Author: Vincent Diepeveen
Date: 02:00:23 10/17/01
Go up one level in this thread
On October 17, 2001 at 04:11:21, Jouni Uski wrote: >... AMD K6-2 450Mhz and Athlon 1200Mhz for SSDF tested programs? So I can check >my calculations for speed-up gain. I hope exact comparison in 4 - 10 middle game >positions for each program. For diep it's not so hard to do. K6-2 was a horrible processor as the L2 cache was busspeed. Now diep is not a small program so doesn't fit in L1 cache unlike some small assembly programs. Also DIEP is a 32 bits program not a 8 bits program and only 8 bits programs could profit from some other feature of K6 namely that 8 bits is 2 times faster nearly on K6 than 32 bits. Speed difference between K6-450 and PII-450 was real huge. Around 30%. PIII versus PII is around 17.3% faster K7 versus PIII is 7% faster (athlon XP or palomino add another small %). So besides the speed difference we have: (1200 / 450) x 1.3 x 1.173 x 1.07 = 4.35x faster We still didn't talk about the extra savings because of hashtable. Nowadays no computer has less than like 256MB ram or something. I remember K6-450 most has perhaps 32MB ram or something. Perhaps 64mb in sweden. Just booting win2000 on that would mean already that you could get like a few megabytes of hashtable at most. Now you can easily get a 150MB hashtable. that's at least another factor of 2 added. So like 8 times faster it is *just like that*. No clue how much RAM the new machines have there, but it's a big step forward. The K7 is a true 32 bits processor with a fast L2 cache. Of course still 8 bits is faster, but that has to do with other things like it is a more compact format whereas the K6 was horrible for 32 bits especially in combination with the slow L2 cache. >Jouni
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.