Ivanhoe plays singlecore in TCEC 13 !!

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

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Guenther
Posts: 2953
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by Guenther » Sat Aug 04, 2018 5:24 pm

Milos wrote:
Sat Aug 04, 2018 4:23 pm
CMCanavessi wrote:
Sat Aug 04, 2018 3:50 pm
Carlos777 wrote:
Sat Aug 04, 2018 3:49 pm
TCEC can fix this if they want, just correct Ivanhoe's settings and replay its games.
Again, TCEC settings are ok. It's the engine's fault.
Technically it is Kranium's fault. He provided the binary. And I am more and more convinced that all this is intentional on Anton's side. Sorry, but too many circumstantial evidence. They simply don't care to have properly working Ivanhoe. And the reason is because they want both NN "engines" to be in division 3. It is pretty obvious that with fully functioning Ivanhoe, there is close to 0% chance for both NN's to go through. It's a scam, period.
And still no information about Tucanos extremely low nps...
(in this case it could be a display error though, as the depths look high)

From current game, Tucano 7.05 is Black:

Code: Select all

Depth
move    white   black
1       NaN     NaN
2       NaN     NaN
3       24      26
4       26      29
5       26      29
6       25      28
7       27      28
8       25      26
9       27      27
10      27      28
11      28      26
12      28      26
13      29      27
14      25      27
15      26      25
16      24      25
17      26      25
18      26      25
19      26      27
20      24      27
21      24      25
22      25      32
23      25      30
24      26      27
25      22      29
26      24      26
27      23      28
28      24      25
29      25      29
30      28      30
31      31      30
32      30      28
33      29      29

Speed
move    white   black
1       NaN     NaN
2       NaN     NaN
3       10,018.929      1,054.255
4       10,198.404      1,234.929
5       10,364.662      1,179.049
6       10,192.882      1,318.178
7       10,451.718      1,321.748
8       10,083.177      1,306.382
9       10,520.03         303.82
10      10,585.231        529.738
11      10,810.212        281.362
12      10,935.495      1,044.053
13      11,445.326        642.428
14      11,145.498      1,029.782
15      10,992.486      1,215.486
16      11,122.792        951.436
17      11,101.299      1,246.276
18      11,037.72       1,212.287
19      11,190.085        980.622
20      11,077.496      1,177.331
21      11,064.12         323.457
22      10,930.517      1,176.836
23      11,319.215        754.579
24      11,392.174        180.042
25      11,574.41       1,150.731
26      11,409.278        437.179
27      10,893.732        491.258
28      11,212.323        554.163
29      11,804.625      1,186.502
30      11,916.579        691.296
31      12,440.173        733.955
32      12,569.352      1,246.682
33      13,155.351        796.561
Post by Joost Buijs » Sat Aug 04, 2018 6:31 am
Yesterday I was watching the game between Deus-X and Tucano and noticed that Tucano shows an incredibly low speed (around 1 mnps, sometimes even lower), so I suspect it was only using one core. Also the configured hash (1024 Mb) seems very low for a lazy-smp engine with these long thinking times. Maybe misconfiguration or is this normal behaviour for Tucano? I don't know.
Last edited by Guenther on Sat Aug 04, 2018 5:43 pm, edited 1 time in total.
Current foe list count : [91]
http://rwbc-chess.de/chronology.htm

User avatar
kranium
Posts: 1824
Joined: Thu May 29, 2008 8:43 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by kranium » Sat Aug 04, 2018 5:40 pm

jp wrote:
Sat Aug 04, 2018 4:41 pm
kranium wrote:
Sat Aug 04, 2018 1:30 pm
Laskos wrote:
Sat Aug 04, 2018 1:14 pm
kranium wrote:
Sat Aug 04, 2018 12:36 pm
personally I'm rooting for both to go thru
I am too rooting for them to go through, but based on merit. That one of the favorites is handicapped by 200-300 Elo points is surely not the best set up of the competition in this division.
Ivanhoe is a very strong engine, it would normally be placed in a much higher division.

