Stockfish bench in i486 & Pentium 75mhz !

Discussion of anything and everything relating to chess playing software and machines.

Moderators: Harvey Williamson, Dann Corbit, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
mwyoung
Posts: 2556
Joined: Wed May 12, 2010 8:00 pm

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by mwyoung » Thu Oct 15, 2020 11:49 pm

Milos wrote:
Thu Oct 15, 2020 11:43 pm
mwyoung wrote:
Thu Oct 15, 2020 11:18 pm
Milos wrote:
Thu Oct 15, 2020 11:13 pm
mwyoung wrote:
Thu Oct 15, 2020 11:08 pm
Milos wrote:
Thu Oct 15, 2020 11:01 pm
mwyoung wrote:
Thu Oct 15, 2020 10:56 pm
Milos wrote:
Thu Oct 15, 2020 10:54 pm
mwyoung wrote:
Thu Oct 15, 2020 10:49 pm
Milos wrote:
Thu Oct 15, 2020 10:40 pm
mwyoung wrote:
Thu Oct 15, 2020 10:32 pm
Milos wrote:
Thu Oct 15, 2020 10:14 pm
lkaufman wrote:
Thu Oct 15, 2020 9:42 pm
mwyoung wrote:
Thu Oct 15, 2020 8:55 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
mclane wrote:
Tue Oct 13, 2020 10:06 am
I would like to see Stockfish run on a 6502 so we can try out how it plays against the old 8 bit dedicated chess computers.
SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess computers.
SF-NNUE on 6502 would have for sure more than 1000nodes per move. It wouldn't be any contest, it would be a slaughter.
You could not even engine test your way out of a paper bag. And your engine testing are as spot on as your conclusions. :lol:
As usual!

Sargon from 1978 vs Stockfish NNUE @ 1 ply (TC=5m+5s)

Live feed: https://www.youtube.com/watch?v=WZzsrGoxy7I
Sargon from 1978?? My recollection was that in 1978 the best chess computer one could buy ( excluding million dollar computers or the like) was no better than 1500 (probably much worse), and that was with 2 hours; at 5+5 should be like 1000 or worse. SF one ply shouldn't be that bad! What am I missing here? Chess computers one could buy weren't half-decent until early 1980s, unless my memory is quite wrong.
You are missing that a retard is playing an engine called Sargon on 2950x CPU and has 50kN/s. As usual all his tests are outright fake.
I know, and did you see the fake Mephisto Gideon professional by Ed Schröder crush Stockfish even worse....Ask Ed if MGP is also fake. :lol:

You can download the "fake" Sargon for UCI here. https://chessengines.blogspot.com/2020/ ... n-100.html
For quality trolling one needs some intelligence. You lack even the basic comprehension. Discussion that you hijacked was about running on 6502. You obviously don't even know what's that.
Milos---"SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess." computers.

Your conclusions, and memory, and your facts. :lol:
To spell for retard, dedicated chess computer is an engine playing on 6502, that's 1975 10um CPU, not playing on 2018 Zen+ 12nm CPU.
Milos---"SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. This is your claim! And is tested easily..... :lol:
Again to spell for retard who is obviously clueless, Elo is a relative measure that without a common base is a meaningless number. 1978 dedicated chess computer is at least 500 Elo weaker than SF-NNUE depth 1 totally irrespective of how SF absolute rating is calculated vs. humans or today's engines or 1970/80s chess computers. You again demonstrate that you are seriously challenged in basic comprehension.
That is why we run test. Who is better. And Stockfish at 1 ply is clearly rated much much worse then MPG from 1993, and much worse then Sargon from 1978. :lol:
You know that talking about oneself in plural is quite worrying indication of schizophrenia?
I was talking about testers, and we see again. You do not test, but just spill B.S. as facts. I know clearly that you do not test.
The only thing that is BS are your tests. Apart from being totally useless they are often fixed. You rename an engine with a fake name, or let it play on 1 core only while giving opponent all the cores, etc.
You sound desperate! :lol:

That is why I broadcast live, and give the engine download links, and testing conditions. I hide nothing....
Professing themselves to be wise, they became fools,
take on me. Foes 0.

