Author: Uri Blass
Date: 01:23:42 10/27/01
Go up one level in this thread
On October 27, 2001 at 04:01:45, Christophe Theron wrote: >On October 26, 2001 at 19:08:29, Thorsten Czub wrote: > >>On October 26, 2001 at 17:06:20, Christophe Theron wrote: >>>No it is just because the machine thinks that the time passes extremely slowly. >>> >>>It believes it can compute 10000 nodes per second because its seconds last much >>>much longer than a real second. >>> >>>That is why I ask you to check Chess Tiger's clock with an external stopwatch. >> >>ok. again. i am stopping with an external clock. >>using 30 mhz. >> >>ct makes xyz nodes . >> >>doing the same in the same time with 54 mhz. >> >>ct makes xyz + difference nodes. >> >> >>now : how can ct make more (difference) nodes when it is slower, >>or not faster ??? >> >>it seems you don't understand what i am doing. > > > >Say that at 30MHz CT can compute 1000 nodes per second (just an example). > >Say that at 30MHz, CT's clock is accurate (it follows closely your stopwatch). > >Now you set the device to 54MHz. But the PLL cannot reach 54MHZ. It reaches >32MHz and cannot go higher. > >BUT the device believes it is running at 54MHz. It will count one second every >time the internal clock ticks 54 million times. But in one REAL second, the >clock ticks only 32 million times. > >One second measured by the device is going to last almost two REAL seconds (the >time needed to count up to 54 million). > >During these two REAL seconds, Chess Tiger is going to compute something like >2000 positions. > >From his point of view, he believes that it can compute 2000 position per >second, hence the incredible performance index. > >In reality, it took almost two REAL seconds to compute those 2000 positions, so >you are still close to 1000 positions per second. > >There is no big gain because the PLL cannot exceed 32MHz, but if you believe the >performance index without checking it you are going to believe that the device >is twice as fast... > > > > Christophe I understand that thorsten is interested in the total number of nodes in a fixed time and not in the number of nodes per second that is displayed by the computer. He does not use the computer clock but an external clock when he trusts the total number of nodes that is displayed by the computer. If I understand correctly in your example 54 Mhz is really slightly faster than 30 Mhz but 54 Mhz is too much and you can use 32 Mhz to get the same total number of nodes when you use external clock. Uri
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.