The Week in Chess

Saturday, November 8, 2014

Sugar 2.0f x64 vs. Fried Shark 141107 x64 - SF Clones War - 100R 1M1S

This is a match of 2 SF clones between the latest Sugar 2.0f x64 released on November 6, 2014 and the latest Fried Shark 141107.

Sugar is currently number 2 in the Owl Rating List but the results showed a devastation of 80.5-19.5 by Shark without winning a single game. Obviously Sugar's score is not the kind of score by a decent SF clone. Something is wrong with the tweaks by the author. There is no quality control before release and needs to be checked.


Sugar 2.0f x64 vs. Fried Shark 141107 x64 - 100R 1M1S
RankEngineScoreFrSuS-B
1Fried Shark 141107 x6480.5/100· ·· ·· ·61-0-39 1569.75 
2Sugar 2.0f x64 19.5/1000-61-39· ·· ·· · 1569.75 


100 games played / Tournament finished

Tournament start: 2014.11.08, 15:00:21
Latest update: 2014.11.08, 16:36:21
Level: Blitz 1/1
Hardware: AMD Phenom(tm) IIX4 945 Processor with 4.0 GB Memory
Operating system: Windows 7 Ultimate Professional Service Pack 1 (Build 7601) 64 bit
Table created with: Arena 3.5
Download the match PGN games here.

6 comments:

  1. There's something wrong somehow. This Sugar version proved to be better than the previous one. To analize. AMD related?

    ReplyDelete
    Replies
    1. I am not sure if this is AMD related but I also run this in another AMD computer which showed the same pattern of zero win before I posted. If this is optimized for another processor, then it is advised that engine authors should not assume of uniform result and instead have their compiles be as generic as possible when it is released to the public.

      Delete
    2. I tested again with Sugar 2.0f x64 Gen which I supposed to be generic but it is the same pattern. Then I run the strongest version Sugar 2.0f x64 SSE4.2 but again was demolished easily. I saw that the depth and nodes per second speed are very low for all the 3 Sugar versions tested. I suspect that it has something to do with a compile using MINGW GCC where it just happily accepts optimization parameters even if it is wrong and then generates slow running executables. This happened to me several times already.

      Delete
  2. Yes.. something wrong with the source code, depth of analysis is too low, this is the reason why I released version now G.
    For some reason I do not understand, add some parameters SPSA makes the engine very slow and weak

    Regards

    Marco Zerbinati

    ReplyDelete
    Replies
    1. Thanks for your comments. I'm glad you were able to see where the problem is. Continue with your good service.

      Delete
    2. I tried to compile your Sugar versions but unfortunately failed using MINGW or Intel C++. Your SFMZ versions are ok.
      Can you check your Sugar source for possible compile errors?

      Delete

Chessdom News