ProDeo
Would you like to react to this message? Create an account in a few clicks or log in to continue.
ProDeo

Computer Chess
 
HomeHome  CalendarCalendar  Latest imagesLatest images  FAQFAQ  SearchSearch  MemberlistMemberlist  UsergroupsUsergroups  RegisterRegister  Log in  

 

 Problem with Rebel 16.2 on AMD threadripper

Go down 
+2
Chris Whittington
Dann Corbit
6 posters
Go to page : 1, 2  Next
AuthorMessage
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 7:15 pm

There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think). My machine has 32 cores, but 64 threads. It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads). However, Rebel seems to be detecting the core count and limiting to 32 threads. Because of this, I cannot use more than half of my CPU power.
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 7:18 pm

Problem with Rebel 16.2 on AMD threadripper Wrong11
Problem with Rebel 16.2 on AMD threadripper Arena11
Back to top Go down
Chris Whittington




Posts : 1254
Join date : 2020-11-17
Location : France

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 7:45 pm

Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

This seems to be an Arena problem (see Ed's post), it compares the engine name with any relics it has in history and then assumes the relic's "saved" UI status. Not much we can do about that.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

Yes, I detected the cores and forbid threads > cores. The rationale is two-fold, first, there's no gain in nps/Elo and second, Windows can get unstable if the thread count gets thrashed too much.

If you can prove an Elo gain, maybe I'll change my mind on this, but I don't like the idea of a gazillion crash bug complaints all over the internet as a result, so even then the best you can hope for is that Ed holds a special protected version for power users who are aware of the potential problems.
Back to top Go down
Dio




Posts : 214
Join date : 2021-08-28

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:06 pm

The problem with Arena is the file "engine.ini". This file can be deleted without any problems. In this file some engines are assigned wrong authors or the engine is declared as WB engine.
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:17 pm

Dio wrote:
The problem with Arena is the file "engine.ini". This file can be deleted without any problems. In this file some engines are assigned wrong authors or the engine is declared as WB engine.
I get half the NPS and 50% CPU utilization.
For me, this makes Prodeo a non-starter.

In short, the algorithm does not work for Threadripper.
How about a radio button to ignore the limitation?

Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:26 pm

Problem with Rebel 16.2 on AMD threadripper Sf-90p10
Problem with Rebel 16.2 on AMD threadripper 70mnps10
Problem with Rebel 16.2 on AMD threadripper 22mnps10
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:27 pm

Problem with Rebel 16.2 on AMD threadripper Utiliz10
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:29 pm

In the images, we can see that Rebel is using half of my compute power.
Stockfish, with 59 out of 64 threads requested, is using 89% of my compute power.
I can set it up to 63 threads and the machine will still be responsive if I don't have other high energy tasks running.
We see that SF is getting 70M Nps and Prodeo is getting 22M Nps.
Back to top Go down
Chris Whittington




Posts : 1254
Join date : 2020-11-17
Location : France

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:35 pm

Dann Corbit wrote:
Dio wrote:
The problem with Arena is the file "engine.ini". This file can be deleted without any problems. In this file some engines are assigned wrong authors or the engine is declared as WB engine.
I get half the NPS and 50% CPU utilization.
For me, this makes Prodeo a non-starter.

In short, the algorithm does not work for Threadripper.
How about a radio button to ignore the limitation?


Betcha there's no more Elo though. Okay, since it's you, Ed will be in touch with you with a suitable solution, once I thought one up.
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:38 pm

Chris Whittington wrote:
Dann Corbit wrote:
Dio wrote:
The problem with Arena is the file "engine.ini". This file can be deleted without any problems. In this file some engines are assigned wrong authors or the engine is declared as WB engine.
I get half the NPS and 50% CPU utilization.
For me, this makes Prodeo a non-starter.

In short, the algorithm does not work for Threadripper.
How about a radio button to ignore the limitation?


Betcha there's no more Elo though. Okay, since it's you, Ed will be in touch with you with a suitable solution, once I thought one up.
Stockfish clearly does better, but if there is an SMP limitation with Prodeo, you might be right.
I would still like to try it, at least.

It is the same reason that I almost never use sting (which is even worse but this one is purely an SMP limitation).
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:40 pm

As an aside, I know some correspondence chess players and all of them use threadrippers.
Back to top Go down
Chris Whittington




Posts : 1254
Join date : 2020-11-17
Location : France

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:52 pm

Chris Whittington wrote:
Dann Corbit wrote:
Dio wrote:
The problem with Arena is the file "engine.ini". This file can be deleted without any problems. In this file some engines are assigned wrong authors or the engine is declared as WB engine.
I get half the NPS and 50% CPU utilization.
For me, this makes Prodeo a non-starter.

In short, the algorithm does not work for Threadripper.
How about a radio button to ignore the limitation?


Betcha there's no more Elo though. Okay, since it's you, Ed will be in touch with you with a suitable solution, once I thought one up.

Okay, done it. If Ed is in accord, he'll be in touch
Back to top Go down
Chris Whittington




Posts : 1254
Join date : 2020-11-17
Location : France

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:54 pm

Dann Corbit wrote:
As an aside, I know some correspondence chess players and all of them use threadrippers.

I use a threadripper 64x - that is not happy with 64 CSTal threads
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:55 pm

Chris Whittington wrote:
Chris Whittington wrote:
Dann Corbit wrote:
Dio wrote:
The problem with Arena is the file "engine.ini". This file can be deleted without any problems. In this file some engines are assigned wrong authors or the engine is declared as WB engine.
I get half the NPS and 50% CPU utilization.
For me, this makes Prodeo a non-starter.

In short, the algorithm does not work for Threadripper.
How about a radio button to ignore the limitation?


Betcha there's no more Elo though. Okay, since it's you, Ed will be in touch with you with a suitable solution, once I thought one up.

