Vitruvius 1.0C v Fritz 13- Seeing Is Believing

[geots]

Vitruvius_1.0C_x32 vs. Fritz 13


Knowing that strange things can happen, and actually seeing them happen are 2 entirely different things. Naturally the name "Fritz" doesn't conjure up the images of 8 or 9 years ago, but you gotta respect any engine that is in the Top 10 32bit list. Go figure.



1CPU/32bit
128MB hash
Bases=NONE (Fritz 13 used Nalimov)
Ponder_Learning=OFF
Perfect 12.32 book w/12-move limit
40/3 Repeating
Match=50 games


Vitruvius_1.0C_x32......+173........+26/-3/=21......73.00%......36.5/50

Fritz 13...........................-173.........+3/-26/=21......27.00%......13.5/50



By the bye-

g
e
o
r
g
e

[gleperlier]

Hi George,

Did you notice also that Vitrivius is using more CPU than alocated ?

I have a 4 Cores (8 threads CPU). When I set Stockfish to play vs Critter and I give them 4 threads each, I can see both of them taking 50% total CPU.

When I am doing the same but Vitrivius vs Stockfish or Vitrivius vs Critter, I can see Vitrivius taking 66 to 70% total CPU and so only average 33% for Stockfish or Critter.

I realised that after some match were Vitrivius was winning to easy...

What is your idea ? As anyone noticed the same ?

Best wishes

Gabriel

[Don]

Hi George,

Did you notice also that Vitrivius is using more CPU than alocated ?

I have a 4 Cores (8 threads CPU). When I set Stockfish to play vs Critter and I give them 4 threads each, I can see both of them taking 50% total CPU.

When I am doing the same but Vitrivius vs Stockfish or Vitrivius vs Critter, I can see Vitrivius taking 66 to 70% total CPU and so only average 33% for Stockfish or Critter.

I realised that after some match were Vitrivius was winning to easy...

What is your idea ? As anyone noticed the same ?

Best wishes

Gabriel

To be sure you can test machines on separate boxes. I think that is really the best way to test MP programs. Of course you need to machines that are exactly the same - or if they are close you can alternate machines.

[geots]

Hi George,

Did you notice also that Vitrivius is using more CPU than alocated ?

I have a 4 Cores (8 threads CPU). When I set Stockfish to play vs Critter and I give them 4 threads each, I can see both of them taking 50% total CPU.

When I am doing the same but Vitrivius vs Stockfish or Vitrivius vs Critter, I can see Vitrivius taking 66 to 70% total CPU and so only average 33% for Stockfish or Critter.

I realised that after some match were Vitrivius was winning to easy...

What is your idea ? As anyone noticed the same ?

Best wishes

Gabriel

I am not sure my comparison would help you, because all my 40/3 repeating matches were run on 1 CPU for each engine. And I check task manager at the beginning, 3 or 4 times during the match, and in games 47-50, etc. And I can assure you that Vitruvius took no cpu time away from Fritz 13. That score is completely fair. Fritz just had no chance. I have been busy with this testing, and so much other stuff going on, I have not even had time yet to get my new 64bit quad loaded and running. Without that, I can't give them more than 1 core. I check a lot because sometimes this mcshield exe, something to do with McAfee I guess, will try to use cpu time, taking 10 to 15% of the thinking for each engine. I have to reboot to stop it. But it has been a couple months since it happened last.

Am I wrong that you are giving them 4 each and running ponder ON? Because with ponder off I don't see how he could be using any cpu time. At any rate, since it would be impossible for me to give them more than 1 cpu on this box, I cannot recreate anything you refer to. I think it would be a good idea for you to get in touch with Roberto on this issue. And I am sorry you are having a problem. Maybe you could get another tester to try to simulate it, tho I know a lot of people don't have 8 cores available.


Best regards,

george

[Don]

Hi George,

Did you notice also that Vitrivius is using more CPU than alocated ?

I have a 4 Cores (8 threads CPU). When I set Stockfish to play vs Critter and I give them 4 threads each, I can see both of them taking 50% total CPU.

When I am doing the same but Vitrivius vs Stockfish or Vitrivius vs Critter, I can see Vitrivius taking 66 to 70% total CPU and so only average 33% for Stockfish or Critter.

I realised that after some match were Vitrivius was winning to easy...

What is your idea ? As anyone noticed the same ?

Best wishes

Gabriel

I am not sure my comparison would help you, because all my 40/3 repeating matches were run on 1 CPU for each engine. And I check task manager at the beginning, 3 or 4 times during the match, and in games 47-50, etc. And I can assure you that Vitruvius took no cpu time away from Fritz 13. That score is completely fair. Fritz just had no chance. I have been busy with this testing, and so much other stuff going on, I have not even had time yet to get my new 64bit quad loaded and running. Without that, I can't give them more than 1 core. I check a lot because sometimes this mcshield exe, something to do with McAfee I guess, will try to use cpu time, taking 10 to 15% of the thinking for each engine. I have to reboot to stop it. But it has been a couple months since it happened last.

