for the suite, because I hadn't seen it and with it we made sure that both engines repeat lines on both sides, right?

Regards, my friend.

[Amos 4ever]

Hey DeeDs, Amos 4ever = MZ

Yes i knew it, he used the same at Chess2U. He no longer has anyone to talk to at OCF, no engine trainers, no learning results, no books tournaments, only ajax errors loool

He said the one who speaks alone with multi accounts.

[dorsz]

You should try Onpaste for your pictures, look :

Games ( 1000)
https://pixeldrain.com/u/JDDuE84u

Regards!

I did not read the whole thread but did anyone notice that in this match there are 19 loses on time?

[deeds]

I did not read the whole thread but did anyone notice that in this match there are 19 loses on time?

Before he published his results, people already don't care about them because as he followed Sedat conditions/settings, instantly we expected "time forfeit + lost experience data + shallow depth"...

[dorsz]

Lots of the games with time loses has drawish eval.

[deeds]

Maybe it was his goal, hard to say with Sedat's fanatics.
It's weird that the same dude who came to explain people how to run Eman in cluster, install mpi, etc, was the same dude who did this test in this way loool
He choose his side without engine trainers, good luck to use learning engines !

[Desvelemosafrica]

Maybe it was his goal, hard to say with Sedat's fanatics.
It's weird that the same dude who came to explain people how to run Eman in cluster, install mpi, etc, was the same dude who did this in this way loool
He choose his side, good luck !

I think you are getting confused. The person who explained to me how to run an MPI engine is Kevin, not Sedat. I don't know where you have read that Sedat has helped me with this. And how can it be otherwise, it was the author of Eman who helped me understand how his cluster works and what changes to make with each update. And sedat is a friend. The fanatic is your perception, but there is no fanaticism.