Okay, done it. If Ed is in accord, he'll be in touch
Thanks for helping a senile old coot, whose whinging twit tendencies can never be fully addressed.
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyThu Feb 23, 2023 8:57 pm

Chris Whittington wrote:
Dann Corbit wrote:
As an aside, I know some correspondence chess players and all of them use threadrippers.

I use a threadripper 64x - that is not happy with 64 CSTal threads

There are two ways the machine can be configured. To use cores (in which case 32 would be a hard limit in my case) or to use threading with two threads per core. I do not remember if I had to do something to change to 64 threads or not, but I guess you might have the core limit as your config.
Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 10:49 am

Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with Stockfish.

Unless Ed has fixed Rebel you gain nothing after 8 cores.

Hello Ed,

I seem to be having a issue with Rebel 15.1 I thought you would like to see it.

Everything looks great in the scaling test. Until I jump from 8 to 16 cores.. I also notice only a slight increase in NPS going from 8 to 16 cores. It is more, but only about 1 to 2 million NPS more. That seems low.

Test is still running. More games to play.
Read message Empty

Read message EmptyThu Jul 28, 2022 1:49 am    Read message EmptyTopic: Re: Rebel 15.1
Hi Mark,

Your observation is correct, the lazy smp is not implemented very well.

It's on the to-do list.

Thanks for testing.

Ed

Code:
Rank Name                          Elo     +/-   Games    Wins  Losses   Draws   Points   Score    Draw
   1 Rebel-15.1-avx2 8 cores        37      16     280      38       8     234    155.0   55.4%   83.6%
   2 Rebel-15.1-avx2 16 cores       23      14     282      26       7     249    150.5   53.4%   88.3%
   3 Rebel-15.1-avx2 4 cores        11      14     280      22      13     245    144.5   51.6%   87.5%
   4 Rebel-15.1-avx2 2 cores       -23      16     282      13      32     237    131.5   46.6%   84.0%
   5 Rebel-15.1-avx2 1 core        -49      16     280       3      42     235    120.5   43.0%   83.9%

702 of 1000 games finished.
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 10:56 am

mwyoung wrote:
Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with stockfish.
It depends.
If I want to use my machine for anything else while I am analyzing, then I can only allocate 31/32 threads, which is 97%. But if I use threads, I can get 63/64=98%.

Sometimes I want 30 threads for engine x and 30 threads for engine y.
The cost of doing things this way is that the CPU runs hotter. But I use a 20 Amp fan to cool my machine.
Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 10:57 am

Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with stockfish.
It depends.
If I want to use my machine for anything else while I am analyzing, then I can only allocate 31/32 threads, which is 97%. But if I use threads, I can get 63/64=98%.

Sometimes I want 30 threads for engine x and 30 threads for engine y.
The cost of doing things this way is that the CPU runs hotter. But I use a 20 Amp fan to cool my machine.

I know it was true with Rebel 15. And with Stockfish.
Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 11:10 am

Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with stockfish.
It depends.
If I want to use my machine for anything else while I am analyzing, then I can only allocate 31/32 threads, which is 97%. But if I use threads, I can get 63/64=98%.

Sometimes I want 30 threads for engine x and 30 threads for engine y.
The cost of doing things this way is that the CPU runs hotter. But I use a 20 Amp fan to cool my machine.

You do know that percent of CPU has nothing to do with how fast Stockfish is really running.
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 11:13 am

mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with stockfish.
It depends.
If I want to use my machine for anything else while I am analyzing, then I can only allocate 31/32 threads, which is 97%. But if I use threads, I can get 63/64=98%.

Sometimes I want 30 threads for engine x and 30 threads for engine y.
The cost of doing things this way is that the CPU runs hotter. But I use a 20 Amp fan to cool my machine.

You do know that percent of CPU has nothing to do with how fast Stockfish is really running.

I have many applications that definitely benefit from large thread count (some database systems, some scientific calculations, etc.)
I use my machine for a lot more than chess.

I like a fine level of control.
Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 11:19 am

Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with stockfish.
It depends.
If I want to use my machine for anything else while I am analyzing, then I can only allocate 31/32 threads, which is 97%. But if I use threads, I can get 63/64=98%.

Sometimes I want 30 threads for engine x and 30 threads for engine y.
The cost of doing things this way is that the CPU runs hotter. But I use a 20 Amp fan to cool my machine.

You do know that percent of CPU has nothing to do with how fast Stockfish is really running.

I have many applications that definitely benefit from large thread count (some database systems, some scientific calculations, etc.)  
I use my machine for a lot more than chess.

I like a fine level of control.

I do nothing but chess.

I can say for a fact. That Stockfish and Rebel 15 have issues with over threading.

And this is easy to test.
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 11:20 am

mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with stockfish.
It depends.
If I want to use my machine for anything else while I am analyzing, then I can only allocate 31/32 threads, which is 97%. But if I use threads, I can get 63/64=98%.

Sometimes I want 30 threads for engine x and 30 threads for engine y.
The cost of doing things this way is that the CPU runs hotter. But I use a 20 Amp fan to cool my machine.

You do know that percent of CPU has nothing to do with how fast Stockfish is really running.

I have many applications that definitely benefit from large thread count (some database systems, some scientific calculations, etc.)  
I use my machine for a lot more than chess.

I like a fine level of control.

I do nothing but chess.

I can say for a fact. That Stockfish and Rebel 15 have issues with over threading.

And this is easy to test.
SF seems to perform well on tcec, which uses a boatload of threads.
Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 11:23 am

Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with stockfish.
It depends.
If I want to use my machine for anything else while I am analyzing, then I can only allocate 31/32 threads, which is 97%. But if I use threads, I can get 63/64=98%.

Sometimes I want 30 threads for engine x and 30 threads for engine y.
The cost of doing things this way is that the CPU runs hotter. But I use a 20 Amp fan to cool my machine.

You do know that percent of CPU has nothing to do with how fast Stockfish is really running.

I have many applications that definitely benefit from large thread count (some database systems, some scientific calculations, etc.)  
I use my machine for a lot more than chess.

I like a fine level of control.

I do nothing but chess.

