Twin 2080Tis and large NN

[shrapnel]

People over at the lc0 Forums wanted me to post in the lichess format, which unfortunately doesn't show many details.
Anyway, here you go...
Shrapnel,Lc0 v0.22.0 (2338) - Deepblueocean,CorChess 5.0 121119 (2518) [B90]
Rated game, 16 min Engine Room, 15.11.2019
[0.31;0.22]
CorChess 5.0 121119 64 BMI2 (12 cores): 33.9 plies; 22,158kN/s Intel(R) Core(TM) i9-7920X CPU @ 2.90GHz 2904MHz, (12 cores, 24 threads), Attacker Pro 4-cubail.ctg, 32768 MB [pgn]1.e4 B/0 0 c5 B/0 0 2.Nf3 B/0 0 d6 B/0 0 3.d4 B/0 0 cxd4 B/0 0 4.Nxd4 B/0 0 Nf6 B/0 0 5.Nc3 B/0 0 a6 B/0 0 6.h3 B/0 0 e5 B/0 0 7.Nde2 B/0 0 h5 B/0 0 8.g3 B/0 0 Nbd7 B/0 0 9.a4 B/0 0 Be7 B/0 0 10.Bg5 B/0 0 b6 B/0 0 11.Bg2 B/0 0 Bb7 B/0 0 12.0-0 B/0 0 Rc8 B/0 0 13.h4 B/0 0 0-0 B/0 0 14.Qd2 B/0 0 Re8 B/0 0 15.Rfd1 B/0 0 Rc4 B/0 0 16.Qd3 0.31/13 14 Rc7 (Rb4) B/0 0 17.Nc1 0.35/19 17 Qc8 (Nc5) B/0 0 18.Nb3 0.36/12 23 Bf8 B/0 0 19.Qe2 0.32/13 28 Rc4 B/0 0 20.Nd2 0.30/14 38 Rb4 0.22/31 51 21.Rab1 (Bh3) 0.28/16 0 21...Rd4 (g6) 0.13/32 1:08 22.Nb3 (Ra1) 0.31/11 58 22...Rb4 0.00/34 18 23.Nd2 0.29/12 0 g6 0.00/32 16 24.Bf1 (Bh3) 0.27/13 53 24...Rd4 0.00/34 35 25.Bh3 (Bg2) 0.26/15 0 25...Qc7 0.15/33 29 26.Bg2 0.25/15 0 Nh7 (Kg7) 0.14/32 17 27.Be3 0.30/14 33 Nhf6 0.21/34 0 28.Nf1 (Nb3) 0.26/14 1:05 28...Rb4 (Rxd1) 0.22/33 42 29.Bg5 (Bd2) 0.35/11 17 29...Nh7 (Rd4) 0.00/32 8 30.Bd2 (Be3) 0.39/13 27 30...Nhf6 0.00/33 17 31.Nb5 (Bg5) 0.54/24 16 31...Qxc2 0.15/38 33 32.Ne3 0.52/29 12 Rxe4 0.25/37 7 33.Nxd6 0.45/30 9 Bxd6 0.23/39 6 34.Nxc2 0.43/29 7 Rxe2 0.30/38 10 35.Bxb7 0.43/29 7 Bc5 (Ne4) 0.14/38 18 36.b4 (Ne3) 0.59/23 54 36...Bxf2+ 0.15/32 15 37.Kf1 0.54/21 0 Rxd2 0.00/33 11 38.Rxd2 0.52/24 4 Bxg3 0.07/32 11 39.Bxa6 0.54/23 0 Ra8 (Bf4) 0.54/34 48 40.Bb5 0.57/21 1:02 Nf8 0.39/35 0 41.Rd6 (Ra1) 0.51/22 1:21 41...Bxh4 0.00/32 18 42.Bc6 (Rxb6) 0.48/28 0 42...Rb8 (Rc8) 0.23/34 29 43.Na3 0.77/23 30 Ng4 0.49/37 0 44.a5 0.68/25 11 Ne3+ (Be7) 0.38/36 0 45.Kg1 1.21/25 30 Nf5 (bxa5) 0.41/38 1 46.Rd3 1.23/17 24 Ne7 0.42/35 41 47.Bg2 1.21/22 0 f5 0.20/35 9 48.Rd6 (Bd5+) 1.18/20 0 48...Nc8 0.22/35 12 49.Rdd1 (Bd5+) 1.14/19 22 49...e4 0.13/30 8 50.a6 1.14/20 0 Bg3 0.00/33 23 51.Bf1 (Nb5) 1.11/18 25 51...Kg7 0.00/32 11 52.Rbc1 1.14/18 0 Bf4 0.35/32 27 53.Rc2 (Rc3) 1.20/17 37 53...Ne6 (Ne7) 0.24/33 35 54.Nb5 1.24/17 20 Ne7 0.34/33 18 55.Bc4 1.12/22 0 Ra8 0.27/31 4 56.a7 (Rd7) 1.04/21 7 56...Kf6 0.31/31 6 57.Rd7 0.99/24 10 h4 (Nc6) 0.39/34 25 58.Bxe6 1.42/19 21 Kxe6 0.32/36 5 59.Nc7+ 1.40/22 3 Bxc7 1.04/32 9 60.Rcxc7 1.37/19 0 Nc8 (Nc6) 1.10/34 14 61.Rd8 1.80/18 16 Rxa7 2.52/36 13 62.Rdxc8 1.80/22 0 Ra4 2.15/29 0 63.Rc6+ (Re8+) 1.80/21 4 63...Ke5 2.00/30 4 64.Rxg6 1.80/19 0 Rxb4 3.06/31 13 65.Rh6 1.81/19 0 Kf4 (Rb1+) 4.44/32 37 66.Rc2 1.95/17 9 Kg5 2.64/29 1 67.Rh8 2.00/16 4 Rb1+ (e3) 2.71/29 11 68.Kh2 2.14/17 6 Rb5 (b5) 147.92/40 0 69.Rf2 (Rg2+) 3.01/10 8 69...e3 (Re5) 147.92/33 3 70.Rf1 3.69/10 4 e2 (Rb2+) 147.92/33 0 71.Rg8+ (Re1) Deepblueocean,CorChess 5.0 121119 resigns (Lag: Av=0.40s, max=2.6s) 8.29/8 5 1-0 [/pgn]

