Author: Robert Hyatt
Date: 09:12:21 05/03/04
Go up one level in this thread
On May 03, 2004 at 12:03:48, Gian-Carlo Pascutto wrote: >On May 03, 2004 at 10:59:09, Robert Hyatt wrote: > >>If you take your numbers it is a 10% change. That is _still_ orders of >>magnitude from what he needed to explain the difference between his search and >>the DTS results... > >The hardware alone makes any direct comparison impossible, so I don't >don't see why it would be needed to explain the performance difference this >way - obviously a shared memory Cray is going to scale better than a cc-NUMA >cluster that is linked together with routers :) The hardware was _identical_ except for CPU speed. All of the quads have the same MB, same memory, same FSB speed. Only difference was your machine had 550mhz pentium III xeons, mine had 700mhz PIII xeons. There were no other differences that affected the test. My box has more (and faster) disks. The tests didn't use disks. > >Given that all data about DTS seems to have vanished from the earth and >the only thing we are left with is an ICCAJ article where the published >data is, well, "calculated in a rather funny and backwards manner", I >would simply not care for those results at all. Totally up to you. The main data was not "calculated in a rather funny and backward manner". (the speedup data). The nodes are simply not measurable either in Crafty or Cray Blitz, except at the precise end of an iteration. If you have a problem with it, so be it... BTW all the data hasn't "vanished". The original log for the 16 cpu game as it was played is around. The actual speedup calculations are around. Just not all the other test log files.. Of course I suppose I could have deleted that, plus all the old crafty versions, the electronic copy of my dissertation, all the log files from all the ACM events I ever played in, etc. I just can't imagine why I would have wanted to do so... > >-- >GCP
This page took 2.87 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.