I can say for a fact. That Stockfish and Rebel 15 have issues with over threading.

And this is easy to test.
SF seems to perform well on tcec, which uses a boatload of threads.

That is because Stockfish is the best no matter how many threads are used.

I suggest you download the PGN files from TCEC as I did, and see how fast stockfish is running on their system. And compare the time to depth times with your system.

You will be shocked!
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptyFri Feb 24, 2023 11:34 am

mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
mwyoung wrote:
Dann Corbit wrote:
There are two things I have found, one is a problem with Arena (Rebel 16.2 is automatically installed as a Winboard engine).
The solution to this one is simple, manually change the type to UCI.

The second problem is a bug in Rebel (I think).  My machine has 32 cores, but 64 threads.  It is AMD 3970x and normally Arena posts a warning about too many cores (e.g. with Stockfish) but lets me use all the threads that I ask for (typically between 59 and 62 threads).  However, Rebel seems to be detecting the core count and limiting to 32 threads.  Because of this, I cannot use more than half of my CPU power.

You are better and faster to use just cores anyone. And the same with stockfish.
It depends.
If I want to use my machine for anything else while I am analyzing, then I can only allocate 31/32 threads, which is 97%. But if I use threads, I can get 63/64=98%.

Sometimes I want 30 threads for engine x and 30 threads for engine y.
The cost of doing things this way is that the CPU runs hotter. But I use a 20 Amp fan to cool my machine.

You do know that percent of CPU has nothing to do with how fast Stockfish is really running.

I have many applications that definitely benefit from large thread count (some database systems, some scientific calculations, etc.)  
I use my machine for a lot more than chess.

I like a fine level of control.

I do nothing but chess.

I can say for a fact. That Stockfish and Rebel 15 have issues with over threading.

And this is easy to test.
SF seems to perform well on tcec, which uses a boatload of threads.

That is because Stockfish is the best no matter how many threads are used.

I suggest you download the PGN files from TCEC as I did, and see how fast stockfish is running on their system. And compare the time to depth times with your system.

You will be shocked!
I may do that, it should be a fairly simple experiment.

My machine is atypical. I have 128gb ram, a gigantic raid disk farm, and 8TB of PCie 4.0 SSD. I also have two Nvidia 2080 super gpus. This is very useful if I want (for example) 60 threads of SMP (perhaps with two different engines) and lc0 using my gpus and two or three threads.

I also do things that hurt performance, since I want the right answers. For instance, a huge Tablebase system probably subtracts Elo, but once in a while it causes better analysis.

I guess if I used only bitbase, there would be better performance, but I am far more concerned about correctness. That's why I am horrified by engines that skip some under-promotions.
Back to top Go down
Dann Corbit




Posts : 182
Join date : 2020-11-26

Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper EmptySat Feb 25, 2023 7:33 am

AMD Ryzen Threadripper 3970X (62 threads used by Stockfish, Test, 32 threads used by Rebel)
2x NVIDIA GeForce RTX 2080 SUPER (both used by LC0, no ponder)

  Engine             Score         St        Lc        Te        Lc        Re    S-B
1: Stockfish-x64-avx2 22.5/34 ········· ======1== ===1===== ====11=1  =1111=11   336.75
2: Lc0-T78            16.5/33 ======0== ········· ========  ========  =====1==   272.75
3: Test-62t           16.0/33 ===0===== ========  ········· ========  ========   268.00
4: Lc0-T60            14.5/33 ====00=0  ========  ========  ········· ====0====  240.25
5: Rebel-16.2         13.5/33 =0000=00  =====0==  ========  ====1==== ·········  216.75

83 of 600 games played
Name of the tournament: Tdog
Site/ Country: DESKTOP-426Q3JG, United States
Level: Blitz 5/2
Hardware: AMD Ryzen Threadripper 3970X 32-Core Processor with 63.9 GB Memory
Operating system: Windows 10 Home Home Edition (Build 9200) 64 bit
PGN-File: C:\Program Files (x86)\Arena\Tournaments\Tdog.pgn
Website:
E-Mail Address:


First game:
[Event "Tdog"]
[Site "DESKTOP-426Q3JG"]
[Date "2023.02.24"]
[Round "1"]
[White "Lc0-T60"]
[Black "Lc0-T78"]
[Result "1/2-1/2"]
[BlackElo "3500"]
[ECO "D12"]
[Opening "Slav"]
[Time "00:10:43"]
[Variation "4.e3 Bf5 5.Nc3 e6 6.Nh4"]
[WhiteElo "3500"]
[TimeControl "300+2"]
[Termination "normal"]
[PlyCount "111"]
[WhiteType "program"]
[BlackType "program"]