[corres]

People over at the lc0 Forums wanted me to post in the lichess format, which unfortunately doesn't show many details.
Anyway, here you go...
Shrapnel,Lc0 v0.22.0 (2338) - Deepblueocean,CorChess 5.0 121119 (2518) [B90]
Rated game, 16 min Engine Room, 15.11.2019
[0.31;0.22]
CorChess 5.0 121119 64 BMI2 (12 cores): 33.9 plies; 22,158kN/s Intel(R) Core(TM) i9-7920X CPU @ 2.90GHz 2904MHz, (12 cores, 24 threads), Attacker Pro 4-cubail.ctg, 32768 MB...

Some note
CorChess is modified for Long Time (correspondent) chess so it is not the best for short games and 12 real cores on 2.9 GHz is not a peak for a derivative of Stockfish.
Please, read my test where I used Threadripper 16 cores with 4.0 GHz against 2 x RTX 2060 (~ 1 X RTX 2080 Ti).
As my test proved against Stockfish (and any derivative) on a machine with 16 x 4 GHz cores is enough only one RTX 2080 Ti.
And you have 2 x RTX 2080 Ti...

[shrapnel]

I don't think everything is cut and dried as far as lc0 is concerned and there are a lot of variables involved.
And I don't choose my opponents ; I play them as they come.
It's thrilling enough for me to DEFEAT AB engine users using an AI Engine using GPUs, something which just a few years ago everybody said was impossible. I hardly care what hardware they use, only what AB Engine they are using.
Even Damir was saying I'm winning because of my twin 2080 Ti GPUs and lc0 is still weak.
So what ? Even AB Engine users who use strong CPUs have an advantage over weak CPU users.
In Chess, all that matters is the WIN, whether because of Software or Hardware hardly matters to me.
In any case, lc0 will only improve with Time and so will the GPUs.
So for me at least, exciting times are ahead as far as AI engines are concerned.

[shrapnel]

Another superb win against latest Stockfish, while using Black ; Sergio NNs of course...
[pgn][Event "Rated game, 16 min"] [Site "Engine Room"] [Date "2019.11.16"] [Round "?"] [White "Stara Voittov, Stockfish 141119 64"] [Black "Shrapnel, Lc0 v0.22.0"] [Result "0-1"] [ECO "C24"] [WhiteElo "2423"] [BlackElo "2356"] [Annotator "0.00;-0.39"] [PlyCount "77"] [EventDate "2019.11.16"] [SourceTitle "playchess.com"] [TimeControl "960"] {Stockfish 141119 64 POPCNT (12 threads): 29.3 plies; 10,395kN/s Intel(R) Core (TM) i7 CPU X 980 @ 3.33GHz 3337MHz, (6 cores, 12 threads), The Outskirts Chess Forum.ctg, 32768 MB} 1. e4 {B 0} e5 {B 0} 2. Bc4 {B 0} Nf6 {B 0} 3. Qf3 { B 0} c6 {B 0} 4. Nc3 {B 0} b5 {B 0} 5. Bb3 {B 0} b4 {B 0} 6. Na4 {0.00/28 26} Be7 {B 0 (d5)} 7. d4 {0.00/30 34} d5 {B 0} 8. exd5 {0.00/30 33} e4 {-0.39/12 13 } 9. Qd1 {0.00/27 7} cxd5 {-0.32/13 8} 10. Ne2 {0.00/28 2} O-O {-0.29/12 26 (Nc6)} 11. O-O {0.00/29 13} Nc6 {-0.29/14 22 (h6)} 12. h3 {0.00/30 24} Na5 { -0.32/14 10} 13. Nc5 {0.00/27 2 (c3)} Rb8 {-0.40/15 19} 14. Qe1 {0.00/32 79 (a3)} Rb6 {-0.70/14 25} 15. Bd2 {-0.45/29 32} Nh5 {-0.62/18 0} 16. Ng3 { -0.38/29 11} Nxg3 {-0.62/18 1} 17. fxg3 {-0.35/31 12} Nc6 {-0.55/17 87 (Rg6)} 18. Be3 {-0.30/29 27} Na5 {-0.52/16 0} 19. Bd2 {-0.63/31 55} Rg6 {-0.50/14 0} 20. Bf4 {-0.50/28 6} h5 {-0.50/16 1 (Qb6)} 21. Qxb4 {-0.32/27 23 (h4)} h4 { -1.02/15 38} 22. Qd2 {-0.83/31 5} hxg3 {-1.09/17 20} 23. Rae1 {-0.53/32 0 (Qe3) } f5 {-1.22/12 36 (Bd6)} 24. Re3 {-0.43/28 9 (c3)} Nxb3 {-1.38/18 22} 25. axb3 {-0.92/36 44 (Rxb3)} Bd6 {-1.45/16 19} 26. Rc3 {-0.79/38 0} Qf6 {-1.57/18 28 (Re8)} 27. Bxd6 {-1.03/31 14} Qxd6 {-1.57/20 0} 28. Qf4 {-1.19/34 28} Qf6 { -1.59/19 4} 29. Rd1 {-1.52/33 21} Kh7 {-1.63/23 0 (a5)} 30. Re3 {-1.41/28 12} Qg5 {-1.67/23 12} 31. Rf1 {-1.42/32 11} Qh6 {-1.70/24 0} 32. Kh1 {-1.33/32 10 (Rc3)} Rff6 {-1.91/20 53 (a5)} 33. Kg1 {-1.38/34 24} a5 {-2.01/20 13} 34. Ree1 {-2.01/34 58 (Rc3)} Qxf4 {-2.38/17 48 (Rf8)} 35. Rxf4 {-2.06/35 21} Rh6 { -2.49/17 0 (Rc6)} 36. h4 {-2.51/31 14} Rc6 {-2.51/17 0} 37. Re3 {-2.80/35 74 (c3)} Rhg6 {-2.77/14 41} 38. Re1 {-3.41/35 0 (Kf1)} Kh6 {-2.72/15 42} 39. Ra1 { -3.80/35 0 Stara Voittov,Stockfish 141119 64 resigns} 0-1 [/pgn]

[corres]

I don't think everything is cut and dried as far as lc0 is concerned and there are a lot of variables involved.
And I don't choose my opponents ; I play them as they come.
It's thrilling enough for me to DEFEAT AB engine users using an AI Engine using GPUs, something which just a few years ago everybody said was impossible. I hardly care what hardware they use, only what AB Engine they are using.
Even Damir was saying I'm winning because of my twin 2080 Ti GPUs and lc0 is still weak.
So what ? Even AB Engine users who use strong CPUs have an advantage over weak CPU users.
In Chess, all that matters is the WIN, whether because of Software or Hardware hardly matters to me.
In any case, lc0 will only improve with Time and so will the GPUs.
So for me at least, exciting times are ahead as far as AI engines are concerned.

These are O.K., but if you want to rally know why you got success you know how strong hardware your partner used. Without this you can not decide about the quality of the net you used.