Stockfish 12 dominating CCRL

[RogerC]

Stockfish 12 is dominating CCRL blitz (Computer Chess Rating Lists) : 3690 Elo points on 1 thread, 695 games played, 540 wins and just 2 losses. 215 Elo points ahead Komodo 14, and 245 Elo points ahead Houdini 6 on 1 thread.

With just 1 thread Stockfish is already the 1st against all engines in blitz, 50 Elo points ahead Lc0 0.25.1 t40-1541 with the RTX2080 graphics card... Wait and see the results on 4 threads !

http://ccrl.chessdom.com/ccrl/404/cgi/e ... _12_64-bit

http://ccrl.chessdom.com/ccrl/404/

[Modern Times]

It is also top of CEGT 40/20 list on just one thread.

[Laskos]

Stockfish 12 is dominating CCRL blitz (Computer Chess Rating Lists) : 3690 Elo points on 1 thread, 695 games played, 540 wins and just 2 losses. 215 Elo points ahead Komodo 14, and 245 Elo points ahead Houdini 6 on 1 thread.

With just 1 thread Stockfish is already the 1st against all engines in blitz, 50 Elo points ahead Lc0 0.25.1 t40-1541 with the RTX2080 graphics card... Wait and see the results on 4 threads !

http://ccrl.chessdom.com/ccrl/404/cgi/e ... _12_64-bit

http://ccrl.chessdom.com/ccrl/404/

Really, 1 threaded SF 12 is stronger than Lc0 SV 1541 on RTX2080? Maybe in rating lists, I guess head to head Lc0 on RTX2080 is still stronger than anything AB on only 1 thread.

[Raphexon]

Stockfish 12 is dominating CCRL blitz (Computer Chess Rating Lists) : 3690 Elo points on 1 thread, 695 games played, 540 wins and just 2 losses. 215 Elo points ahead Komodo 14, and 245 Elo points ahead Houdini 6 on 1 thread.

With just 1 thread Stockfish is already the 1st against all engines in blitz, 50 Elo points ahead Lc0 0.25.1 t40-1541 with the RTX2080 graphics card... Wait and see the results on 4 threads !

http://ccrl.chessdom.com/ccrl/404/cgi/e ... _12_64-bit

http://ccrl.chessdom.com/ccrl/404/

Really, 1 threaded SF 12 is stronger than Lc0 SV 1541 on RTX2080? Maybe in rating lists, I guess head to head Lc0 on RTX2080 is still stronger than anything AB on only 1 thread.

Taken from the Lc0 Discord (Source goes to Clementin)
2600 should be 2060 obviously.
One thread on a strong netbook CPU = roughly a 2060.

Round-Robin: SF NNue 1CPU, Lc0, Allie, CFish 11a 8CPU
Time control: 100s + 1s
Hardware: Ryzen 4900H, RTX 2600
Openings: TCEC SuFi 2016 (50 pos.) + 6-ply openings (50 pos.) as in Alphazero-Stockfish simulated rematch at CCC 2020/08
CFish 11a: Threads=8, Hash = 512, zen2_avx2
SF 170820: Threads=1, Hash=64, avx2_Ryzen (Chessman)
Allie 120820: MaxBatchSize=120, UseFP16=true, UseCustomWinograd=true
Lc0 26.2 SV-5300: backend-opts=custom_winograd=false, minibatch-size=120, max-prefetch=54, max-out-of-order-evals-factor=2, syzygy-fast-play=false
Lc0 26.1 J92-100: cuda 11 + cudnn 8.0.2 (Borg), minibatch-size=112, max-prefetch=70, max-out-of-order-evals-factor=2.4, moves-left-max-effect=0.1, moves-left-threshold=0.862976, moves-left-slope=0.007, syzygy-fast-play=false
Bench startpos (3.3s): SF 170820 NNUE = 1400 knps. CFish 11a 8CPU = 15000 knps. Lc0 J92-100 = 4.3 knps. Lc0 SV-5300 = 90 knps. Allie=50 knps.
TBs and adj.: Syzygy 5-men, resign movecount=5 score=600, draw movenumber=35 movecount=5 score=5
Comment: 1 CPU SF NNue on par with the 10x128 net here.
Games: https://gofile.io/d/FePTwM , Head-to-head statistics: https://gofile.io/d/goeYLw

Code: Select all

 # PLAYER                      :  RATING  ERROR  PLAYED    (%)   CFS    W    D    L   D(%)
   1 Lc0 26.1 + J92-100 MLH      :    22.4    9.0    1000  52.85    99  158  741  101  74.10
   2 SF 170820 + SV-1705 1CPU    :     5.4    8.9    1000  50.00    71  125  750  125  75.00
   3 Allie 120820 + LS 15.0      :     1.8    7.8    1000  49.40    50  128  732  140  73.20
   4 Lc0 26.2 + SV-5300          :     1.8    9.3    1000  49.40    55  113  762  125  76.20
   5 CFish 11a 8CPU              :     0.9    9.1    1000  49.25    55  143  699  158  69.90
   6 SF 170820 + SV-2257 1CPU    :     0.0    9.0    1000  49.10   ---  120  742  138  74.20

White advantage = 49.09 +/- 3.10
Draw rate (equal opponents) = 77.00 % +/- 0.80

[Laskos]

Stockfish 12 is dominating CCRL blitz (Computer Chess Rating Lists) : 3690 Elo points on 1 thread, 695 games played, 540 wins and just 2 losses. 215 Elo points ahead Komodo 14, and 245 Elo points ahead Houdini 6 on 1 thread.

With just 1 thread Stockfish is already the 1st against all engines in blitz, 50 Elo points ahead Lc0 0.25.1 t40-1541 with the RTX2080 graphics card... Wait and see the results on 4 threads !

http://ccrl.chessdom.com/ccrl/404/cgi/e ... _12_64-bit

http://ccrl.chessdom.com/ccrl/404/

Really, 1 threaded SF 12 is stronger than Lc0 SV 1541 on RTX2080? Maybe in rating lists, I guess head to head Lc0 on RTX2080 is still stronger than anything AB on only 1 thread.

Taken from the Lc0 Discord (Source goes to Clementin)
2600 should be 2060 obviously.
One thread on a strong netbook CPU = roughly a 2060.

Round-Robin: SF NNue 1CPU, Lc0, Allie, CFish 11a 8CPU
Time control: 100s + 1s
Hardware: Ryzen 4900H, RTX 2600
Openings: TCEC SuFi 2016 (50 pos.) + 6-ply openings (50 pos.) as in Alphazero-Stockfish simulated rematch at CCC 2020/08
CFish 11a: Threads=8, Hash = 512, zen2_avx2
SF 170820: Threads=1, Hash=64, avx2_Ryzen (Chessman)
Allie 120820: MaxBatchSize=120, UseFP16=true, UseCustomWinograd=true
Lc0 26.2 SV-5300: backend-opts=custom_winograd=false, minibatch-size=120, max-prefetch=54, max-out-of-order-evals-factor=2, syzygy-fast-play=false
Lc0 26.1 J92-100: cuda 11 + cudnn 8.0.2 (Borg), minibatch-size=112, max-prefetch=70, max-out-of-order-evals-factor=2.4, moves-left-max-effect=0.1, moves-left-threshold=0.862976, moves-left-slope=0.007, syzygy-fast-play=false
Bench startpos (3.3s): SF 170820 NNUE = 1400 knps. CFish 11a 8CPU = 15000 knps. Lc0 J92-100 = 4.3 knps. Lc0 SV-5300 = 90 knps. Allie=50 knps.
TBs and adj.: Syzygy 5-men, resign movecount=5 score=600, draw movenumber=35 movecount=5 score=5
Comment: 1 CPU SF NNue on par with the 10x128 net here.
Games: https://gofile.io/d/FePTwM , Head-to-head statistics: https://gofile.io/d/goeYLw

Code: Select all

 # PLAYER                      :  RATING  ERROR  PLAYED    (%)   CFS    W    D    L   D(%)
   1 Lc0 26.1 + J92-100 MLH      :    22.4    9.0    1000  52.85    99  158  741  101  74.10
   2 SF 170820 + SV-1705 1CPU    :     5.4    8.9    1000  50.00    71  125  750  125  75.00
   3 Allie 120820 + LS 15.0      :     1.8    7.8    1000  49.40    50  128  732  140  73.20
   4 Lc0 26.2 + SV-5300          :     1.8    9.3    1000  49.40    55  113  762  125  76.20
   5 CFish 11a 8CPU              :     0.9    9.1    1000  49.25    55  143  699  158  69.90
   6 SF 170820 + SV-2257 1CPU    :     0.0    9.0    1000  49.10   ---  120  742  138  74.20

White advantage = 49.09 +/- 3.10
Draw rate (equal opponents) = 77.00 % +/- 0.80

Thanks, so it seems RTX2060 is a good match for 1 fast thread. Still, it's close, RTX2080 is not even 2 times faster. CCRL seems to show 50 Elo advantage of SF 12 over Lc0 1541 2080, which I guess are either $$$ or an artifact of the rating list with many much weaker opponents.

[Jouni]

What the point to play SF12 with one core against 2060 and 1920 cores? Fair ?

[RogerC]

Stockfish 12 is dominating CCRL blitz (Computer Chess Rating Lists) : 3690 Elo points on 1 thread, 695 games played, 540 wins and just 2 losses. 215 Elo points ahead Komodo 14, and 245 Elo points ahead Houdini 6 on 1 thread.

With just 1 thread Stockfish is already the 1st against all engines in blitz, 50 Elo points ahead Lc0 0.25.1 t40-1541 with the RTX2080 graphics card... Wait and see the results on 4 threads !

http://ccrl.chessdom.com/ccrl/404/cgi/e ... _12_64-bit

http://ccrl.chessdom.com/ccrl/404/

Code: Select all

 # PLAYER                      :  RATING  ERROR  PLAYED    (%)   CFS    W    D    L   D(%)
   1 Lc0 26.1 + J92-100 MLH      :    22.4    9.0    1000  52.85    99  158  741  101  74.10
   2 SF 170820 + SV-1705 1CPU    :     5.4    8.9    1000  50.00    71  125  750  125  75.00
   3 Allie 120820 + LS 15.0      :     1.8    7.8    1000  49.40    50  128  732  140  73.20
   4 Lc0 26.2 + SV-5300          :     1.8    9.3    1000  49.40    55  113  762  125  76.20
   5 CFish 11a 8CPU              :     0.9    9.1    1000  49.25    55  143  699  158  69.90
   6 SF 170820 + SV-2257 1CPU    :     0.0    9.0    1000  49.10   ---  120  742  138  74.20

White advantage = 49.09 +/- 3.10
Draw rate (equal opponents) = 77.00 % +/- 0.80

SF170820 IS NOT SF12 !

[Raphexon]

Stockfish 12 is dominating CCRL blitz (Computer Chess Rating Lists) : 3690 Elo points on 1 thread, 695 games played, 540 wins and just 2 losses. 215 Elo points ahead Komodo 14, and 245 Elo points ahead Houdini 6 on 1 thread.

With just 1 thread Stockfish is already the 1st against all engines in blitz, 50 Elo points ahead Lc0 0.25.1 t40-1541 with the RTX2080 graphics card... Wait and see the results on 4 threads !

http://ccrl.chessdom.com/ccrl/404/cgi/e ... _12_64-bit

http://ccrl.chessdom.com/ccrl/404/

Code: Select all

 # PLAYER                      :  RATING  ERROR  PLAYED    (%)   CFS    W    D    L   D(%)
   1 Lc0 26.1 + J92-100 MLH      :    22.4    9.0    1000  52.85    99  158  741  101  74.10
   2 SF 170820 + SV-1705 1CPU    :     5.4    8.9    1000  50.00    71  125  750  125  75.00
   3 Allie 120820 + LS 15.0      :     1.8    7.8    1000  49.40    50  128  732  140  73.20
   4 Lc0 26.2 + SV-5300          :     1.8    9.3    1000  49.40    55  113  762  125  76.20
   5 CFish 11a 8CPU              :     0.9    9.1    1000  49.25    55  143  699  158  69.90
   6 SF 170820 + SV-2257 1CPU    :     0.0    9.0    1000  49.10   ---  120  742  138  74.20

White advantage = 49.09 +/- 3.10
Draw rate (equal opponents) = 77.00 % +/- 0.80

SF170820 IS NOT SF12 !

It's close enough to demonstrate a point.

[Modern Times]

Thanks, so it seems RTX2060 is a good match for 1 fast thread. Still, it's close, RTX2080 is not even 2 times faster. CCRL seems to show 50 Elo advantage of SF 12 over Lc0 1541 2080, which I guess are either $$$ or an artifact of the rating list with many much weaker opponents.

SF12 has not actually played any of the RTX2080 engines head-to-head (different tester has the RTX) so we need to take care with that initial rating. I think things could change a little once those games happen.

[RogerC]

Thanks, so it seems RTX2060 is a good match for 1 fast thread. Still, it's close, RTX2080 is not even 2 times faster. CCRL seems to show 50 Elo advantage of SF 12 over Lc0 1541 2080, which I guess are either $$$ or an artifact of the rating list with many much weaker opponents.

SF12 has not actually played any of the RTX2080 engines head-to-head (different tester has the RTX) so we need to take care with that initial rating. I think things could change a little once those games happen.

As i said, Stockfish 12 is dominating CCRL , but CEGT also !
http://www.cegt.net/40_40%20Rating%20Li ... liste.html
Here are the results of the matchs between SF12 and opponents : SF12 (1 CPU) vs LCZero 0.25.1 Cuda (LS15.0) : 100 games , 11 wins, 86 draws, 3 losses.
http://www.cegt.net/40_40%20Rating%20Li ... ons/1.html

SF12 on 8 cores will crush LC0... LC0 is the past, and is finished because it needs more and more Graphical power (2000+ Cores) to compute and evaluate half dozen kn/s. On the other hand SF do an intelligent mix of classical and NNUE evaluations on just a single CPU Core. And finally, NNUE nets are just at the very beginning and early of there trainings (2 month old). See you in 12 months !