1. d4 {(d4 Nf6 c4 e6 Nf3 d5 Nc3 c6 e3 Nbd7 Be2 Bd6 0-0 0-0 b3) +0.07/8 6}
d5 {(d5 c4 c6 Nc3 Nf6 Nf3 e6 Bg5 h6 Bh4 dxc4 e4 g5 Bg3 b5 Be2 Bb7 h4 g4 Ne5
Nbd7 0-0 Nxe5 Bxe5 Rg8 g3 a6 Re1 Rg6) -0.12/11 9} 2. c4 {(c4 c6 Nf3 Nf6 e3
Bf5 Nh4 Bc8 Qc2 e6 Nf3 Bd6 b3 0-0 Bb2 Nbd7 Bd3) +0.09/11 5} c6 {(c6 Nf3 Nf6
Nc3 e6 e3 Nbd7 Qc2 Bd6 Bd3 0-0 0-0 e5 cxd5 cxd5 e4 exd4 Nxd5 Nxd5 exd5 h6
Nxd4 Qh4 Nf3 Qh5 Bh7+ Kh8 Qf5 Qxf5) -0.13/13 4} 3. Nf3 {(Nf3 Nf6 e3 Bf5 Nc3
e6 Nh4 Bg6 Nxg6 hxg6 Bd2 Nbd7 cxd5 exd5 Qc2 Bd6 0-0-0 0-0 Kb1 Rc8 h4)
+0.10/12 11} Nf6 {(Nf6 Nc3 e6 Bg5 h6 Bh4 dxc4 e4 g5 Bg3 b5 Be2 Bb7 h4 g4
Ne5 Nbd7 0-0 Nxe5 Bxe5 Rg8 g3 a6 Re1 Rg6 Qc2 Nh5 d5 Bc5 Rad1 Qb6 Bf1)
-0.12/12 4} 4. e3 {(e3 Bf5 Nc3 e6 Nh4 Bg6 Bd2 Nbd7 Nxg6 hxg6 cxd5 exd5 Qc2
Bd6 0-0-0 0-0 Kb1 b5 f3 Rc8 e4) +0.11/12 4} Bf5 {(Bf5 Nh4 Be6 Nd2 g6 b3 Bg7
Bb2 0-0 Be2 c5 Nhf3 Ne4 0-0 Nc6 Rc1 cxd4 Nxe4) -0.11/10 7} 5. Nc3 {(Nc3 e6
Nh4 Bg6 Nxg6 hxg6 Bd2 Nbd7 cxd5 exd5 Qc2 Bd6 0-0-0 0-0 Kb1 Rc8 h4 b5)
+0.11/13 7} e6 {(e6 Nh4 Bg6 Bd2 Nbd7 Nxg6 hxg6 Qb3 Qb6 Be2 Be7 h3 0-0 g4
dxc4 Bxc4) -0.11/11 6} 6. Nh4 {(Nh4 Be4 f3 Bg6 Bd2 Nbd7 cxd5 Nxd5 Nxg6 hxg6
g3 Be7 Kf2 0-0 h4 N5f6 Kg2 c5 dxc5 Nxc5 e4) +0.12/15 2} Bg6 {(Bg6 Bd2 Nbd7
Nxg6 hxg6 cxd5 exd5 Qc2 Bd6 0-0-0 Ng4 Be1 Rxh2 Rxh2 Nxh2 Be2 Qg5 g3 Nf6 f4
Qf5 Bd3 Qh5 e4 dxe4 Nxe4 Nxe4) -0.12/12 10} 7. Bd2 {(Bd2 Nbd7 Nxg6 hxg6
cxd5 exd5 Qc2 Bd6 0-0-0 0-0 Kb1 Rc8 f3 b5 e4 c5 Nxb5 cxd4 Qa4 Bc5 Ba5 Nb6
Qb3) +0.12/11 17} Nbd7 {(Nbd7 Nxg6 hxg6 cxd5 exd5 Qc2 Bd6 0-0-0 Ng4 Be1
Rxh2 Rxh2 Nxh2 Be2 Qg5 g3 Nf6 f4 Qf5 Bd3 Qh3 e4 dxe4 Nxe4 Nxe4 Bxe4 Kf8 d5
c5) -0.11/10 5} 8. Nxg6 {(Nxg6 hxg6 cxd5 exd5 Qc2 Bd6 0-0-0 0-0 Kb1 Rc8 f3
b5 e4 c5 Nxb5 cxd4 Qa4 Bc5 Ba5 Nb6 Qb3) +0.12/11 2} hxg6 {(hxg6 Qb3 Rb8 Be2
Be7 0-0 0-0 Rfd1 Re8 a4 a5 g3) -0.12/9 4} 9. cxd5 {(cxd5 exd5 Qc2 Bd6 0-0-0
0-0 f3 b5 Kb1 Rc8 g4 b4 Na4 c5 dxc5 Nxc5 Nxc5 Rxc5 Qd3 a5 h4 Qc7 h5 g5 h6)
+0.13/11 12} exd5 {(exd5 Qc2 Bb4 f3 Qc7 g3 Bd6 0-0-0 Bxg3 Kb1 0-0-0 e4 dxe4
fxe4 Bf4 Bc4 Bxd2 Rxd2 Nb6 Bb3 Kb8 Rf1 Rd7 a4 a5 e5 Nfd5) -0.08/11 8} 10.
Qc2 {(Qc2 Bd6 0-0-0 0-0 f3 Rc8 Kb1 b5 g4 b4 Na4 c5 dxc5 Nxc5 Nxc5 Rxc5 Qd3
Qb6 g5 Nd7 f4 Ra5 h4 Qc6 Bg2 Qa4) +0.15/13 2} Bd6 {(Bd6 0-0-0 Ng4 Be1 Rxh2
Rxh2 Nxh2 Be2 Qg5 g3 Nf6 f4 Qf5 Bd3 Qh3 e4 dxe4 Nxe4 Nxe4 Bxe4 0-0-0 Bxc6
bxc6 Qxc6+ Kb8 Qb5+ Ka8 Qc6+ Kb8) -0.08/17 14} 11. O-O-O {(0-0-0 0-0 h4 b5