Milos
Posts: 3971
Joined: Wed Nov 25, 2009 12:47 am

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by Milos » Fri Oct 16, 2020 12:04 am

mwyoung wrote:
Thu Oct 15, 2020 11:49 pm
Milos wrote:
Thu Oct 15, 2020 11:43 pm
mwyoung wrote:
Thu Oct 15, 2020 11:18 pm
Milos wrote:
Thu Oct 15, 2020 11:13 pm
mwyoung wrote:
Thu Oct 15, 2020 11:08 pm
Milos wrote:
Thu Oct 15, 2020 11:01 pm
mwyoung wrote:
Thu Oct 15, 2020 10:56 pm
Milos wrote:
Thu Oct 15, 2020 10:54 pm
mwyoung wrote:
Thu Oct 15, 2020 10:49 pm
Milos wrote:
Thu Oct 15, 2020 10:40 pm
mwyoung wrote:
Thu Oct 15, 2020 10:32 pm
Milos wrote:
Thu Oct 15, 2020 10:14 pm
lkaufman wrote:
Thu Oct 15, 2020 9:42 pm
mwyoung wrote:
Thu Oct 15, 2020 8:55 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
mclane wrote:
Tue Oct 13, 2020 10:06 am
I would like to see Stockfish run on a 6502 so we can try out how it plays against the old 8 bit dedicated chess computers.
SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess computers.
SF-NNUE on 6502 would have for sure more than 1000nodes per move. It wouldn't be any contest, it would be a slaughter.
You could not even engine test your way out of a paper bag. And your engine testing are as spot on as your conclusions. :lol:
As usual!

Sargon from 1978 vs Stockfish NNUE @ 1 ply (TC=5m+5s)

Live feed: https://www.youtube.com/watch?v=WZzsrGoxy7I
Sargon from 1978?? My recollection was that in 1978 the best chess computer one could buy ( excluding million dollar computers or the like) was no better than 1500 (probably much worse), and that was with 2 hours; at 5+5 should be like 1000 or worse. SF one ply shouldn't be that bad! What am I missing here? Chess computers one could buy weren't half-decent until early 1980s, unless my memory is quite wrong.
You are missing that a retard is playing an engine called Sargon on 2950x CPU and has 50kN/s. As usual all his tests are outright fake.
I know, and did you see the fake Mephisto Gideon professional by Ed Schröder crush Stockfish even worse....Ask Ed if MGP is also fake. :lol:

You can download the "fake" Sargon for UCI here. https://chessengines.blogspot.com/2020/ ... n-100.html
For quality trolling one needs some intelligence. You lack even the basic comprehension. Discussion that you hijacked was about running on 6502. You obviously don't even know what's that.
Milos---"SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess." computers.

Your conclusions, and memory, and your facts. :lol:
To spell for retard, dedicated chess computer is an engine playing on 6502, that's 1975 10um CPU, not playing on 2018 Zen+ 12nm CPU.
Milos---"SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. This is your claim! And is tested easily..... :lol:
Again to spell for retard who is obviously clueless, Elo is a relative measure that without a common base is a meaningless number. 1978 dedicated chess computer is at least 500 Elo weaker than SF-NNUE depth 1 totally irrespective of how SF absolute rating is calculated vs. humans or today's engines or 1970/80s chess computers. You again demonstrate that you are seriously challenged in basic comprehension.
That is why we run test. Who is better. And Stockfish at 1 ply is clearly rated much much worse then MPG from 1993, and much worse then Sargon from 1978. :lol:
You know that talking about oneself in plural is quite worrying indication of schizophrenia?
I was talking about testers, and we see again. You do not test, but just spill B.S. as facts. I know clearly that you do not test.
The only thing that is BS are your tests. Apart from being totally useless they are often fixed. You rename an engine with a fake name, or let it play on 1 core only while giving opponent all the cores, etc.
You sound desperate! :lol:

