TCEC update .......

[Jouni]

OK actually FGRL 60 + 0,6 has SF8 - H4 84% = very similar! But stunning anyway.

[Dokterchen]

OK actually FGRL 60 + 0,6 has SF8 - H4 84% = very similar! But stunning anyway.

I have so far (180+1), S8:
Houdini 4 : 60 (+ 39,= 20,- 1), 81.7 %

Also in the same range.

[FriedmannC]

the topic should speak about the future Houdini, not the past Houdini. Sadly, still no news...It makes no sense a final between Houdini 20.07.2016 and Stockfish 8, the latter is too strong. The result would clearly favour Stockfish

[Nay Lin Tun]

From TCEC facebook page"
Both Stockfish and Houdini are going to play with new versions in the TCEC final this year. The engines will bring their best for the Season 9 title.
The Superfinal preparations need a few more days and the games are expected to start after Nov 9th. More precise date and time will be published on Chessdom.com
We believe it is good to extend the period for engine submissions until Nov 9th (noon time CET) if there are last minute updates. And we remind both Stockfish and Houdini that stability is key, as there will be no testing.

[Lyudmil Tsvetkov]

From TCEC facebook page"
Both Stockfish and Houdini are going to play with new versions in the TCEC final this year. The engines will bring their best for the Season 9 title.
The Superfinal preparations need a few more days and the games are expected to start after Nov 9th. More precise date and time will be published on Chessdom.com
We believe it is good to extend the period for engine submissions until Nov 9th (noon time CET) if there are last minute updates. And we remind both Stockfish and Houdini that stability is key, as there will be no testing.

Houdart needs to release in the meantime...

[shrapnel]

Houdart needs to release in the meantime...

Yes. Houdart has had enough time.
I get the feeling while watching online matches that Stockfish 8 has some weaknesses, which even Komodo 10.2 can exploit, but not consistently. Also Komodo seems to lose its way sometimes, even in a winning position.
Maybe its a matter of finding the correct settings, which is not easy, as there are so many Parameters to adjust.
But Houdini is much more ruthless, and I think Houdini 5 will simply POUND Stockfish into the dust !
Anyway, should be interesting.

[Nay Lin Tun]

Martin is testing both SF and Houdini in TCEC 9 server
After 4 minutes from starting position, 32 GB hash, 44 cores

ASM fish 56 million nodes / sec, depth 34,
SF 8 - 49 million nodes/sec, depth 36
H5Dev - 64 million nodes/sec, depth 30

So, ASM fish this version is much faster than officially submitted SF,

https://github.com/tthsqe12/asm/blob/ma ... 4_bmi2.exe

[Milos]

Martin is testing both SF and Houdini in TCEC 9 server
After 4 minutes from starting position, 32 GB hash, 44 cores

ASM fish 56 million nodes / sec, depth 34,
SF 8 - 49 million nodes/sec, depth 36
H5Dev - 64 million nodes/sec, depth 30

So, ASM fish this version is much faster than officially submitted SF,

https://github.com/tthsqe12/asm/blob/ma ... 4_bmi2.exe

That's very unimpressive for 44 cores, kind of what you could get from an old quad E5-2670 server build that would cost under 2k. They must be using low power Xeons.

[APassionForCriminalJustic]

Martin is testing both SF and Houdini in TCEC 9 server
After 4 minutes from starting position, 32 GB hash, 44 cores

ASM fish 56 million nodes / sec, depth 34,
SF 8 - 49 million nodes/sec, depth 36
H5Dev - 64 million nodes/sec, depth 30

So, ASM fish this version is much faster than officially submitted SF,

https://github.com/tthsqe12/asm/blob/ma ... 4_bmi2.exe

That's very unimpressive for 44 cores, kind of what you could get from an old quad E5-2670 server build that would cost under 2k. They must be using low power Xeons.

Your comment here absolutely proves that you have no idea what you are talking about. Lol. What you would expect from a 2670 rig would be WAY LESS than 49 million nodes per second (not even in the same area code). By the way - there are no "quad 2670s". The 2670 Xeons are eight-core Sandybridge, and they are not great by today's standards. Ha-ha. Low power Xeons you say? Sorry - it is more like two 2699 v4s. They are pretty much the best Xeons that one could buy for the dual-socket configuration other than some rare experimental Xeon CPUs that are not readily available to the public. The scaling of the engine comes down to Stockfish; asmFish is clearly faster. Houdini is extremely fast but its depth reached is trash.

[Milos]

Your comment here absolutely proves that you have no idea what you are talking about. Lol. What you would expect from a 2670 rig would be WAY LESS than 40 million nodes per second. By the way - there are no "quad 2670s". The 2670 Xeons are eight-core Sandybridge, and they are not great by today's standards. Ha-ha. Low power Xeons you say? Sorry - it is more like two 2699 v4s. They are pretty much the best Xeons that one could buy for the dual-socket configuration other than some rare experimental Xeon CPUs that are not readily available to the public. The scaling of the engine comes down to Stockfish; asmFish is clearly faster. Houdini is extremely fast but its depth reached is trash.

Are you dumb or something, or you really have problem with understanding english?
Quad means 4, 4x E5-2670, 32 cores all running at 3GHz. Cinebench >4000. I bet your machine is way less than that, especially with those crappy ES cores.
Btw. v4 Xeon at 2.2GHz provides worse nps, than v1 at 3GHz. Try first informing yourself before writing nonsense.