Kb1 Rc8 e4 b4 e5 bxc3 Bxc3 c5 exd6 Ne4 h5 cxd4 Rxd4 Nxc3+ bxc3 Qb6+ Qb3
Qxd6 Qxd5 Rb8+ Kc2 Qa3 Bb5) +0.18/15 3} Ng4 {(Ng4 Be1 Rxh2 Rxh2 Nxh2 Be2
Qg5 g3 Nf6 f4 Qf5 Bd3 Qh3 e4 dxe4 Nxe4 Nxe4 Bxe4 0-0-0 Bxc6 bxc6 Qxc6+ Kb8
Qb5+ Ka8 Qc6+ Kb8) -0.06/22 1} 12. Be1 {(Be1 Rxh2 Rxh2 Nxh2 Be2 Qg5 g3 Ng4
Ne4 dxe4 Qxe4+ Kf8 Qxg4 Qxg4 Bxg4 f5 Bf3 Nf6 Bc3 g5 Rh1 Kf7 Kc2 g4 Bg2 Re8
Kd3 a6 a4) +0.14/21 7} Rxh2 {(Rxh2 Rxh2 Nxh2 Be2 Qg5 g3 Nf6 f4 Qf5 Bd3 Qh3
e4 dxe4 Nxe4 Nxe4 Bxe4 0-0-0 d5 c5 b4 c4 Bf2 f5 Qxc4+ Kb8 Qd4 b6 Bd3 Ng4
Ba6 Qh8 Be1 Qe8 Qxg7 Rd7 Qd4 Re7 Kb2 Re4 Qg7 Re7 Qc3 Rc7 Qd4 Ne3 Rc1 Rxc1
Kxc1 Nxd5 Bd2 Qc6+ Kb1 b5) -0.07/25 2} 13. Rxh2 {(Rxh2 Nxh2 Be2 Qg5 g3 Ng4
Ne4 dxe4 Qxe4+ Kf8 Qxg4 Qxg4 Bxg4 f5 Bf3 Nf6 Bc3 g5 Rh1 Kf7 Kc2 g4 Bg2 Re8
Kd3 a6 a4) +0.12/16 0} Nxh2 {(Nxh2 Be2 Qg5 g3 Nf6 f4 Qf5 Bd3 Qh3 e4 dxe4
Nxe4 Nxe4 Bxe4 0-0-0 d5 c5 b4 Kb8 bxc5 Bxc5 d6 Bxd6 Kb1 Be7 Rxd8+ Bxd8 Qd2
Kc8 Qc3+ Bc7 Qb4 Qf1 Qxb7+ Kd7 Qc6+ Kd8 Qd5+ Ke8 Qc6+ Kd8) -0.06/27 4} 14.
Be2 {(Be2 Qg5 g3 Ng4 Ne4 dxe4 Qxe4+ Kf8 Qxg4 Qxg4 Bxg4 f5 Bf3 Nf6 Bc3 g5
Rh1 Kf7 Kc2 g4 Bg2 Re8 Kd3 a6 a4) +0.11/20 8} Qg5 {(Qg5 g3 Nf6 f4 Qf5 Bd3
Qh3 e4 dxe4 Nxe4 Nxe4 Bxe4 0-0-0 d5 c5 b4 Kb8 bxc5 Bxc5 d6 Bxd6 Kb1 Be7
Rxd8+ Bxd8 Qd2 Kc8 Qc3+ Bc7 Qb4 Qf1 Qxb7+ Kd7 Qc6+ Kd8 Qd5+ Ke8 Qc6+ Kd8)
-0.06/27 0} 15. g3 {(g3 Ng4 Ne4 dxe4 Qxe4+ Kf8 Qxg4 Qxg4 Bxg4 f5 Bf3 Nf6
Bc3 g5 Rh1 Kf7 Kc2 g4 Bg2 Re8 Kd3 a6 a4) +0.11/11 1} Nf6 {(Nf6 f4 Qf5 Bd3
Qh3 e4 dxe4 Nxe4 Nxe4 Bxe4 0-0-0 d5 c5 b4 Kb8 bxc5 Bxc5 d6 Bxd6 Kb1 Be7
Rxd8+ Bxd8 Qd2 Kc8 Qc3+ Bc7 Qb4 Qf1 Qxb7+ Kd7 Qc6+ Kd8 Qc3 Qb5+ Kc1 Bd6 Bd3
Qc5 Qxc5 Bxc5 Kd1 Nf3 Bc3) -0.01/24 11} 16. f4 {(f4 Qf5 Bd3 Qh3 e4 Be7 exd5
cxd5 f5 gxf5 Bxf5 Nhg4 Qb3 Rd8 Qxb7 Kf8 Kb1 g6 Bxg4 Qxg4 Qxa7 Ne4 Qa4 Nxg3
Bxg3 Qxg3 Qb5 Kg7 a4 Bf6 Ka2 Qg4 Qe2) +0.02/17 14} Qf5 {(Qf5 Bd3 Qh3 e4
dxe4 Nxe4 Nxe4 Bxe4 0-0-0 d5 c5 b4 Kb8 bxc5 Bxc5 d6 Bxd6 Kb1 Be7 Rxd8+ Bxd8
Qd2 Kc8 Qc3+ Bc7 Qb4 Qf1 Qxb7+ Kd7 Qc6+ Kd8 Qc3 Qb5+ Kc1 Bd6 Bd3 Qc5 Qxc5
Bxc5 Kd1 Nf3 Bc3) -0.03/21 0} 17. Bd3 {(Bd3 Qh3 e4 Be7 Qb3 0-0-0 exd5 cxd5
Kb1 Kb8 Na4 Nf3 Ba6 Qd7 Bxb7 Qxb7 Qxf3 Ne4 Nc5 Bxc5 dxc5 Nxc5 Ba5 Rd7 Qc3
Qb5 Rc1 Qd3+) +0.02/18 7} Qh3 {(Qh3 e4 dxe4 Nxe4 Nxe4 Bxe4 0-0-0 d5 c5 b4
Kb8 bxc5 Bxc5 d6 Bxd6 Kb1 Be7 Rxd8+ Bxd8 Qd2 Kc8 Qc3+ Bc7 Qb4 Qf1 Qxb7+ Kd7
Qc6+ Kd8 Qc3 Qb5+ Kc1 Bd6 Bc2 Qc5 Qxg7 Qe3+ Kd1 Qf3+ Kc1 Qe3+) -0.02/22 8}
18. e4 {(e4 Be7 exd5 cxd5 Qb3 0-0-0 Na4 Kb8 Kb1 Nf3 Ba6 Qd7 Bxb7 Qxb7 Qxf3