That is why I broadcast live, and give the engine download links, and testing conditions. I hide nothing....
And never give detailed pgn, yeah right, real boy-scout. :lol: :lol: :lol:

mwyoung
Posts: 2556
Joined: Wed May 12, 2010 8:00 pm

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by mwyoung » Fri Oct 16, 2020 12:06 am

Milos wrote:
Fri Oct 16, 2020 12:04 am
mwyoung wrote:
Thu Oct 15, 2020 11:49 pm
Milos wrote:
Thu Oct 15, 2020 11:43 pm
mwyoung wrote:
Thu Oct 15, 2020 11:18 pm
Milos wrote:
Thu Oct 15, 2020 11:13 pm
mwyoung wrote:
Thu Oct 15, 2020 11:08 pm
Milos wrote:
Thu Oct 15, 2020 11:01 pm
mwyoung wrote:
Thu Oct 15, 2020 10:56 pm
Milos wrote:
Thu Oct 15, 2020 10:54 pm
mwyoung wrote:
Thu Oct 15, 2020 10:49 pm
Milos wrote:
Thu Oct 15, 2020 10:40 pm
mwyoung wrote:
Thu Oct 15, 2020 10:32 pm
Milos wrote:
Thu Oct 15, 2020 10:14 pm
lkaufman wrote:
Thu Oct 15, 2020 9:42 pm
mwyoung wrote:
Thu Oct 15, 2020 8:55 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
mclane wrote:
Tue Oct 13, 2020 10:06 am
I would like to see Stockfish run on a 6502 so we can try out how it plays against the old 8 bit dedicated chess computers.
SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess computers.
SF-NNUE on 6502 would have for sure more than 1000nodes per move. It wouldn't be any contest, it would be a slaughter.
You could not even engine test your way out of a paper bag. And your engine testing are as spot on as your conclusions. :lol:
As usual!

Sargon from 1978 vs Stockfish NNUE @ 1 ply (TC=5m+5s)

Live feed: https://www.youtube.com/watch?v=WZzsrGoxy7I
Sargon from 1978?? My recollection was that in 1978 the best chess computer one could buy ( excluding million dollar computers or the like) was no better than 1500 (probably much worse), and that was with 2 hours; at 5+5 should be like 1000 or worse. SF one ply shouldn't be that bad! What am I missing here? Chess computers one could buy weren't half-decent until early 1980s, unless my memory is quite wrong.
You are missing that a retard is playing an engine called Sargon on 2950x CPU and has 50kN/s. As usual all his tests are outright fake.
I know, and did you see the fake Mephisto Gideon professional by Ed Schröder crush Stockfish even worse....Ask Ed if MGP is also fake. :lol:

You can download the "fake" Sargon for UCI here. https://chessengines.blogspot.com/2020/ ... n-100.html
For quality trolling one needs some intelligence. You lack even the basic comprehension. Discussion that you hijacked was about running on 6502. You obviously don't even know what's that.
Milos---"SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess." computers.

Your conclusions, and memory, and your facts. :lol:
To spell for retard, dedicated chess computer is an engine playing on 6502, that's 1975 10um CPU, not playing on 2018 Zen+ 12nm CPU.
Milos---"SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. This is your claim! And is tested easily..... :lol:
Again to spell for retard who is obviously clueless, Elo is a relative measure that without a common base is a meaningless number. 1978 dedicated chess computer is at least 500 Elo weaker than SF-NNUE depth 1 totally irrespective of how SF absolute rating is calculated vs. humans or today's engines or 1970/80s chess computers. You again demonstrate that you are seriously challenged in basic comprehension.
That is why we run test. Who is better. And Stockfish at 1 ply is clearly rated much much worse then MPG from 1993, and much worse then Sargon from 1978. :lol:
You know that talking about oneself in plural is quite worrying indication of schizophrenia?
I was talking about testers, and we see again. You do not test, but just spill B.S. as facts. I know clearly that you do not test.
The only thing that is BS are your tests. Apart from being totally useless they are often fixed. You rename an engine with a fake name, or let it play on 1 core only while giving opponent all the cores, etc.
You sound desperate! :lol:

That is why I broadcast live, and give the engine download links, and testing conditions. I hide nothing....
And never give detailed pgn, yeah right, real boy-scout. :lol: :lol: :lol:
I give games, and you can have these. If that will help with your desperation.... :lol:
Professing themselves to be wise, they became fools,
take on me. Foes 0.