Am I wrong that you are giving them 4 each and running ponder ON? Because with ponder off I don't see how he could be using any cpu time. At any rate, since it would be impossible for me to give them more than 1 cpu on this box, I cannot recreate anything you refer to. I think it would be a good idea for you to get in touch with Roberto on this issue. And I am sorry you are having a problem. Maybe you could get another tester to try to simulate it, tho I know a lot of people don't have 8 cores available.


Best regards,

george

A program can still cheat even on 1 processor. I'm not saying that is happening, but with ponder off an improperly written program could consume CPU time during it's thinking cycles which would affect your results. On a single CPU system that would be a serious advantage for a program to do this. If I can find the program I will check it on my linux system. It could be different on Linux as the I/O is somewhat different.

You can probably check task manager to see if anything fishy is going on.

[geots]

Hi George,

Did you notice also that Vitrivius is using more CPU than alocated ?

I have a 4 Cores (8 threads CPU). When I set Stockfish to play vs Critter and I give them 4 threads each, I can see both of them taking 50% total CPU.

When I am doing the same but Vitrivius vs Stockfish or Vitrivius vs Critter, I can see Vitrivius taking 66 to 70% total CPU and so only average 33% for Stockfish or Critter.

I realised that after some match were Vitrivius was winning to easy...

What is your idea ? As anyone noticed the same ?

Best wishes

Gabriel

I am not sure my comparison would help you, because all my 40/3 repeating matches were run on 1 CPU for each engine. And I check task manager at the beginning, 3 or 4 times during the match, and in games 47-50, etc. And I can assure you that Vitruvius took no cpu time away from Fritz 13. That score is completely fair. Fritz just had no chance. I have been busy with this testing, and so much other stuff going on, I have not even had time yet to get my new 64bit quad loaded and running. Without that, I can't give them more than 1 core. I check a lot because sometimes this mcshield exe, something to do with McAfee I guess, will try to use cpu time, taking 10 to 15% of the thinking for each engine. I have to reboot to stop it. But it has been a couple months since it happened last.

Am I wrong that you are giving them 4 each and running ponder ON? Because with ponder off I don't see how he could be using any cpu time. At any rate, since it would be impossible for me to give them more than 1 cpu on this box, I cannot recreate anything you refer to. I think it would be a good idea for you to get in touch with Roberto on this issue. And I am sorry you are having a problem. Maybe you could get another tester to try to simulate it, tho I know a lot of people don't have 8 cores available.


Best regards,

george

A program can still cheat even on 1 processor. I'm not saying that is happening, but with ponder off an improperly written program could consume CPU time during it's thinking cycles which would affect your results. On a single CPU system that would be a serious advantage for a program to do this. If I can find the program I will check it on my linux system. It could be different on Linux as the I/O is somewhat different.

You can probably check task manager to see if anything fishy is going on.

The only thing is that for an engine to be doing that on 1cpu with ponder off, he would have to be doing it without it showing in task manager. Hiding the fact, so to speak. Because I check so often, and at times I will leave task mgr. on the screen for a short time, glancing at it fairly regularly. Unless a lot of other programs were running, the user will show up to 99 in usage, that's as high as it will ever show- doesn't go to 100. So with ponder OFF, if one is showing 99 in usage, it is impossible for the other to be showing anything anyway. And believe me, if one engine is showing consistently in the low 90's I will be scrolling to find out which process is getting the other 7 or 8. I admit checking like I do is not the norm, and I don't always, but never a match goes by without checking at all. I have had problems with a couple of exes in the past, and it has just made me gunshy. (I wish I had the money to run 4 cores each with ponder ON.)

george

[Don]

Hi George,

Did you notice also that Vitrivius is using more CPU than alocated ?

I have a 4 Cores (8 threads CPU). When I set Stockfish to play vs Critter and I give them 4 threads each, I can see both of them taking 50% total CPU.

When I am doing the same but Vitrivius vs Stockfish or Vitrivius vs Critter, I can see Vitrivius taking 66 to 70% total CPU and so only average 33% for Stockfish or Critter.

I realised that after some match were Vitrivius was winning to easy...

What is your idea ? As anyone noticed the same ?

Best wishes

Gabriel

I am not sure my comparison would help you, because all my 40/3 repeating matches were run on 1 CPU for each engine. And I check task manager at the beginning, 3 or 4 times during the match, and in games 47-50, etc. And I can assure you that Vitruvius took no cpu time away from Fritz 13. That score is completely fair. Fritz just had no chance. I have been busy with this testing, and so much other stuff going on, I have not even had time yet to get my new 64bit quad loaded and running. Without that, I can't give them more than 1 core. I check a lot because sometimes this mcshield exe, something to do with McAfee I guess, will try to use cpu time, taking 10 to 15% of the thinking for each engine. I have to reboot to stop it. But it has been a couple months since it happened last.