Ne4 Nc5 Nxc5 dxc5 Bxc5 Ba5 Bb6 Bxb6 axb6 Qe3 Qc6 Qe5+ Kb7 Qxg7 Qe6 Qd4 f5)
0.00/18 7} dxe4 {(dxe4 Nxe4 Nxe4 Bxe4 0-0-0 d5 c5 b4 Kb8 bxc5 Bxc5 d6 Bxd6
Kb1 Be7 Rxd8+ Bxd8 Qd2 Kc8 Qc3+ Bc7 Qb4 Qf1 Qxb7+ Kd7 Qc6+ Kd8 Qc3 Qb5+ Kc1
Bd6 Bc2 Qc5 Qd3 Ng4 Bc3 Qe3+ Kb2 Qxd3) -0.01/22 5} 19. Nxe4 {(Nxe4 Nxe4
Bxe4 0-0-0 Bxc6 bxc6 Qxc6+ Kb8 Qb5+ Ka8 Qc6+ Kb8) +0.03/14 11} Nxe4 {(Nxe4
Bxe4 0-0-0 d5 c5 b4 Kb8 bxc5 Bxc5 d6 Bxd6 Kb1 Be7 Rxd8+ Bxd8 Qd2 Kc8 Qc3+
Bc7 Qb4 Qf1 Qxb7+ Kd7 Qc6+ Kd8 Qc3 Qb5+ Kc1 Bd6 Bc2 Qc5 Qd3 Ng4 Kb1 Ke7
Qe4+ Kf8 Qa8+) 0.00/20 6} 20. Bxe4 {(Bxe4 0-0-0 Bxc6 bxc6 Qxc6+ Kb8 Qb5+
Ka8 Qc6+ Kb8) +0.03/15 0} O-O-O {(0-0-0 d5 c5 b4 Kb8 bxc5 Bxc5 d6 Bxd6 Kb1
Be7 Rxd8+ Bxd8 Qd2 Kc8 Qc3+ Bc7 Qb4 Qf1 Qxb7+ Kd7 Qc6+ Kd8 Qc3 Qb5+ Kc1 Bd6
Bc2 Qc5 Qd3 Ng4 Kb1 Ke7 Qe4+ Kf8 Qa8+) 0.00/17 6} 21. Bxc6 {(Bxc6 bxc6
Qxc6+ Kb8 Qb5+ Ka8 Qc6+ Kb8) +0.01/13 0} bxc6 {(bxc6 Ba5 Kd7 Bxd8 Kxd8 Qxc6
Qc8 Qxc8+ Kxc8 Rh1 Nf3 Rh8+ Kc7 Rg8 Kd7 Ra8 Nxd4 Rxa7+ Bc7 Ra3 Ne6 b4 Bd6
Rd3 Kc6 b5+ Kc5 a4 g5 fxg5 Nxg5) 0.00/17 18} 22. Ba5 {(Ba5 Kd7 Bxd8 Kxd8
Qxc6 Qc8 Qxc8+ Kxc8 Rh1 Nf3 Rh8+ Kc7 Rg8 Kd7 Ra8 Nxd4 Rxa7+ Bc7 Ra3 Ne6 Kc2
g5 fxg5 Nxg5 g4 g6 Rd3+ Kc6 Rc3+ Kb7 Rc5 Ne4) +0.02/16 33} Kd7 {(Kd7 Bxd8
Kxd8 Qxc6 Qc8 Qxc8+ Kxc8 Rh1 Nf3 Rh8+ Kc7 Rg8 Kd7 Ra8 Nxd4 Rxa7+ Bc7 Ra3
Bd6 Rd3 Ne6 b4 Kc7 a3 g5 f5 Nd8 Kc2 Nc6 Kb3 g4 b5 Na5+) 0.00/22 9} 23. Bxd8
{(Bxd8 Kxd8 Qxc6 Qc8 Qxc8+ Kxc8 Rh1 Nf3 Rh8+ Kc7 Rg8 Kd7 Ra8 Nxd4 Rxa7+ Bc7
Ra3 Ne6 Kc2 g5 fxg5 Nxg5 g4 g6 Rd3+ Kc6 Rc3+ Kb7 Rc5 Ne4) +0.01/22 0} Kxd8
{(Kxd8 Qxc6 Qc8 Qxc8+ Kxc8 Rh1 Nf3 Rh8+ Kc7 Rg8 Kd7 Ra8 Nxd4 Rxa7+ Bc7 Ra3
Ne6 Kc2 g5 fxg5 Nxg5 g4 g6 Rd3+ Kc6 Rc3+ Kb7 a4 f5) 0.00/22 1} 24. Qxc6
{(Qxc6 Qc8 Qxc8+ Kxc8 Rh1 Nf3 Rh8+ Kc7 Rg8 Kd7 Ra8 Nxd4 Rxa7+ Bc7 Ra3 Ne6
Kc2 g5 fxg5 Nxg5 g4 g6 Rd3+ Kc6 Rc3+ Kb7 Rc5 Ne4) +0.01/21 1} Qc8 {(Qc8
Qxc8+ Kxc8 Rh1 Nf3 Rh8+ Kb7 Rg8 Nxd4 Rxg7 Ne2+ Kc2 Nxg3 Rxf7+ Bc7 Kd3 Nh5
f5 gxf5 Rxf5 Ng7 Rf7 Ne6 Kc4 Nd8 Rh7 Nc6 Kc5) -0.01/21 32} 25. Qxc8+
{(Qxc8+ Kxc8 Rh1 Nf3 Rh8+ Kc7 Rg8 Kd7 Ra8 Nxd4 Rxa7+ Bc7 Ra3 Bd6 Rd3 Ne6 b4
Kc7 a3 f5 Rd5 Kc6 Ra5 Bc7 Ra6+ Bb6 Ra8 Bf2 Ra6+) +0.02/19 2} Kxc8 {(Kxc8
Rh1 Nf3 Rh8+ Kb7 Rg8 Nxd4 Rxg7 Ne2+ Kc2 Nxg3 Rxf7+ Bc7 Kd3 Nh5 f5 gxf5 Rxf5
Ng7 Rf7 Ne6 Kc4 Nd8 Rh7 Nc6 Kc5) -0.02/19 0} 26. Rh1 {(Rh1 Nf3 Rh8+ Kc7 Rg8
Kd7 Ra8 Nxd4 Rxa7+ Bc7 Ra3 Bd6 Rd3 Ne6 Kc2 Kc6 a4 Nc5 Rd4 Ne6 Rc4+ Kd5 Kc3