Milos
Posts: 3971
Joined: Wed Nov 25, 2009 12:47 am

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by Milos » Fri Oct 16, 2020 12:15 am

mwyoung wrote:
Fri Oct 16, 2020 12:06 am
Milos wrote:
Fri Oct 16, 2020 12:04 am
mwyoung wrote:
Thu Oct 15, 2020 11:49 pm
Milos wrote:
Thu Oct 15, 2020 11:43 pm
mwyoung wrote:
Thu Oct 15, 2020 11:18 pm
Milos wrote:
Thu Oct 15, 2020 11:13 pm
mwyoung wrote:
Thu Oct 15, 2020 11:08 pm
Milos wrote:
Thu Oct 15, 2020 11:01 pm
mwyoung wrote:
Thu Oct 15, 2020 10:56 pm
Milos wrote:
Thu Oct 15, 2020 10:54 pm
mwyoung wrote:
Thu Oct 15, 2020 10:49 pm
Milos wrote:
Thu Oct 15, 2020 10:40 pm
mwyoung wrote:
Thu Oct 15, 2020 10:32 pm
Milos wrote:
Thu Oct 15, 2020 10:14 pm
lkaufman wrote:
Thu Oct 15, 2020 9:42 pm
mwyoung wrote:
Thu Oct 15, 2020 8:55 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
mclane wrote:
Tue Oct 13, 2020 10:06 am
I would like to see Stockfish run on a 6502 so we can try out how it plays against the old 8 bit dedicated chess computers.
SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess computers.
SF-NNUE on 6502 would have for sure more than 1000nodes per move. It wouldn't be any contest, it would be a slaughter.
You could not even engine test your way out of a paper bag. And your engine testing are as spot on as your conclusions. :lol:
As usual!

Sargon from 1978 vs Stockfish NNUE @ 1 ply (TC=5m+5s)

Live feed: https://www.youtube.com/watch?v=WZzsrGoxy7I
Sargon from 1978?? My recollection was that in 1978 the best chess computer one could buy ( excluding million dollar computers or the like) was no better than 1500 (probably much worse), and that was with 2 hours; at 5+5 should be like 1000 or worse. SF one ply shouldn't be that bad! What am I missing here? Chess computers one could buy weren't half-decent until early 1980s, unless my memory is quite wrong.
You are missing that a retard is playing an engine called Sargon on 2950x CPU and has 50kN/s. As usual all his tests are outright fake.
I know, and did you see the fake Mephisto Gideon professional by Ed Schröder crush Stockfish even worse....Ask Ed if MGP is also fake. :lol:

You can download the "fake" Sargon for UCI here. https://chessengines.blogspot.com/2020/ ... n-100.html
For quality trolling one needs some intelligence. You lack even the basic comprehension. Discussion that you hijacked was about running on 6502. You obviously don't even know what's that.
Milos---"SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess." computers.

