Author: Christophe Theron
Date: 17:59:46 10/10/01
Go up one level in this thread
On October 10, 2001 at 19:00:08, Roy Eassa wrote:
>On October 10, 2001 at 13:58:28, Christophe Theron wrote:
>
>>On October 10, 2001 at 13:09:50, Roy Eassa wrote:
>>
>>>On October 08, 2001 at 19:39:02, Christophe Theron wrote:
>>>
>>>>On October 08, 2001 at 16:25:37, Roy Eassa wrote:
>>>>
>>>>>Tiger's clock is slow when I tell AfterBurner to "Adjust Ticks". If I do NOT
>>>>>choose that setting, Tiger's clock seems OK. This is the exact OPPOSITE result
>>>>>as I expected from that setting.
>>>>
>>>>
>>>>You MUST run the "Speed test" before you have a look at Chess Tiger's clock,
>>>>because the "Speed test" does some adjusting job.
>>>>
>>>>Did you do that every time you have changed the Palm's clock frequency?
>>>>
>>>>If not, you should try again.
>>>>
>>>>0) check "adjust ticks" in AfterBurner
>>>>1) change Palm's frequency
>>>>2) **IMPORTANT**>>>> run "Speed test" <<<<**IMPORTANT**
>>>>3) check if Tiger's clock matches your stopwatch
>>>>4) if the clocks do not match, start again at (1) with a lower frequency.
>>>>
>>>>
>>>>>Well, I guess that explains that.
>>>>
>>>>
>>>>No it does not.
>>>>
>>>>
>>>>>Now, with "Adjust Ticks" turned OFF, I get a TigerMark index of 1.25 with my
>>>>>Prism at 54 MHz, which is kind of disappointing.
>>>>
>>>>
>>>
>>>OK, I tried the exact sequence you suggested. With "adjust ticks" checked, the
>>>speed index came out relatively high (as high as 1.74), but the CLOCK NEVER
>>>MATCHED a stopwatch. It was always slow, though less and less slow as I
>>>decreased the MHz (and YES, I definitely did the speed index first thing after
>>>adjusting the MHz).
>>>
>>>With "adjust ticks" UNchecked, the speed index comes out relatively low (1.24),
>>>but Tiger's clock (which I checked AFTER doing the speed index test) matched a
>>>stopwatch perfectly.
>>>
>>>Now I'm even more confused!
>>
>>
>>
>>If you don't give me the full picture, I can't do much.
>>
>>Until what frequency did you go in the first case?
>>
>>What frequencies did you try in the second case?
>>
>>Yes, this problem is confusing, but I hope we will find a simple procedure to
>>get out of trouble.
>>
>
>
>In both cases, I went as high as 54 MHz. In the first case (adjust ticks), that
>led to a speed index of 1.74, but the clock was way too slow. Lowering the
>speed to 41 MHz (speed index 1.31) led to the clock being slow by a much smaller
>amount, but I KNOW the Prism can do better than 41 MHz, so I went back to
>UNchecking "adjust ticks". At 54 MHz, that gave a speed index of 1.24. The
>clock was perfect.
You ***BELIEVE*** your Prism can run higher than 41MHz.
>I thus conclude that I should NOT check "adjust ticks" if I want an accurate
>clock, and that 54 MHz is then OK on my Prism but gives a speed index of only
>1.24.
Now what you need to do is to check solution times.
It would be funny to discover that your Prism solves problem faster at 40MHz
than at 54MHz, isn't it? Well that's what happen because of the known PLL
problem.
Check this with Chess Tiger:
1) set the speed to whatever first setting you want with AfterBurner
2) execute the Speed test in CT
3) load a position and set the level to "ply depth"
4) measure the time CT takes to finish the given plies
5) set the speed to whatever second setting you want
6) execute the Speed test in CT
7) load the SAME position and set the level to "ply depth" (it is important to
reload the position, or else CT will keep track of the previous search in its
hash table)
8) measure the time CT takes to finish the given plies
Christophe
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.