If Ivanhoe is indeed playing on single core, which certainly appears to be possible..
that would actually bring it down to a suitable level for Division 4
That does not make it okay. How do you expect Ivanhoe to get to the much higher division where you claim it belongs if you cripple it so it loses in Division 4?
Ivanhoe hasn't played in TCEC in years
I couldn't care less if it advances...
with 2 days left to the start Anton was still short 2 entries...I simply tried to help him out with I consider to be a very interesting engine

Joost Buijs
Posts: 927
Joined: Thu Jul 16, 2009 8:47 am
Location: Almere, The Netherlands

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by Joost Buijs » Sat Aug 04, 2018 6:10 pm

Guenther wrote:
Sat Aug 04, 2018 5:24 pm
Milos wrote:
Sat Aug 04, 2018 4:23 pm
CMCanavessi wrote:
Sat Aug 04, 2018 3:50 pm
Carlos777 wrote:
Sat Aug 04, 2018 3:49 pm
TCEC can fix this if they want, just correct Ivanhoe's settings and replay its games.
Again, TCEC settings are ok. It's the engine's fault.
Technically it is Kranium's fault. He provided the binary. And I am more and more convinced that all this is intentional on Anton's side. Sorry, but too many circumstantial evidence. They simply don't care to have properly working Ivanhoe. And the reason is because they want both NN "engines" to be in division 3. It is pretty obvious that with fully functioning Ivanhoe, there is close to 0% chance for both NN's to go through. It's a scam, period.
And still no information about Tucanos extremely low nps...
(in this case it could be a display error though, as the depths look high)

From current game, Tucano 7.05 is Black:

Code: Select all

Depth
move    white   black
1       NaN     NaN
2       NaN     NaN
3       24      26
4       26      29
5       26      29
6       25      28
7       27      28
8       25      26
9       27      27
10      27      28
11      28      26
12      28      26
13      29      27
14      25      27
15      26      25
16      24      25
17      26      25
18      26      25
19      26      27
20      24      27
21      24      25
22      25      32
23      25      30
24      26      27
25      22      29
26      24      26
27      23      28
28      24      25
29      25      29
30      28      30
31      31      30
32      30      28
33      29      29

Speed
move    white   black
1       NaN     NaN
2       NaN     NaN
3       10,018.929      1,054.255
4       10,198.404      1,234.929
5       10,364.662      1,179.049
6       10,192.882      1,318.178
7       10,451.718      1,321.748
8       10,083.177      1,306.382
9       10,520.03         303.82
10      10,585.231        529.738
11      10,810.212        281.362
12      10,935.495      1,044.053
13      11,445.326        642.428
14      11,145.498      1,029.782
15      10,992.486      1,215.486
16      11,122.792        951.436
17      11,101.299      1,246.276
18      11,037.72       1,212.287
19      11,190.085        980.622
20      11,077.496      1,177.331
21      11,064.12         323.457
22      10,930.517      1,176.836
23      11,319.215        754.579
24      11,392.174        180.042
25      11,574.41       1,150.731
26      11,409.278        437.179
27      10,893.732        491.258
28      11,212.323        554.163
29      11,804.625      1,186.502
30      11,916.579        691.296
31      12,440.173        733.955
32      12,569.352      1,246.682
33      13,155.351        796.561
Post by Joost Buijs » Sat Aug 04, 2018 6:31 am
Yesterday I was watching the game between Deus-X and Tucano and noticed that Tucano shows an incredibly low speed (around 1 mnps, sometimes even lower), so I suspect it was only using one core. Also the configured hash (1024 Mb) seems very low for a lazy-smp engine with these long thinking times. Maybe misconfiguration or is this normal behaviour for Tucano? I don't know.
I just looked at the Tucano TCEC version and it does not support more than 1 GB hash, with regard to SMP it supports 8 threads max, with the command-line parameter -threads 8. On a single core I get ~2700 knps and when using 8 cores the reported speed remains the same but the CPU utilization goes to ~50% on my 6950X @ 3.8 GHz. So it shows the speed of the main thread only and the question remains whether it is running on 1 core or on 8 cores, configuring it to use 43 cores like they did in TCEC is wrong. The reported speed at TCEC is very low, that is probably due to the hardware it is running on, I believe the machine has 44 cores, but the clock-speed is probably not more than 2.2 GHz. or something like that, and it has ECC memory which also slows it down.

Milos
Posts: 3383
Joined: Wed Nov 25, 2009 12:47 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by Milos » Sat Aug 04, 2018 6:54 pm

Joost Buijs wrote:
Sat Aug 04, 2018 6:10 pm
On a single core I get ~2700 knps and when using 8 cores the reported speed remains the same but the CPU utilization goes to ~50% on my 6950X @ 3.8 GHz. So it shows the speed of the main thread only and the question remains whether it is running on 1 core or on 8 cores, configuring it to use 43 cores like they did in TCEC is wrong. The reported speed at TCEC is very low, that is probably due to the hardware it is running on, I believe the machine has 44 cores, but the clock-speed is probably not more than 2.2 GHz. or something like that, and it has ECC memory which also slows it down.
2699v4 used in TCEC has 2.8GHz all cores frequency and when using 4 cores per CPU they are running at 3.4GHz, which is only 12% slower than your 6950X.
ECC memory doesn't slow down anything, no idea where you that impression?

User avatar
CMCanavessi
Posts: 834
Joined: Thu Dec 28, 2017 3:06 pm
Location: Argentina

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by CMCanavessi » Sat Aug 04, 2018 7:28 pm

Joost Buijs wrote:
Sat Aug 04, 2018 6:10 pm
Guenther wrote:
Sat Aug 04, 2018 5:24 pm
Milos wrote:
Sat Aug 04, 2018 4:23 pm
CMCanavessi wrote:
Sat Aug 04, 2018 3:50 pm
Carlos777 wrote:
Sat Aug 04, 2018 3:49 pm
TCEC can fix this if they want, just correct Ivanhoe's settings and replay its games.
Again, TCEC settings are ok. It's the engine's fault.
Technically it is Kranium's fault. He provided the binary. And I am more and more convinced that all this is intentional on Anton's side. Sorry, but too many circumstantial evidence. They simply don't care to have properly working Ivanhoe. And the reason is because they want both NN "engines" to be in division 3. It is pretty obvious that with fully functioning Ivanhoe, there is close to 0% chance for both NN's to go through. It's a scam, period.
And still no information about Tucanos extremely low nps...
(in this case it could be a display error though, as the depths look high)

From current game, Tucano 7.05 is Black:

Code: Select all

Depth
move    white   black
1       NaN     NaN
2       NaN     NaN
3       24      26
4       26      29
5       26      29
6       25      28
7       27      28
8       25      26
9       27      27
10      27      28
11      28      26
12      28      26
13      29      27
14      25      27
15      26      25
16      24      25
17      26      25
18      26      25
19      26      27
20      24      27
21      24      25
22      25      32
23      25      30
24      26      27
25      22      29
26      24      26
27      23      28
28      24      25
29      25      29
30      28      30
31      31      30
32      30      28
33      29      29

Speed
move    white   black
1       NaN     NaN
2       NaN     NaN
3       10,018.929      1,054.255
4       10,198.404      1,234.929
5       10,364.662      1,179.049
6       10,192.882      1,318.178
7       10,451.718      1,321.748
8       10,083.177      1,306.382
9       10,520.03         303.82
10      10,585.231        529.738
11      10,810.212        281.362
12      10,935.495      1,044.053
13      11,445.326        642.428
14      11,145.498      1,029.782
15      10,992.486      1,215.486
16      11,122.792        951.436
17      11,101.299      1,246.276
18      11,037.72       1,212.287
19      11,190.085        980.622
20      11,077.496      1,177.331
21      11,064.12         323.457
22      10,930.517      1,176.836
23      11,319.215        754.579
24      11,392.174        180.042
25      11,574.41       1,150.731
26      11,409.278        437.179
27      10,893.732        491.258
28      11,212.323        554.163
29      11,804.625      1,186.502
30      11,916.579        691.296
31      12,440.173        733.955
32      12,569.352      1,246.682
33      13,155.351        796.561
Post by Joost Buijs » Sat Aug 04, 2018 6:31 am
Yesterday I was watching the game between Deus-X and Tucano and noticed that Tucano shows an incredibly low speed (around 1 mnps, sometimes even lower), so I suspect it was only using one core. Also the configured hash (1024 Mb) seems very low for a lazy-smp engine with these long thinking times. Maybe misconfiguration or is this normal behaviour for Tucano? I don't know.
I just looked at the Tucano TCEC version and it does not support more than 1 GB hash, with regard to SMP it supports 8 threads max, with the command-line parameter -threads 8. On a single core I get ~2700 knps and when using 8 cores the reported speed remains the same but the CPU utilization goes to ~50% on my 6950X @ 3.8 GHz. So it shows the speed of the main thread only and the question remains whether it is running on 1 core or on 8 cores, configuring it to use 43 cores like they did in TCEC is wrong. The reported speed at TCEC is very low, that is probably due to the hardware it is running on, I believe the machine has 44 cores, but the clock-speed is probably not more than 2.2 GHz. or something like that, and it has ECC memory which also slows it down.
True about hash size max being 1gb, but where did you get your 8 cores max thing? This is directly from the engine:

Code: Select all

protover
feature setboard=1
feature myname="tucano_7.05"
feature colors=0
feature analyze=1
feature option="Hash -spin 64 8 1024"
feature option="Threads -spin 1 1 64"
feature done=1
You can clearly see "Threads -spin 1 1 64" which means the highest value possible is 64. Wether Tucano actually uses all those or not is another matter, and yes, the reported speed always looks like from only 1 thread. Depth changes accordingly with more threads though.
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls

arunsoorya1309
Posts: 202
Joined: Tue Nov 03, 2015 3:55 pm

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by arunsoorya1309 » Sun Aug 05, 2018 1:48 am

If anyone has a better version of ivanhoe, can that not be submitted to TCEC?

Uri Blass
Posts: 8530
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by Uri Blass » Sun Aug 05, 2018 2:30 am

kranium wrote:
Sat Aug 04, 2018 5:40 pm
jp wrote:
Sat Aug 04, 2018 4:41 pm
kranium wrote:
Sat Aug 04, 2018 1:30 pm
Laskos wrote:
Sat Aug 04, 2018 1:14 pm
kranium wrote:
Sat Aug 04, 2018 12:36 pm
personally I'm rooting for both to go thru
I am too rooting for them to go through, but based on merit. That one of the favorites is handicapped by 200-300 Elo points is surely not the best set up of the competition in this division.
Ivanhoe is a very strong engine, it would normally be placed in a much higher division.

If Ivanhoe is indeed playing on single core, which certainly appears to be possible..
that would actually bring it down to a suitable level for Division 4
That does not make it okay. How do you expect Ivanhoe to get to the much higher division where you claim it belongs if you cripple it so it loses in Division 4?
Ivanhoe hasn't played in TCEC in years
I couldn't care less if it advances...
with 2 days left to the start Anton was still short 2 entries...I simply tried to help him out with I consider to be a very interesting engine
I agree and there is no reason to support it.

Based on wikipedia this engine is from 8.12.2011 and I do not know if people tested that it does not crash with 43 cores.

https://en.wikipedia.org/wiki/IPPOLIT

based on search with 1 core it seems to be slightly weaker than rybka4

best players based on CCRL(with 4 cores):

1)Stockfish premier division
2)Komodo premier division
3)Houdini premier division
4)Fire permier division
5)Shredder does not participate
6)Fizbo division 1
7)Andscacs premier division
8)Laser division 1
9)Booot division 1
10)Fritz division 1
11)Chiron division 1
12)Ethereal division 3
13)Nirvana division 2
14)Gull division 2
15)Xiphos division 2
16)Equinox does not participate
17)Critter does not participate
18)Hannibal division 3
19)Bouquet does not participate
20)Rybka does not participate and of course Fritz is better by the same author
21)Texel division 2
22)Protector does not participate
23)Alfil does not participate
24)chessbrain division 2
25)BlackMamba doea nor participate
26)Arasan division 3
27)Senpai division 4
28)Strelka does not particpate
29)Vajolet division 2
30)Nemorino division 3
31)Naum does not participate
32)Pedone division 3

best candidate to particpate based on CCRL rating could be:
1)Shredder
2)Equinox
3)Critter
4)Bouquet

User avatar
CMCanavessi
Posts: 834
Joined: Thu Dec 28, 2017 3:06 pm
Location: Argentina

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by CMCanavessi » Sun Aug 05, 2018 3:43 am

Uri Blass wrote:
Sun Aug 05, 2018 2:30 am
kranium wrote:
Sat Aug 04, 2018 5:40 pm
jp wrote:
Sat Aug 04, 2018 4:41 pm
kranium wrote:
Sat Aug 04, 2018 1:30 pm
Laskos wrote:
Sat Aug 04, 2018 1:14 pm
kranium wrote:
Sat Aug 04, 2018 12:36 pm
personally I'm rooting for both to go thru
I am too rooting for them to go through, but based on merit. That one of the favorites is handicapped by 200-300 Elo points is surely not the best set up of the competition in this division.
Ivanhoe is a very strong engine, it would normally be placed in a much higher division.

If Ivanhoe is indeed playing on single core, which certainly appears to be possible..
that would actually bring it down to a suitable level for Division 4
That does not make it okay. How do you expect Ivanhoe to get to the much higher division where you claim it belongs if you cripple it so it loses in Division 4?
Ivanhoe hasn't played in TCEC in years
I couldn't care less if it advances...
with 2 days left to the start Anton was still short 2 entries...I simply tried to help him out with I consider to be a very interesting engine
I agree and there is no reason to support it.

Based on wikipedia this engine is from 8.12.2011 and I do not know if people tested that it does not crash with 43 cores.

https://en.wikipedia.org/wiki/IPPOLIT

based on search with 1 core it seems to be slightly weaker than rybka4

best players based on CCRL(with 4 cores):

1)Stockfish premier division
2)Komodo premier division
3)Houdini premier division
4)Fire permier division
5)Shredder does not participate
6)Fizbo division 1
7)Andscacs premier division
8)Laser division 1
9)Booot division 1
10)Fritz division 1
11)Chiron division 1
12)Ethereal division 3
13)Nirvana division 2
14)Gull division 2
15)Xiphos division 2
16)Equinox does not participate
17)Critter does not participate
18)Hannibal division 3
19)Bouquet does not participate
20)Rybka does not participate and of course Fritz is better by the same author
21)Texel division 2
22)Protector does not participate
23)Alfil does not participate
24)chessbrain division 2
25)BlackMamba doea nor participate
26)Arasan division 3
27)Senpai division 4
28)Strelka does not particpate
29)Vajolet division 2
30)Nemorino division 3
31)Naum does not participate
32)Pedone division 3

best candidate to particpate based on CCRL rating could be:
1)Shredder
2)Equinox
3)Critter
4)Bouquet
Bouquet is another of the IvanHoe clones anyways :P
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls

jp
Posts: 701
Joined: Mon Apr 23, 2018 5:54 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by jp » Sun Aug 05, 2018 4:39 am

kranium wrote:
Sat Aug 04, 2018 5:40 pm
with 2 days left to the start Anton was still short 2 entries...I simply tried to help him out with I consider to be a very interesting engine
okay.

Joost Buijs
Posts: 927
Joined: Thu Jul 16, 2009 8:47 am
Location: Almere, The Netherlands

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by Joost Buijs » Sun Aug 05, 2018 5:02 am

Milos wrote:
Sat Aug 04, 2018 6:54 pm
Joost Buijs wrote:
Sat Aug 04, 2018 6:10 pm
On a single core I get ~2700 knps and when using 8 cores the reported speed remains the same but the CPU utilization goes to ~50% on my 6950X @ 3.8 GHz. So it shows the speed of the main thread only and the question remains whether it is running on 1 core or on 8 cores, configuring it to use 43 cores like they did in TCEC is wrong. The reported speed at TCEC is very low, that is probably due to the hardware it is running on, I believe the machine has 44 cores, but the clock-speed is probably not more than 2.2 GHz. or something like that, and it has ECC memory which also slows it down.
2699v4 used in TCEC has 2.8GHz all cores frequency and when using 4 cores per CPU they are running at 3.4GHz, which is only 12% slower than your 6950X.
ECC memory doesn't slow down anything, no idea where you that impression?
I have no clue what kind of machine or processor TCEC uses, I only know that it has 2x 22 core Xeon. When I compare the (single core) speed of Tucano TCEC on my machine ~2700 knps vs ~1200 knps on the TCEC machine, I get the impression that the clock-speed must be even lower than 2.2 GHz. Maybe it is throttling or whatever, dont ask me.

In the past I build several multi-processor machines with buffered ECC memory, the performance was always less compared to machines with CPU's of the same generation without ECC memory. So yes... in my experience ECC memory has a negative impact on performance.

Post Reply