Your conclusions, and memory, and your facts. :lol:
To spell for retard, dedicated chess computer is an engine playing on 6502, that's 1975 10um CPU, not playing on 2018 Zen+ 12nm CPU.
Milos---"SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. This is your claim! And is tested easily..... :lol:
Again to spell for retard who is obviously clueless, Elo is a relative measure that without a common base is a meaningless number. 1978 dedicated chess computer is at least 500 Elo weaker than SF-NNUE depth 1 totally irrespective of how SF absolute rating is calculated vs. humans or today's engines or 1970/80s chess computers. You again demonstrate that you are seriously challenged in basic comprehension.
That is why we run test. Who is better. And Stockfish at 1 ply is clearly rated much much worse then MPG from 1993, and much worse then Sargon from 1978. :lol:
You know that talking about oneself in plural is quite worrying indication of schizophrenia?
I was talking about testers, and we see again. You do not test, but just spill B.S. as facts. I know clearly that you do not test.
The only thing that is BS are your tests. Apart from being totally useless they are often fixed. You rename an engine with a fake name, or let it play on 1 core only while giving opponent all the cores, etc.
You sound desperate! :lol:

That is why I broadcast live, and give the engine download links, and testing conditions. I hide nothing....
And never give detailed pgn, yeah right, real boy-scout. :lol: :lol: :lol:
I give games, and you can have these. If that will help with your desperation.... :lol:
Yes, you give fake, useless games. You call it testing, others call it trolling.

mwyoung
Posts: 2556
Joined: Wed May 12, 2010 8:00 pm

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by mwyoung » Fri Oct 16, 2020 12:16 am

mwyoung wrote:
Thu Oct 15, 2020 8:55 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
mclane wrote:
Tue Oct 13, 2020 10:06 am
I would like to see Stockfish run on a 6502 so we can try out how it plays against the old 8 bit dedicated chess computers.
SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess computers.
SF-NNUE on 6502 would have for sure more than 1000nodes per move. It wouldn't be any contest, it would be a slaughter.
You could not even engine test your way out of a paper bag. And your engine testing are as spot on as your conclusions. :lol:
As usual!

Sargon from 1978 vs Stockfish NNUE @ 1 ply (TC=5m+5s)

Live feed: https://www.youtube.com/watch?v=WZzsrGoxy7I

End of match.

Code: Select all

Result:
-------------------------------------------------------------------------------
  #  name               games    wins   draws  losses   score    los%  elo+/-
  1. Sargon 1978 V1.00     50      34       4      12    36.0    99.9   164.1
  2. Stockfish 141020      50      12       4      34    14.0     0.1  -164.1

Cross table:
-------------------------------------------------------------------------------
  #  name                  score   games                                                  1                                                  2
  1. Sargon 1978 V1.00      36.0      50                                                  x 101111=1111101101011=10100=110101110111=1111011110
  2. Stockfish 141020       14.0      50 010000=0000010010100=01011=001010001000=0000100001                                                  x

Tech:
-------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  #  name                 nodes/m         NPS  depth/m   time/m    moves     time
  1. Sargon 1978 V1.00       226K       31344      5.8      7.2     38.1    274.1
  2. Stockfish 141020          9K      861534      1.0      0.0     38.4      0.4
     all ---                 114K       32576      3.4      3.6     38.2    137.3
Professing themselves to be wise, they became fools,
take on me. Foes 0.

mwyoung
Posts: 2556
Joined: Wed May 12, 2010 8:00 pm

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by mwyoung » Fri Oct 16, 2020 12:48 am

mwyoung wrote:
Fri Oct 16, 2020 12:16 am
mwyoung wrote:
Thu Oct 15, 2020 8:55 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
mclane wrote:
Tue Oct 13, 2020 10:06 am
I would like to see Stockfish run on a 6502 so we can try out how it plays against the old 8 bit dedicated chess computers.
SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess computers.
SF-NNUE on 6502 would have for sure more than 1000nodes per move. It wouldn't be any contest, it would be a slaughter.
You could not even engine test your way out of a paper bag. And your engine testing are as spot on as your conclusions. :lol:
As usual!

Sargon from 1978 vs Stockfish NNUE @ 1 ply (TC=5m+5s)

Live feed: https://www.youtube.com/watch?v=WZzsrGoxy7I

End of match.

