Author: Lin Harper
Date: 16:10:39 11/27/00
Go up one level in this thread
On November 27, 2000 at 17:51:00, Robert Hyatt wrote:
>On November 27, 2000 at 15:52:04, Uri Blass wrote:
>
>>I found in my engine-engine games that the number of nodes per second was too
>>low and I suspected that something is wrong with it so I decided to download
>>Fritz6light as the only active engine.
>>
>>Hiarcs7.32 was the only active program and Fritz6light was the only active
>>engine(32 mbytes hash tables).
>>
>>I found the same behaviour so I decided to check the numbers of nodes and the
>>times in the position after 1.a3 a6 2.e4
>>
>>I got the following results on my PIII450:
>>20 seconds 4737 knodes
>>56 seconds 13024 knodes
>>
>>This means only slightly more than 200 knodes per scond and it seems too low
>
>because I know that the normal knodes per second is bigger.
>>
>>I decided to restart my computer and after doing it I found the following
>>results at the same conditions(Hiarcs7.32 is the only active program and
>>Fritz6light is the only active engine with 32 mbytes hash tables).
>>
>>14 seconds 4737 knodes.
>>41 seconds 13024 knodes.
>>
>>1)Can someone explain the reason that my computer became faster after
>>restarting?
>>
>>2)Is there a way to know that my computer is slower without doing a test like
>>this test.
>>
>>I want to know if it is slower during an engine-engine game without giving it
>>the position after 1.a3 a6 2.e4 and checking the times.
>>
>>I can only suspect by looking at the number of the nodes per second because it
>>is possible that the number is smaller because of different position.
>>
>>Uri
>
>
>The most common reason is a problem with how the problem is laid out in
>memory. Since memory is direct-mapped into sets in cache, it is possible
>to get a memory layout where many memory pages map to the same cache set.
>Overloading that set and under-utilizing other sets. This is a tough
>problem to overcome. There is a memory layout algorithm that is often
>called "page coloring" that prevents this. I don't know whether windows
>uses it or not...
>
>This reboot/test/reboot/test is a pretty common thing to see at computer
>chess events. :)
I don't mean to be negative here, but there is some monkey business going
on in the computer industry. I've heard reports from fairly reliable
sources that processors are actually being sent to some part of Asia where
they get restamped! ie, what it says on top of your processor is not correct.
I'm not suggesting this has happened in your case. This is the reason I
buy big name brands, that way I can be fairly sure of what I'm getting.
I got diddled some years ago by a shonky operator, did'nt find out about
it until about a year later when I went to get some extra RAM fitted, that
was back in the 486 era.
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.