Am I wrong that you are giving them 4 each and running ponder ON? Because with ponder off I don't see how he could be using any cpu time. At any rate, since it would be impossible for me to give them more than 1 cpu on this box, I cannot recreate anything you refer to. I think it would be a good idea for you to get in touch with Roberto on this issue. And I am sorry you are having a problem. Maybe you could get another tester to try to simulate it, tho I know a lot of people don't have 8 cores available.


Best regards,

george

A program can still cheat even on 1 processor. I'm not saying that is happening, but with ponder off an improperly written program could consume CPU time during it's thinking cycles which would affect your results. On a single CPU system that would be a serious advantage for a program to do this. If I can find the program I will check it on my linux system. It could be different on Linux as the I/O is somewhat different.

You can probably check task manager to see if anything fishy is going on.

The only thing is that for an engine to be doing that on 1cpu with ponder off, he would have to be doing it without it showing in task manager. Hiding the fact, so to speak. Because I check so often, and at times I will leave task mgr. on the screen for a short time, glancing at it fairly regularly. Unless a lot of other programs were running, the user will show up to 99 in usage, that's as high as it will ever show- doesn't go to 100. So with ponder OFF, if one is showing 99 in usage, it is impossible for the other to be showing anything anyway. And believe me, if one engine is showing consistently in the low 90's I will be scrolling to find out which process is getting the other 7 or 8. I admit checking like I do is not the norm, and I don't always, but never a match goes by without checking at all. I have had problems with a couple of exes in the past, and it has just made me gunshy.

george

Well then it seems like it's probably ok.

It may be an MP issue to check when you get your quad working.

[geots]

Hi George,

Did you notice also that Vitrivius is using more CPU than alocated ?

I have a 4 Cores (8 threads CPU). When I set Stockfish to play vs Critter and I give them 4 threads each, I can see both of them taking 50% total CPU.

When I am doing the same but Vitrivius vs Stockfish or Vitrivius vs Critter, I can see Vitrivius taking 66 to 70% total CPU and so only average 33% for Stockfish or Critter.

I realised that after some match were Vitrivius was winning to easy...

What is your idea ? As anyone noticed the same ?

Best wishes

Gabriel

I am not sure my comparison would help you, because all my 40/3 repeating matches were run on 1 CPU for each engine. And I check task manager at the beginning, 3 or 4 times during the match, and in games 47-50, etc. And I can assure you that Vitruvius took no cpu time away from Fritz 13. That score is completely fair. Fritz just had no chance. I have been busy with this testing, and so much other stuff going on, I have not even had time yet to get my new 64bit quad loaded and running. Without that, I can't give them more than 1 core. I check a lot because sometimes this mcshield exe, something to do with McAfee I guess, will try to use cpu time, taking 10 to 15% of the thinking for each engine. I have to reboot to stop it. But it has been a couple months since it happened last.

Am I wrong that you are giving them 4 each and running ponder ON? Because with ponder off I don't see how he could be using any cpu time. At any rate, since it would be impossible for me to give them more than 1 cpu on this box, I cannot recreate anything you refer to. I think it would be a good idea for you to get in touch with Roberto on this issue. And I am sorry you are having a problem. Maybe you could get another tester to try to simulate it, tho I know a lot of people don't have 8 cores available.


Best regards,

george

A program can still cheat even on 1 processor. I'm not saying that is happening, but with ponder off an improperly written program could consume CPU time during it's thinking cycles which would affect your results. On a single CPU system that would be a serious advantage for a program to do this. If I can find the program I will check it on my linux system. It could be different on Linux as the I/O is somewhat different.

You can probably check task manager to see if anything fishy is going on.

The only thing is that for an engine to be doing that on 1cpu with ponder off, he would have to be doing it without it showing in task manager. Hiding the fact, so to speak. Because I check so often, and at times I will leave task mgr. on the screen for a short time, glancing at it fairly regularly. Unless a lot of other programs were running, the user will show up to 99 in usage, that's as high as it will ever show- doesn't go to 100. So with ponder OFF, if one is showing 99 in usage, it is impossible for the other to be showing anything anyway. And believe me, if one engine is showing consistently in the low 90's I will be scrolling to find out which process is getting the other 7 or 8. I admit checking like I do is not the norm, and I don't always, but never a match goes by without checking at all. I have had problems with a couple of exes in the past, and it has just made me gunshy.

george

Well then it seems like it's probably ok.

It may be an MP issue to check when you get your quad working.

I had just been talking to Roberto when I came over here, so after I answered Gabriel I sent a mail to Roberto telling him what I was told. So he should be checking it out if he saw my last mail. I would think any number of people could check it easily, because from what I have seen, it seems like everyone who has it has at minimum a quad.


george

[Carotino]

I never had this problem ... Although I possess only a small lap-top dual-core.
In theory, with ponder = off, should not there be absolutely, no" theft" of time...
In any case, I will run tests from a friend of mine who owns a Quad. If there is a problem, I will try to resolve it immediately.

Roberto.

P.S. After a while, I will release an updated version of Vitruvius, contains all the bug-fixes and enhancements found so far...