Code: Select all

Result:
-------------------------------------------------------------------------------
  #  name               games    wins   draws  losses   score    los%  elo+/-
  1. Sargon 1978 V1.00     50      34       4      12    36.0    99.9   164.1
  2. Stockfish 141020      50      12       4      34    14.0     0.1  -164.1

Cross table:
-------------------------------------------------------------------------------
  #  name                  score   games                                                  1                                                  2
  1. Sargon 1978 V1.00      36.0      50                                                  x 101111=1111101101011=10100=110101110111=1111011110
  2. Stockfish 141020       14.0      50 010000=0000010010100=01011=001010001000=0000100001                                                  x

Tech:
-------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  #  name                 nodes/m         NPS  depth/m   time/m    moves     time
  1. Sargon 1978 V1.00       226K       31344      5.8      7.2     38.1    274.1
  2. Stockfish 141020          9K      861534      1.0      0.0     38.4      0.4
     all ---                 114K       32576      3.4      3.6     38.2    137.3
Here is the talkchess link about Sargon UCI. For those that think it is fake. Most questions are answered in the thread. Fun Match!

viewtopic.php?f=2&t=74027
Professing themselves to be wise, they became fools,
take on me. Foes 0.

Vinvin
Posts: 4782
Joined: Thu Mar 09, 2006 8:40 am
Full name: Vincent Lejeune

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by Vinvin » Fri Oct 16, 2020 1:54 am

mwyoung wrote:
Thu Oct 15, 2020 8:16 pm
lkaufman wrote:
Thu Oct 15, 2020 6:41 pm
mar wrote:
Wed Oct 14, 2020 10:03 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo.
That sounds unbelievable to me. Can you cite evidence for this 2400 figure? One ply is so tactically weak that I can't imagine how this could be even close to correct. I may run some tests myself.
Do not be distressed. Tactics still rule! :lol:

Mephisto Gideon Professional 2445 Elo is smashing Stockfish NNUE at 1 ply into little Stockfish NNUE meatballs.
The "ply" is not a standardized unit and so, is useless to compare engines.

mwyoung
Posts: 2556
Joined: Wed May 12, 2010 8:00 pm

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by mwyoung » Fri Oct 16, 2020 2:07 am

Vinvin wrote:
Fri Oct 16, 2020 1:54 am
mwyoung wrote:
Thu Oct 15, 2020 8:16 pm
lkaufman wrote:
Thu Oct 15, 2020 6:41 pm
mar wrote:
Wed Oct 14, 2020 10:03 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
SF-NNUE depth 1 which is SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo..
That sounds unbelievable to me. Can you cite evidence for this 2400 figure? One ply is so tactically weak that I can't imagine how this could be even close to correct. I may run some tests myself.
Do not be distressed. Tactics still rule! :lol:

Mephisto Gideon Professional 2445 Elo is smashing Stockfish NNUE at 1 ply into little Stockfish NNUE meatballs.
The "ply" is not a standardized unit and so, is useless to compare engines.
But it is by Milos. One Stockfish NNUE ply = 2400 Elo!

And Milos was even smart enough to give the node count. "SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo". :lol:
Professing themselves to be wise, they became fools,
take on me. Foes 0.

lkaufman
Posts: 4423
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by lkaufman » Fri Oct 16, 2020 4:01 am

mwyoung wrote:
Fri Oct 16, 2020 12:48 am
mwyoung wrote:
Fri Oct 16, 2020 12:16 am
mwyoung wrote:
Thu Oct 15, 2020 8:55 pm
Milos wrote:
Wed Oct 14, 2020 9:00 pm
mclane wrote:
Tue Oct 13, 2020 10:06 am
I would like to see Stockfish run on a 6502 so we can try out how it plays against the old 8 bit dedicated chess computers.
SF-NNUE depth 1 which is 500-1000 nodes is 2400Elo. That is at least 500 Elo above the best dedicated chess computers.
SF-NNUE on 6502 would have for sure more than 1000nodes per move. It wouldn't be any contest, it would be a slaughter.
You could not even engine test your way out of a paper bag. And your engine testing are as spot on as your conclusions. :lol:
As usual!

Sargon from 1978 vs Stockfish NNUE @ 1 ply (TC=5m+5s)

Live feed: https://www.youtube.com/watch?v=WZzsrGoxy7I

End of match.

Code: Select all

Result:
-------------------------------------------------------------------------------
  #  name               games    wins   draws  losses   score    los%  elo+/-
  1. Sargon 1978 V1.00     50      34       4      12    36.0    99.9   164.1
  2. Stockfish 141020      50      12       4      34    14.0     0.1  -164.1
  
       This is interesting to me, I had no idea such an old program as Sargon had been made to run on a current model computer. So I decided to determine what CCRL blitz rating Sargon would get. I happened to have a copy of program "Irina 0.15" on my 5 GHz laptop, which is rated 1435 by CCRL blitz. So I ran a 400 game match between Irina and Sargon at the CCRL blitz 2' + 1" level. Result: 48 wins for Sargon, 220 wins for Irina, 132 draws. This works out to 160 elo gap, which gives Sargon a predicted CCRL blitz rating of 1275 on an i7 computer. Presumably this means that on the computers it actually ran on forty years ago it would have a blitz rating somewhere near zero, which may or may not be accurate. Now it is important to note that a 1275 CCRL blitz rating does not really mean that a human rated 1275 would score 50% against it at 2' + 1", because humans play much worse at blitz relative to engines. My best guess from playing engines and looking at various human results is that it means that a human rated 1275 would score about 50% against it given something like 10x the time, i.e. 20' + 10" vs 2' + 1". So based on your result, the CCRL blitz rating for Stockfish NNUE 1 ply would be 1111. If my interpretation of the meaning of CCRL ratings is correct, it means that a human rated 1111 playing at a Rapid time control like 20' + 10" would be fairly matched with SFNNUE level one, at least if play is started from the terminal position of short opening books. I don't really believe this, but I don't see where I might be making a huge error. Perhaps some actual human players of average rating can play SF12 depth 1 at a casual pace like 20' + 10" and report your ratings and results. I'm sure the real performance will be between 1111 and 2400, but that's a huge range!
Komodo rules!

mwyoung
Posts: 2556
Joined: Wed May 12, 2010 8:00 pm

Re: Stockfish bench in i486 & Pentium 75mhz !

Post by mwyoung » Fri Oct 16, 2020 4:15 am

This is the post by LK. I took it out of the code window so it could be read.

LK - This is interesting to me, I had no idea such an old program as Sargon had been made to run on a current model computer. So I decided to determine what CCRL blitz rating Sargon would get. I happened to have a copy of program "Irina 0.15" on my 5 GHz laptop, which is rated 1435 by CCRL blitz. So I ran a 400 game match between Irina and Sargon at the CCRL blitz 2' + 1" level. Result: 48 wins for Sargon, 220 wins for Irina, 132 draws. This works out to 160 elo gap, which gives Sargon a predicted CCRL blitz rating of 1275 on an i7 computer. Presumably this means that on the computers it actually ran on forty years ago it would have a blitz rating somewhere near zero, which may or may not be accurate. Now it is important to note that a 1275 CCRL blitz rating does not really mean that a human rated 1275 would score 50% against it at 2' + 1", because humans play much worse at blitz relative to engines. My best guess from playing engines and looking at various human results is that it means that a human rated 1275 would score about 50% against it given something like 10x the time, i.e. 20' + 10" vs 2' + 1". So based on your result, the CCRL blitz rating for Stockfish NNUE 1 ply would be 1111. If my interpretation of the meaning of CCRL ratings is correct, it means that a human rated 1111 playing at a Rapid time control like 20' + 10" would be fairly matched with SFNNUE level one, at least if play is started from the terminal position of short opening books. I don't really believe this, but I don't see where I might be making a huge error. Perhaps some actual human players of average rating can play SF12 depth 1 at a casual pace like 20' + 10" and report your ratings and results. I'm sure the real performance will be between 1111 and 2400, but that's a huge range!
Professing themselves to be wise, they became fools,
take on me. Foes 0.

Post Reply