Author: James T. Walker
Date: 04:24:43 06/17/99
Go up one level in this thread
On June 16, 1999 at 22:10:26, Melvin S. Schwartz wrote: > >On June 16, 1999 at 21:18:39, James T. Walker wrote: > >>On June 16, 1999 at 09:30:29, Melvin S. Schwartz wrote: >> >>> >>>On June 16, 1999 at 07:29:58, James T. Walker wrote: >>> >>>>On June 16, 1999 at 07:04:46, Terry Ripple wrote: >>>> >>>>> Also, would it be Best to use less for Faster time controls like 30 min.,- >>>>>15 min. Speed Chess? And what do you think would be the Max. i could use for >>>>>Slower time controls having Only 64 RAM and using an AMD K6 at 266mhz? >>>>> >>>>> Thankyou so much, >>>>> Terry >>>> >>>>Hello Terry, >>>>With 64 Meg Ram you can probably use 44 Meg for hash tables with no problem at >>>>the longer time controls. It's one of the buttons under "Load engine" and >>>>probably there because it's recommended for computers with 64 M RAM. I use 16 >>>>Meg for game/5 min but 8 meg is plenty for that time control especially on a >>>>slow computer like you have. You can also run some test using the "HiarcsMark" >>>>and try different hash table settings to find which gives you the highest >>>>"HiarcsMark". >>> >>>Hello Jim, >>> >>>What Hiarcs mark? In my Hiarcs 7.32 it says Fritz mark!! Also, the so-called >>>manual with Fritz 5.32 on top, gives the same formula for hash-table as Fritz - >>>even Hiarcs 7.32 does the same under levels. >>> >>>Mel >>>>Jim Walker >> >>Hello Mel, >>I don't know about your program but my HIARCS 7.32 has a HiarcsMark. It is in >>the same place as the "FritzMarK" but definitely labeled "HiarcsMarK". It's >>under the Menus in this order:Levels,Engine research,HiarcsMark. And when you >>click on HiarcsMark the box that comes up is also labeled HiarcsMark at the top >>and shows the (hash=16384kB) beside the label. If your Hiarcs 7.32 does not > >Hello Jim, > >You are correct - it does say HiarcsMark. I could have sworn it said FritzMark >when I first looked at it - I wonder...Now, if you look in Levels and click on >system configuration - I am looking at it right now - it states: "Optimum hash >size (200MHz): 72000KB". Take a look at it and let me know what you see. This is >the same formula as for Fritz. > >What you saw initially - 16384KB was a default setting but does not correlate >with the info under Optimum hash size. If you raise the hash figure to the >highest total, you will find out what your hardware can produce for hash tables. >Also, you will find in the mimimum hash proposed a figure different than >16384KB. In fact, you will see a minimum to a figure what you should have in >system configuration under Levels. This is the same formula and same figures you >will find on Fritz 5.32. Is it correct? That is the question I have put to >ChessBase without an answer as of this writing. > >Mel ********************** Hello Mel, Yes, I know all this. The 16384 that shows up when doing the HiarcsMark test is simply what you have selecte at the time you run the test. I only mentioned that because that's what I used for the "Short" test. I think you will find that for the short test which I think is a good performance test for Blitz type games you can experiment with your hash settings and find that 8 meg is slightly faster than 16 meg. At least on mine it's 1 HiarcsMark faster(Not worth bothering with). If you run the 'Long' test you will probably find that 64 meg is faster than 128 meg (It was on my PC). So 64 Meg is plenty of hash for tournament games(40/2). The difference in these figures is not worth worrying about though. By the way, Nimzo99 has a "NimzoMark" as does Junior with it's "JuniorMark". Jim Walker ************************* > > >>have this then I would complain! :-) Also since Hiarcs searchs about1/10 the >>speed as Fritz it should need nowhere near the hash tables size Fritz needs. I >>use 16 meg for Fritz at game/5 minutes and never see the 100% sign come on. So >>I'm sure Hiarcs will work fine with 8 Meg at blitz. My HiarcsMarks are: >>PII 333 short:122(57.25s) 34Kn/sec >>K6-2 400 short: 110 (63.4s) 31Kn/sec >>My FritzMarks are: >>PII 333 short 250 364Kn/s >>K6-2 400 short 288 420Kn/s >>So you can see that Fritz is at least 10 times faster than Hiarcs in nodes/sec. >>I believe this rate is directly related to how fast you can fill the hash >>tables. >>Jim Walker
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.