Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / New (w/ NNUE) Stockfish on an old (x64) CPU
- - By MrKris (***) Date 2021-02-16 22:30
This worked out very well:
abrok Stockfish vs. the the same set to no NNUE (Classical).

Both plain "x64" versions, no avx2 nor any other advanced instructions that would have made NNUE faster.
For AMD PhIIX6 (6 'real cores', it has no HT, popcnt could be added, alter the Makefile, but I used abrok exe.)

Speed comparison, initial position, 6 cores, 60 secs., 1GBh, Sy.6
Stockfish_21021123_x64 defaults: 2,409knps 35/46 144,514kn
Stockfish_21021123_x64  no NNUE: 5,089knps 32/45 305,371kn

4kbnr/4pppp/8/8/8/8/4PPPP/4KBNR w Kk - 0 1

Stockfish_21021123_x64 defaults: 2,547knps 37/40 152,796kn
Stockfish_21021123_x64  no NNUE: 6,384knps 40/53 383,026kn

_ _ _ _ _ _

Test:
AMD PhIIX6 ; 1-core each @ G/60"+1" (concurrency=6)
128 MBh , Syzygy-6 , UHO 4-moves colors-reversed
                                   Games     Score    Draws
1    Stockfish_21021123_x64  :  426.5 / 600  71.1%    38.8%
2 Sf-Classical_21021123_x64  :  173.5 / 600  28.9%    38.8%


Stockfish_21021123_x64 vs Sf-Classical_21021123_x64:
        +310  -57  =233 , Elo +156 +-22

The shortest win, then Stockfish drew as black.
[Event "G/60''+1'' 128MBh Sy.6"]
[Site "AMD PhIIX6 Sy.6 6x1core"]
[Date "2021.02.16"]
[Round "176"]
[White "Stockfish_21021123_x64"]
[Black "Sf-Classical_21021123_x64"]
[Result "1-0"]
[ECO "B01"]
[Opening "Scandinavian"]
[Variation "Schiller Defence"]
[TimeControl "60+1"]
[Termination "adjudication"]
[PlyCount "50"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} d5 {book} 2. exd5 {book} Qxd5 {book} 3. Nc3 {book} Qd6 {book}
4. d4 {book} a6 {book} 5. Nf3 {+1.23/19 2.6s}
( {>>> Sf-Cl 1/2-1/2 Sf} 5. Bd3 Nf6 6. Nf3 g6 7. O-O Bg7
8. Re1 O-O 9. Ne4 Nxe4 10. Rxe4 Bf5 11. Bf4 Qd7 12. Re3 c5 13. dxc5 Bxb2
14. Rb1 Bf6 15. Ne5 Bxe5 16. Rxe5 Bxd3 17. Qxd3 Qxd3 18. cxd3 Nc6 19. Rd5
Rfd8 20. Rxd8+ Nxd8 21. d4 Rc8 22. Be3 Rc7 23. d5 Rd7 24. Rd1 Ne6 25. Rd2
Nc7 26. d6 Ne6 27. dxe7 Rxe7 28. g4 f5 29. gxf5 gxf5 30. Rd6 f4 31. Bd2
Nxc5 32. Bxf4 Rd7 33. Rb6 Na4 34. Rb3 Kf7 {1/2-1/2} )
5... Bf5 {-0.75/24 6.7s} 6. g3
{+1.23/19 1.8s} Nc6 {-0.59/18 0.69s} 7. Bg2 {+1.09/22 4.1s} e6 {-1.01/22
3.2s} 8. O-O {+1.58/21 3.4s} O-O-O {-0.87/21 1.8s} 9. Qe2 {+1.85/18 0.78s}
Nf6 {-0.86/20 1.1s} 10. Rd1 {+1.89/21 1.5s} h5 {-0.90/26 15s} 11. h3
{+1.69/19 1.1s} Qe7 {-0.71/20 0.89s} 12. a3 {+2.68/18 1.2s} h4 {-0.75/21
2.3s} 13. g4 {+2.71/23 3.3s} Nxg4 {-1.21/21 3.3s <<< N offer.} 14. b4
{+1.97/22 3.9s <<< Piece declined - pawn sacrifice.} Na7 {-1.90/26 17s} 15.
b5 {+4.19/20 2.5s} Rd6 {-3.44/22 7.2s} 16. a4 {+5.01/17 2.6s} Nf6 {-3.10/20
4.5s} 17. Ne5 {+4.89/19 2.5s} Nd5 {-2.94/20 1.1s} 18. Ne4 {+5.13/18 1.2s}
f6 {-3.06/22 2.8s} 19. Nd3 {+4.81/20 3.6s} Rb6 {-3.64/21 9.0s} 20. c4
{+4.90/18 1.3s} g5 {-3.60/18 0.61s} 21. Ba3 {+6.10/20 3.3s} Qe8 {-4.16/18
0.64s} 22. Bc5 {+6.49/19 0.96s} Bxe4 {-5.71/19 1.7s} 23. Qxe4 {+6.85/20
1.8s} f5 {-5.20/19 0.76s} 24. Qe1 {+7.66/21 1.8s} Nf4 {-5.78/17 0.37s} 25.
Nxf4 {+8.11/20 1.5s} gxf4 {-6.75/19 1.4s, White wins by adjudication} 1-0

The shortest win by black. Sf-Classical gave itself an advantage until after it was lost, then Stockfish made a B offer it could not accept. Then Stockfish won with white also.
[Event "G/60''+1'' 128MBh Sy.6"]
[Site "AMD PhIIX6 Sy.6 6x1core"]
[Date "2021.02.16"]
[Round "65"]
[White "Sf-Classical_21021123_x64"]
[Black "Stockfish_21021123_x64"]
[Result "0-1"]
[ECO "A41"]
[Opening "Neo-Old Indian"]
[Variation "2.Nf3"]
[TimeControl "60+1"]
[Termination "adjudication"]
[PlyCount "69"]

1. d4 {book} d6 {book} 2. Nf3 {book} Nc6 {book} 3. c4 {book} e5 {book} 4.
d5 {book} Nce7 {book} 5. e4 {+1.00/23 2.3s} Nf6 {-1.22/24 12s} 6. Nc3
{+1.04/23 1.0s}  b6 {-1.01/22 5.0s}
( {>>> Sf 1-0 Sf-Cl} 6... Ng6 {-0.58/24 0.91s} 7. h4 {+1.26/23 4.3s} h6 {-0.78/26
9.7s} 8. g3 {+1.28/18 0.85s} Be7 {-0.40/20 2.3s} 9. Be2 {+1.31/20 1.3s} c6
{-0.52/24 6.1s} 10. a4 {+1.35/25 14s} Bg4 {-0.62/19 1.2s} 11. Nh2 {+1.35/22
5.5s} Bxe2 {-0.54/20 1.4s} 12. Qxe2 {+1.29/19 1.4s} Nf8 {-0.59/22 1.4s} 13.
Be3 {+1.29/20 3.3s} cxd5 {-0.51/24 3.0s} 14. cxd5 {+1.43/20 1.7s} Qa5
{-0.55/24 1.2s} 15. O-O {+1.62/18 1.0s} N8d7 {-0.50/27 6.0s} 16. Rfb1
{+1.60/19 2.0s} O-O {-0.55/20 0.94s} 17. Nf1 {+1.70/20 1.4s} Qa6 {-0.68/24
5.7s} 18. Qxa6 {+1.62/20 2.0s} bxa6 {-0.78/24 1.4s} 19. a5 {+1.62/21 5.2s}
Rab8 {-0.51/26 7.6s} 20. f3 {+1.55/20 1.5s} Rb7 {-0.66/25 1.7s} 21. b4
{+1.91/20 1.3s} Rfb8 {-0.57/21 1.2s} 22. Ra4 {+1.25/28 20s} Ne8 {-0.44/26
6.8s} 23. Kg2 {+1.26/22 2.3s} Kh7 {-0.58/26 5.0s} 24. Bd2 {+1.74/20 1.2s}
g6 {-0.58/25 0.69s} 25. Nd1 {+0.89/22 8.6s} Nc7 {-0.44/25 3.6s} 26. Rc1
{+1.90/16 0.33s} Nb5 {-0.12/19 0.68s} 27. Nf2 {+1.48/20 3.1s} f5 {-0.36/24
1.9s} 28. Rc6 {+1.41/16 0.42s} Rf8 {-0.44/22 1.7s} 29. Ra1 {+1.83/16 0.28s}
Nb8 {-0.50/28 8.9s} 30. Rc4 {+1.61/22 1.9s} Nd7 {-0.62/22 0.44s} 31. Rac1
{+1.79/18 0.24s} f4 {-0.61/21 0.84s} 32. gxf4 {+1.67/19 0.50s} exf4
{-0.57/20 0.30s} 33. Nd3 {+1.55/17 0.65s} Ne5 {-0.66/21 0.62s} 34. Nxe5
{+1.30/20 0.79s} dxe5 {-0.50/19 0.48s} 35. Be1 {+1.46/17 0.34s} Rg8
{-0.74/20 0.91s} 36. Nd2 {+1.19/21 3.3s} g5 {-0.57/20 0.47s} 37. h5
{+1.50/20 1.9s} g4 {-1.06/23 4.9s} 38. Nb3 {+1.45/19 0.33s} gxf3+ {-1.01/18
0.32s} 39. Kxf3 {+1.52/19 0.43s} Bd6 {-1.59/24 2.3s} 40. Nc5 {+1.67/19
0.46s} Rc7 {-1.40/19 0.37s} 41. Ne6 {+1.69/19 0.29s} Rxc4 {-1.48/19 0.22s}
42. Rxc4 {+2.13/19 0.53s} Rg1 {-1.63/19 0.35s} 43. Bf2 {+1.87/18 0.67s} Rb1
{-1.38/19 1.0s} 44. Bc5 {+2.02/19 3.1s} Kg8 {-2.14/20 0.73s} 45. Kg4
{+1.93/16 0.25s} Kf7 {-1.95/21 1.3s} 46. Kf5 {+2.91/17 0.28s} Rh1 {-2.15/21
0.73s} 47. Nd8+ {+2.78/17 0.58s} Ke8 {-1.87/20 0.34s} 48. Nc6 {+3.05/21
1.2s} f3 {-1.91/20 0.37s} 49. Rc2 {+2.62/21 0.95s} Rh4 {-2.01/20 0.48s} 50.
Rf2 {+2.32/20 1.2s} Bxc5 {-1.98/22 1.1s} 51. bxc5 {+2.67/20 0.58s} Nc3
{-2.35/22 1.6s} 52. Rxf3 {+2.71/20 0.94s} Rxh5+ {-2.64/18 0.71s} 53. Ke6
{+2.63/19 0.51s} Nxe4 {-2.90/21 1.9s} 54. Rc3 {+2.53/18 0.88s} Ng5+
{-2.79/18 0.42s} 55. Kf5 {+2.49/19 1.0s} Nf7+ {-2.64/19 0.49s} 56. Kf6
{+2.44/18 0.76s} e4 {-3.74/22 3.5s} 57. Nb4 {+2.55/18 0.58s} Rh1 {-2.09/17
0.47s} 58. c6 {+3.25/17 0.50s} Rf1+ {-1.56/15 0.28s} 59. Ke6 {+3.95/19
0.96s} Nd8+ {-1.46/18 0.90s} 60. Kd6 {+4.09/18 0.60s} Nf7+ {-1.75/19 0.52s}
61. Kc7 {+2.57/22 1.8s} Ke7 {-2.71/20 1.2s} 62. Rg3 {+2.69/19 0.72s} h5
{-2.81/21 0.98s} 63. Rg7 {+5.01/20 2.3s} Kf8 {-2.54/19 1.2s} 64. Rg6
{+4.91/19 0.64s} h4 {-2.64/18 0.37s} 65. Nxa6 {+5.40/17 0.45s} Ne5
{-2.85/19 1.4s} 66. Re6 {+5.71/19 1.6s} h3 {-3.26/18 1.2s} 67. Kb8
{+5.48/19 0.87s} Rb1+ {-3.93/21 3.6s} 68. Kxa7 {+6.00/20 0.55s} h2
{-3.86/16 0.24s} 69. c7 {+5.98/21 0.69s} Rc1 {-2.81/16 0.29s} 70. Rh6
{+6.60/19 0.54s} e3 {-6.43/18 2.7s} 71. Rxh2 {+5.94/19 1.7s} Ke7 {-6.12/15
0.15s} 72. Re2 {+6.44/18 0.90s} Ng4 {-7.25/17 1.3s, White wins by
adjudication 1-0} )
7. a4 {+1.40/22 6.3s} a5 {-0.67/18
0.77s} 8. g3 {+1.50/23 4.7s} Nd7 {-0.50/19 1.5s} 9. Bh3 {+1.55/22 2.1s} g6
{-0.30/18 1.1s} 10. O-O {+0.63/22 8.4s} Bg7 {-0.16/19 1.2s} 11. Ne1
{+0.49/21 6.8s} O-O {-0.13/25 5.2s} 12. Nd3 {+0.48/20 0.70s} f5 {-0.11/21
0.74s} 13. Bg2 {+0.50/23 4.8s} Nc5 {+0.57/20 1.1s} 14. Nxc5 {+0.72/21 1.5s}
bxc5 {+0.59/18 1.1s} 15. Bd2 {+0.25/24 10s} f4 {+0.64/17 1.6s} 16. Nb5
{+0.28/21 0.74s} g5 {+0.87/22 8.4s} 17. Bc3 {+0.36/25 5.5s} g4 {+1.42/20
2.1s} 18. Qd2 {+0.75/19 0.68s} f3 {+0.86/24 3.3s} 19. Bh1 {-0.05/22 3.0s}
Rf7 {+1.85/21 0.97s} 20. Qg5 {0.00/23 6.2s} Kh8 {+2.48/21 3.0s} 21. Rad1
{+0.21/23 4.6s <<< Sf-Cl still gies itself += but it is lost.} Bf6
{+3.31/17 0.77s} 22. Qc1 {-1.06/24 8.9s} Ng8 {+3.48/21 2.1s} 23. Rd3
{-1.67/19 0.98s} h5 {+3.84/19 1.5s} 24. Bd2 {-1.56/17 0.32s} Ne7 {+3.69/18
2.4s} 25. Qe1 {-1.12/20 2.4s} Ng6 {+3.72/19 2.0s} 26. Qe3 {-1.06/19 0.82s}
Qg8 {+3.89/21 5.1s} 27. h3 {-1.71/20 2.1s} Bd8 {+3.90/19 1.8s} 28. h4
{-1.42/18 0.63s} Bxh4 {+6.27/17 1.3s} 29. Qh6+ {-4.22/17 1.5s} (29. gxh4
Nf4 {!} (29. .. g3 {??} 30. fxg3 Nxh4 31. Qh6+ Rh7 32. Qf6+ {+-}) 30. Kh2
Qd8 {-+}) 29. .. Rh7 {+6.94/21 1.4s} 30. Qe3 {-3.89/17 0.15s} Bd8 {+6.94/20
1.9s} 31. Re1 {-5.68/18 1.8s} h4 {+7.13/20 1.5s} 32. Bg2 {-5.98/17 0.41s}
Qg7 {+7.78/22 6.3s} 33. Ra3 {-6.00/20 1.2s} Bd7 {+8.32/18 2.3s} 34. Rf1
{-6.03/17 0.35s} Nf4 {+10.18/19 0.94s} 35. Bxf3 {-10.01/19 0.91s, Black
wins by adjudication} 0-1
Attachment: Sf-SClas_1core.pgn - Games (1639k)
Parent - - By Ray (****) Date 2021-02-17 02:10 Upvotes 1

> For AMD PhIIX6 (6 'real cores', it has no HT, popcnt could be added, alter the Makefile, but I used abrok exe.)


Yes, great CPUs, I've only just stopped using mine. At at time when Intel really only offered 4 cores with hyperthreading, I could get so much more done with six proper cores.
Parent - By MrKris (***) Date 2021-02-17 06:37
I was shocked by Stockfish (with NNUE) doing so well with about 1/2 the nps of Sf-Classical (use NNUE=no) because Classical never has utilized avx etc. etc. that NNUE can use on Intel's (even older ones I think) and newer AMD's.
Up Topic The Rybka Lounge / Computer Chess / New (w/ NNUE) Stockfish on an old (x64) CPU

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill