CCC has serious hardware update!

[Jouni]

"Now, on CCC and TCEC, SF runs at about the same speed : 100 Mn/s in the opening and 150 Mn/s in late middlegame." Maybe SF prefers Intel Xeon architecture? BMI and better hyperthreading?

[the_real_greco]

I was very happy to see the CCC upgraded their system. What I want to see now, is who still dares to make the silly "bullet isn't real chess" argument, when "bullet" on CCC is played at 200M nodes per move.

From what I can tell, the advantage of having huge hardware is to maintain quality of play, while having games that don't take all day.

[Modern Times]

From what I can tell, the advantage of having huge hardware is to maintain quality of play, while having games that don't take all day.

Totally agree !

[Jouni]

Clearly SF and Lc0 are dominating now both in CCC and TCEC!

[jshriver]

CPUs: 2 x AMD EPYC 7H12
GPU: 2x A100 (40 GB GPU memory)
Cores: 256 cores (128 physical)
RAM: 512GB DIMM DDR4 2933 MHz (0.3 ns)
SSD: 2x Micron 5210 MTFD (2TB) in RAID1
OS: CentOS 8

What's TCEC answer ?

That is a beautiful machine!

[Albert Silver]

Actually, the CPU difference is smaller, since the GPU server uses its own CPU cores, which are:
CPU: Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz, 32 vcores
So they are :
2.5/2.6 * 100 = 96% of the speed or about 4% slower than the AMD cores.

I don't think you can compute their relative performances that way.

[Nordlandia]

The SSD could have been bigger to accommodate more 7-peices.

The question if this is right as 7-pieces can cause memory to get filled up quite quickly as syzygy get catched.

[Dann Corbit]

Actually, the CPU difference is smaller, since the GPU server uses its own CPU cores, which are:
CPU: Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz, 32 vcores
So they are :
2.5/2.6 * 100 = 96% of the speed or about 4% slower than the AMD cores.

I don't think you can compute their relative performances that way.

It's just a back of the envelope estimation.
According to my understanding, CCC and TCEC CPU engines run at about the same speed now.

[mwyoung]

Clearly SF and Lc0 are dominating now both in CCC and TCEC!

Yes, but you do not need massive hardware. SF, and Lc0 also dominate on much less hardware, even with fast time controls.
The best chess will always be played by the best software. Current tournament at 1m+1s 32 threads.

Code: Select all

End of Year Tournament 1m+1s, Blitz 1.0  2020

                                   1            2            3            4            5            6            7            8            9            0            1            2            
1   Stockfish 12                   ************ ½½½½½        ½½½½½        ½1½½½        1½½½½        ½½1½½        1111         ½½1½½        1111½        ½1½½         111½1        111½          36.0/52  865.25
2   Stockfish 311220               ½½½½½        ************ ½1½½½        1½½½         ½½½½½        ½½½½         ½1½½½        1½½½½        1½11         11½11        ½1111        11111         36.0/52  804.50
3   Lc0 v0.26.3                    ½½½½½        ½0½½½        ************ ½1½½½        ½½½½1        1½½½         ½½½1½        ½1½½1        ½½½1         ½111½        1111         1111½         34.5/52  801.00
4   Dragon by Komodo Chess,64-bit  ½0½½½        0½½½         ½0½½½        ************ 1½1½         ½½½½½        11½½½        ½1½½½        ½1½½½        1111         11111        11111         34.5/52  754.50
5   Stockfish 11 64 POPCNT         0½½½½        ½½½½½        ½½½½0        0½0½         ************ ½½½½         ½1½½½        1½½1½        1½½11        111½1        1½111        1111          33.0/52
6   Fat Fritz (in lc0) 1.2 rtx     ½½0½½        ½½½½         0½½½         ½½½½½        ½½½½         ************ ½½½0½        ½11½1        1½111        ½½101        ½1½11        11111         32.5/52
7   Stockfish 10 64 POPCNT         0000         ½0½½½        ½½½0½        00½½½        ½0½½½        ½½½1½        ************ ½½½½         1½½1½        ½½111        11½½1        11½11         28.5/53
8   Komodo 14.1 64-bit             ½½0½½        0½½½½        ½0½½0        ½0½½½        0½½0½        ½00½0        ½½½½         ************ 11½1         ½1½½½        1111         11½11         27.0/52
9   Ethereal 12.50 (POPCNT)        0000½        0½00         ½½½0         ½0½½½        0½½00        0½000        0½½0½        00½0         ************ ½½½1½        11½1½        11111         20.0/52
10  Xiphos 0.6 NO-POPCNT           ½0½½         00½00        ½000½        0000         000½0        ½½010        ½½000        ½0½½½        ½½½0½        ************ 01½½½        1½11          16.5/52
11  Fritz 17 Popcnt                000½0        ½0000        0000         00000        0½000        ½0½00        00½½0        0000         00½0½        10½½½        ************ 11111         12.0/53
12  Deep Fritz 14 x64              000½         00000        0000½        00000        0000         00000        00½00        00½00        00000        0½00         00000        ************   2.5/52

[Albert Silver]

Actually, the CPU difference is smaller, since the GPU server uses its own CPU cores, which are:
CPU: Intel(R) Xeon(R) Platinum 8163 CPU @ 2.50GHz, 32 vcores
So they are :
2.5/2.6 * 100 = 96% of the speed or about 4% slower than the AMD cores.

I don't think you can compute their relative performances that way.

It's just a back of the envelope estimation.
According to my understanding, CCC and TCEC CPU engines run at about the same speed now.

My point was that not all CPUs share similar IPCs. Right now I think AMD has a very serious edge.