Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Question: Re: HiarcsMark

Author: Melvin S. Schwartz

Date: 19:35:50 06/27/99

Go up one level in this thread



On June 27, 1999 at 17:49:14, Mark Young wrote:

>On June 27, 1999 at 17:26:29, Melvin S. Schwartz wrote:
>
>>
>>On June 27, 1999 at 15:33:48, Terry Ripple wrote:
>>
>>>On June 27, 1999 at 14:00:33, Melvin S. Schwartz wrote:
>>>
>>>>
>>>>Does anyone have factual information regarding the reason for both the short and
>>>>long test?
>>>>
>>>>Question #2: How does these tests relate to the playing of a game at 40/2?
>>>>
>>>>Question #3: How many of these tests, either short or long are we supposed to
>>>>run?
>>>>
>>>>Question #4: On what criteria do you base your answers to the above?
>>>>
>>>>Now, I ran a short test with 32 megs for hash tables and the score was 133
>>>>HiarcsMark 39kN/S. When I ran the same test with 64 megs, the score was 131
>>>>HiarcsMark 39 kN/S.
>>>>At 145 megs the score was lower in kN/S and HiarcsMark.
>>>>
>>>>When I ran the long test with 32 megs of HT, the score was 280 HiarcsMark with
>>>>35 kN/S. When I ran the same test with 64 megs for HT, the score was still 280
>>>>HiarcsMark but 34kN/S. Please keep in mind that I only ran one test for each
>>>>set-up. Now, there is absolutely nothing in regards to how these tests relate to
>>>>actual play or how many tests should be run - with Fritz 5.32 I believe it's all
>>>>automatic.
>>>>
>>>>I thank you in advance for any assistance you can provide.
>>>>
>>>>Regards,
>>>>Mel, Presdent of the Hiarcs Confusion Club. New menbers are welcome. There is no
>>>>annual membership fee. The only requirement is to own Hiarcs 7.32 and the rest
>>>>should come naturally.
>>>--------
>>>Hi Mel,
>>>  Did you know that the time control is in direct relation with how many MB of
>>>Hash Size you use and the speed of your computer is in direct relation with how
>>>high the KN/S will be! Whats your CPU Speed and total RAM?
>>>  My CPU Speed is a AMD K6-2, 266mhz MMX with 64RAM: My average HI-Mark is 93,
>>>and i get a 26 KN/S when following the Optimum Hash Size in Configurations.
>>>For instance, info recommends for Game/60min.= 24MB Hash Size but i get a bit
>>>better HI-Mark using 18.432MB which gives me a 93 Hi-Mark and 26 KN/S.
>>>At 30min. it recommends Optimum of 12MB and i use 12.288MB which gives me a -
>>>94 Hi-Mark with 26 KN/S.
>>>  Now the Confusion Starts: I discovered that if i ignor all the Recommendations
>>>and set the Hash Size for Only 1 or 2 MB for any Time Control- Setting, i will
>>>get a Wopping Hi-Mark of 193 and 27 KN/S.
>>>  I got a recommendation to use this Formula: 2 x KN/S x Average Sec.per Move:
>>>Lets say a time of 40/2hour.= 180 seconds per move in the above formula so here
>>>we go: 2 x 26 x 180 = 9360 KB = 9.360 MB of Hash Size! I know it seems low but
>>>don`t forget that Fritz fills up the Hash Table 10 times Faster than Hiarcs does
>>>and that Optimum Hash recommendation in the Configuration settings was for meant
>>>for Fritz!! I didn`t mean to cofuse you even more but after all, thats why we
>>>joined the Confusion Club of Hiarcs!!!!! Ha Ha!!
>>
>>Hello Terry,
>>
>>You can consider yourself an honary member of the Club. Thanks for adding more
>>confusion and causing me to take two aspirin. Now, you didn't really address the
>>questions as to how many of these tests we are supposed to run,
>
>I would run it 1 time and then never run it again...unless for some reason you
>think your computer is running slow for some reason, the hiarcsmark makes an ok
>check for this.
>
>
> and what is the
>>purpose for having a short and long test?
>
>Just to check the system speed at a short and long time period. With Hiarcs I
>would only use the long test.
>
> Remember, I'm a Vitamin expert, not a
>>computer whiz.
>>
>>So the formula in System Configuration is wrong? Hmmm? Maybe that's why
>>ChessBase doesn't answer my question about that; however, Gambit-Soft support
>>said it was correct! Anyway, how could Hiarcs only need such a small amount of
>>megs even though it's different than Fritz? Are you saying that with more megs
>>for HT it will perform worse???
>
>
>Don't worry about hash size with Hiarcs 7. You gain nothing by using more then
>64mb, but it does not hurt you either. I see nothing that shows using 64mb hurts
>you at fast time controls. I would just pick a setting you like in your own mind
>and keep it there.
>
Hello Mark,

I thank you for making life a little easier for me. You have put my mind at
ease. I will now ask you instead of ConfusionBase, err, I mean ChessBase. Now,
Mark, I have two more questions for you:

1. Would I be correct in assuming that the HiarcsMark is more important than
kN/S? In other words, are we really after a higher HiarcsMark as opposed to a
higher kN/S score?

2. Here is a game where in the opening Hiarcs played a move that the info screen
highlighted in red as "no":

1.e4...c5  2.Nf3...e6  3.d4...cxd4  4.Nxd4...Nf6  5.Nc3...Nc6  6.Nxc6...bxc6
7.e5...Nd5  8.Ne4...f5  9.Nd6+ and here Hiarcs played Bxd6 when the info window
displayed Bxd6 no with the no highlghted in red and was the only move listed.
How do you explain Hiarcs making a move that is displayed as not to do? Please
try it on your Hiarcs and let me know what happens.

Mel
>
>If you want to get a clearer understanding how hash effects hiarcs 7.32 speed,
>just runs set of problems with hiarcs 7.32 and see how it performs with
>different hash sizes.
>
>>
>>I think all the CM6000K guys must be having a good laugh about all this - maybe
>>I should get CM6000K. This whole situation with Hiarcs is in my opinion -
>>ABSURD. The wrong manual, wrong hash table info, no explanation about
>>HiarcsMark, tablebase confusion - I consider it fortunate we were able to run
>>the program at all. ChessBase deserves a lot of credit for being unresponsive
>>and totally without regard for the consumer. Either they don't answer, or they
>>can't answer. Even Mark Uniacke never responded to my query about hash tables.
>>Well, at this point, I don't know if what you say above about "your" formula for
>>HT is true. I would like to hear from Mark Uniacke here and get the truth. Also,
>>I am still interested in what the reason is for the short and long tests? And
>>how many tests are we supposed to run? Furthermore, is it fact that a higher
>>HiarcsMark with a lower kN/S is better than a lower HiarcsMark with a higher
>>kN/S?
>>
>>For your information: I have an AMD K6 400MHz processor with 192 MB of RAM and
>>10.0 GB on the hard drive.
>>
>>Regards,
>>Mel>
>>>Best Regards,
>>>Terry



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.