85th Amateur Series Division 4

[Graham Banks]

Congratulations to Vivien (Minic author) for the convincing victory, and promotion for the next series.

Thanks to everybody who took an interest, either through visiting the broadcast or following progress.

[Angle]

Cheng far behind Fruit, Deuterium, and Godel? Something should be wrong here ...

[xr_a_y]

Thanks a lot for the tourney Graham.

I'm still a be confused with Minic's not so good results on the CCRL 40/40 list especially with NNUE such as this 3.04 with noisy notch.
I convinced myself that this is linked with using NNUE without AVX2 hardware. Can you post more spec on the "i5 quad you are running one". But on the other end, this does not seem to affect Igel and Nemo or Rubi. Even Seer 1.2 is stronger than Minic here, and they both using more or less the same net architecture.

On the blitz list, Minic is rated so much higher than on the 40/40 (https://ccrl.chessdom.com/ccrl/404/cgi/ ... librate=no)

On the CEGT 40/40 list, Minic 3.04 + noisy notch is more than 100Elo stronger than 2.46, while they will be even on the CCRL 40/40 list.

Another question would be, which Minic executable are you ising for this test ? and which OS ? I probably should release specific binaries for arch between Nehalem and Skylake...

[Graham Banks]

I used the core2 exe for this tournament. (Windows 7 64-bit)

[xr_a_y]

I used the core2 exe for this tournament. (Windows 7 64-bit)

For sure this is why it is both slow and a bit weak. I probably shall release more exe so that better perf are available on a wider range of cpu.

What is a exact model of your current cpu please ?

[Graham Banks]

Cheng far behind Fruit, Deuterium, and Godel? Something should be wrong here ...

The games are there for you to look through.

O the CCRL 40/15 list, Godel is 2981, Cheng is 2959 and Fruit Reloaded is 2906.

Just a bad tournament for Cheng.

[Graham Banks]

I used the core2 exe for this tournament. (Windows 7 64-bit)

For sure this is why it is both slow and a bit weak. I probably shall release more exe so that better perf are available on a wider range of cpu.

What is a exact model of your current cpu please ?

You'd be best doing what others do - a non-popcount exe, a popcount exe and a bmi2 exe.

[RubiChess]

I used the core2 exe for this tournament. (Windows 7 64-bit)

For sure this is why it is both slow and a bit weak. I probably shall release more exe so that better perf are available on a wider range of cpu.

What is a exact model of your current cpu please ?

You'd be best doing what others do - a non-popcount exe, a popcount exe and a bmi2 exe.

Little addition: All of Grahams computers (even the non-popcount one) support SSSE3 and earlier SIMD instructions.

[xr_a_y]

I used the core2 exe for this tournament. (Windows 7 64-bit)

For sure this is why it is both slow and a bit weak. I probably shall release more exe so that better perf are available on a wider range of cpu.

What is a exact model of your current cpu please ?

You'd be best doing what others do - a non-popcount exe, a popcount exe and a bmi2 exe.

This is the case now more or less. But using NNUE without good vectorisation is often counter productive. Does your i5 has bmi2 and avx or avx2?

[Graham Banks]

I used the core2 exe for this tournament. (Windows 7 64-bit)

For sure this is why it is both slow and a bit weak. I probably shall release more exe so that better perf are available on a wider range of cpu.

What is a exact model of your current cpu please ?

You'd be best doing what others do - a non-popcount exe, a popcount exe and a bmi2 exe.

This is the case now more or less. But using NNUE without good vectorisation is often counter productive. Does your i5 has bmi2 and avx or avx2?

No, but my i7 does.