g5 fxg5) +0.03/17 4} Nf3 {(Nf3 Rh8+ Kb7 Rg8 Nxd4 Rxg7 Ne2+ Kc2 Nxg3 Rxf7+
Bc7 Kd3 Nh5 f5 gxf5 Rxf5 Ng7 Rf7 Ne6 Kc4 Nd8 Rh7 Nc6 Kc5) -0.02/17 0} 27.
Rh8+ {(Rh8+ Kc7 Rg8 Kd7 Ra8 Nxd4 Rxa7+ Bc7 Ra3 Bd6 Rd3 Ne6 Kc2 Kc6 a4 g5
fxg5 Nxg5 Rc3+ Kb6 g4 Ne6 Rc4 Ka5) +0.03/15 5} Kb7 {(Kb7 Rg8 Nxd4 Rxg7 Ne2+
Kc2 Nxg3 Rxf7+ Bc7 Kd3 Nh5 f5 gxf5 Rxf5 Ng7 Rf7 Ne6 Kc4 Nd8 Rh7 Nc6 Kc5)
-0.03/16 13} 28. Rg8 {(Rg8 Nxd4 Rxg7 Ne2+ Kc2 Nxg3 Rxf7+ Bc7 Kd3 Nh5 f5
gxf5 Rxf5 Ng7 Rf8 Ne6 Rf7 Nd8 Rh7 Kc6) +0.04/14 16} Nxd4 {(Nxd4 Rxg7 Ne2+
Kc2 Nxg3 Rxf7+ Bc7 Kd3 Nh5 f5 gxf5 Rxf5 Ng7 Rf7 Ne6 Kc4 Nd8 Rh7 Nc6 Kc5)
-0.03/14 0} 29. Rxg7 {(Rxg7 Ne2+ Kc2 Nxg3 Rxf7+ Bc7 Kd3 Nh5 f5 gxf5 Rxf5
Ng7 Rf8 Ne6 Rf7 Nc5+ Kc4 Ne4) +0.03/13 4} Ne2+ {(Ne2+ Kc2 Nxg3 Rxf7+ Bc7
Kd3 Nh5 f5 gxf5 Rxf5 Ng7 Rf7 Ne8 Re7 Nd6 b4 Kc6 a4) -0.02/12 0} 30. Kc2
{(Kc2 Nxg3 Rxf7+ Bc7 Kd3 Nh5 f5 gxf5 Rxf5 Ng7 Rf8 Ne6 Rf7 Nd8 Rh7 Kc6 Kc4)
+0.02/11 28} Nxg3 {(Nxg3 Rxf7+ Bc7 Kd3 Nh5 f5 gxf5 Rxf5 Ng7 Rf6 Ne8 Rg6 Nd6
Kd4 Nc8 Kc5 Bb6+) -0.01/11 23} 31. Rxf7+ {(Rxf7+ Bc7 Kd3 Nh5 f5 gxf5 Rxf5
Ng7 Rf8 Ne6 Rf7 Nd8 Rh7 Kc6 Kc4) +0.02/10 1} Bc7 {(Bc7 Kd3 Nh5 f5 gxf5 Rxf5
Ng7 Rf6 Ne8 Rg6 Nd6 Kd4 Nc8 Kc5 Bb6+) -0.01/10 0} 32. Kd3 {(Kd3 Nh5 f5 gxf5
Rxf5 Ng7 Rf8 Ne6 Rf7 Nd8 Rh7 Kc6 Kc4) +0.01/9 8} Nh5 {(Nh5 f5 gxf5 Rxf5 Ng7
Rf7 Ne6 Kc4 Nd8 Rh7 Nc6 Kc5 Ne5 b4) 0.00/9 16} 33. f5 {(f5 gxf5 Rxf5 Ng7
Rf8 Ne6 Rf7 Nd8 Rh7 Nc6 Kc4 Kb6) +0.02/9 8} gxf5 {(gxf5 Rxf5 Ng7 Rf7 Ne6
Kc4 Nd8 Rh7 Nc6 Kc5 Ne5 b4 Nd3+) -0.01/9 8} 34. Rxf5 {(Rxf5 Ng7 Rf8 Ne6 Rf7
Nd8 Rh7 Nc6 Kc4 Kb6) +0.03/8 4} Ng7 {(Ng7 Rf7 Ne8 Kd4 Kc6 Rh7 Nd6 Rh5 Bb6+
Kd3) -0.02/8 14} 35. Rf8 {(Rf8 Ne6 Rf7 Nd8 Rh7 Kc6 b4 a6 a4 Nb7) +0.03/7
17} Ne6 {(Ne6 Rf5 Nd8 Kc4 Nc6 Rb5+ Bb6 Rh5) -0.01/7 7} 36. Rf7 {(Rf7 Nd8
Rh7 Kc6 b4 a6 a4 Nb7) +0.01/6 8} Nd8 {(Nd8 Rh7 Nc6 Kc4 Ne5+ Kd5 Nd3 b3 Nb4+
Kc4) -0.01/6 10} 37. Rh7 {(Rh7 Nc6 Kc4 Kb6 Rh6 Kb7 Kc5 Nb8 Rh7) +0.01/6 7}
Nc6 {(Nc6 Kc4 a6 Kc5 Ne5 b4 Nd3+ Kc4 Ne5+ Kd5 Nd3) -0.01/6 8} 38. Kc4 {(Kc4
a6 Kd5 Nb4+ Kc4 Nc6) +0.02/6 7} Ne5+ {(Ne5+ Kd4 Nc6+ Kc5 Ne5 b4 Nd3+ Kc4
Ne5+ Kd5 Nd3 a3) -0.02/6 7} 39. Kd4 {(Kd4 Nc6+ Kc5 Nb8 b4 Kc8 b5 Nd7+ Kc6
Nb8+ Kd5) +0.03/6 6} Nc6+ {(Nc6+ Kc5 Ne5 b4 Nd3+ Kc4 Ne5+ Kd5 Nd3 a3 Nf4+
Kc4) -0.02/7 5} 40. Kc5 {(Kc5 Nb8 b4 Na6+ Kc4 Nb8 b5 Kc8 a4 Nd7 Kb4 Bd6+)
+0.02/7 7} Ne5 {(Ne5 b4 Nd3+ Kc4 Ne5+ Kd5 Nd3 a3 Nf4+ Kc4 Ng6 a4) -0.02/7
8} 41. Kd4 {(Kd4 Nc6+ Kd5 Nb4+ Kc4 Nxa2 Rh1 a5 Kb5 Nb4 Rh7 Nd3) +0.03/7 7}
Nc6+ {(Nc6+ Kc5 Ne5 b4 Nd3+ Kc4 Ne5+ Kd5 Nd3 a3 Nf4+ Kc4 Ng6) -0.02/7 5}
42. Kc5 {(Kc5 Ne5 Kd5 Nd3 b3 Nb4+ Kc5 Nxa2 Kc4 a5 Kb5) +0.02/7 7} Ne5 {(Ne5
b4 Nd3+ Kc4 Ne5+ Kd5 Nd3 a3 Nf4+ Kc4 Ng6 a4 Ne5+) -0.02/7 6} 43. b4 {(b4
Nd3+ Kc4 Ne5+ Kd5 Nd3 a3 a6 Kd4 Nf4 Rh6) +0.03/6 6} Nd3+ {(Nd3+ Kc4 Ne5+
Kd5 Nd3 a3 Nb2 Kc5 Na4+ Kb5 Nb6 a4) -0.02/7 8} 44. Kc4 {(Kc4 Ne5+ Kd5 Nd3
a3 a6 Rf7 Kb6 Kd4 Nb2) +0.03/7 5} Ne5+ {(Ne5+ Kd5 Nd3 a3 Nb2 Kc5 Na4+ Kb5
Nc3+ Kc4 Ne4) -0.02/7 5} 45. Kd5 {(Kd5 Nd3 b5 Kb6 a4 Nb2 a5+ Kxa5 Rxc7)
+0.03/7 7} Nd3 {(Nd3 Kc4 Ne5+ Kd4 Nc6+ Kc5 Ne5 a4 Nd3+ Kc4 Ne5+ Kd5 Nd3 b5)
-0.02/7 6} 46. Kc4 {(Kc4 Ne5+ Kd4 Nc6+ Kc5 Ne5 b5 Nd3+ Kc4 Ne5+ Kd5 Nd3 a4)
+0.03/7 5} Ne5+ {(Ne5+ Kd4 Nc6+ Kc5 a6 a4 Na7 Rg7 Kb8 Rg6 Kb7) -0.02/7 5}
47. Kd5 {(Kd5 Nd3 a3 a6 Re7 Kb6 Kd4 Nf4 a4 Kc6 Rh7) +0.03/7 6} Nd3 {(Nd3 a3
Nb2 Kc5 Na4+ Kc4 Nb6+ Kc5 Na4+) -0.01/7 5} 48. a3 {(a3 a6 Re7 Kb6 Kc4 Ne5+
Kd5 Nd3) +0.03/7 4} a6 {(a6 Rf7 Nb2 Rg7 Na4 Kc4 Nb6+ Kc5 Na4+ Kc4) -0.01/7
7} 49. Kc4 {(Kc4 Nf4 a4 Ne6 Kd5 Nf4+ Ke4 Ne2 Kd3) +0.02/6 6} Nb2+ {(Nb2+
Kc3 Na4+ Kb3 Nb6 a4 Nc8 Kc4 Nb6+ Kb3) -0.02/7 5} 50. Kd4 {(Kd4 Na4 Kd5 Nc3+
Kc4 Nb5 a4 Nd6+ Kd5 Nc8) +0.02/7 5} Na4 {(Na4 Kc4 Nb6+ Kc5 Na4+ Kd4 Nb6 Rh6
Na4 Kc4 Nb6+) -0.01/6 3} 51. Kc4 {(Kc4 Nb6+ Kc5 Na4+ Kc4) +0.02/6 4} Nb6+
{(Nb6+ Kc5 Na4+ Kd4 Nb6 Kc5) -0.01/7 4} 52. Kc5 {(Kc5 Na4+ Kc4 Nb6+)
+0.01/6 4} Na4+ {(Na4+ Kc4 Nb6+ Kb3 Nc8 Kc4 Kc6 Rh5 Nd6+ Kb3) -0.02/7 5}
53. Kc4 {(Kc4 Nb6+ Kc5 Na4+) +0.02/7 4} Nb6+ {(Nb6+ Kc5 Na4+ Kd4 Nb6 Kc5)
-0.01/7 3} 54. Kc5 {(Kc5 Na4+ Kd4 Nb6 Kc5) +0.01/7 3} Na4+ {(Na4+ Kd4 Nb6
Kc5) -0.01/7 3} 55. Kd4 {(Kd4 Nb6 Kc5) +0.01/6 4} Nb6 {(Nb6 Kc5) 0.00/6 4}
56. Kc5 {(Kc5) 0.00/6 3 3-fold repetition} 1/2-1/2
Back to top Go down
Sponsored content





Problem with Rebel 16.2 on AMD threadripper Empty
PostSubject: Re: Problem with Rebel 16.2 on AMD threadripper   Problem with Rebel 16.2 on AMD threadripper Empty

Back to top Go down
 
Problem with Rebel 16.2 on AMD threadripper
Back to top 
Page 1 of 2Go to page : 1, 2  Next
 Similar topics
-
» The Stockfish ELO problem
» New test set "Problem of the Week"
» Small Somu 1.7a problem ...
» Rebel EAS and CS Tal Bug
» Rebel in my FACE -Tournaments

Permissions in this forum:You cannot reply to topics in this forum
ProDeo :: Computer Chess-
Jump to: