Author: Robert Hyatt
Date: 10:23:07 12/17/02
Go up one level in this thread
On December 17, 2002 at 12:14:16, Vincent Diepeveen wrote: >On December 17, 2002 at 12:03:38, Matt Taylor wrote: > >Matt, dr hyatt hasn't showed *any* data regarding 2 cpu's without HT >versus 4 CPU's with HT. I certainly _did_ post such data. And I explained why it is a pain to do since I have to reboot to disable SMT. Once more, in typical fashion, you make a statement that is simply 100% _wrong_. But in any case, I posted the one/two thread (no SMT) plus three/four threads (SMT on) numbers _again_ so that you can say you have never seen them one more time... > >He has only showed you something from which he admits he already knew >in advance that he was producing intel friendly results. What??? > >He admits somewhere else in this thread that he KNEW that running >2 threads at HT turned on at his dual Xeon, that this would run worse >than 2 threads without HT because of the 33% chance it would go wrong. Certainly. Everyone with half a brain knows this (present company excluded). It has been a topin on the linux-smp kernel users group. Intel has more than one paper posted on their web site describing what an operating system needs to do to avoid the above problem. We have discussed it here at length. I really don't know what kind a vacuum you live in, but you need to get out, look around, and wake up... > >So he knew then that his test could be intel friendly and that the >real test, which he never performed, wasn't done at all. I performed the _real_ test. one and two threads with SMT off. three and four with SMT on. I _know_ better than to run two threads only with SMT on until the linux kernel guys (or myself) fix this problem. Windows .net knows how to solve this... > >So much for 'objective testing' from Hyatt. SO much for information from "crapola diepeveen"... > >Best regards, >Vincent > >>On December 17, 2002 at 11:58:21, Gian-Carlo Pascutto wrote: >> >>>On December 17, 2002 at 11:27:18, Matt Taylor wrote: >>> >>>>Crafty gets better results with HT, >>> >>>In addition to what Vincent said, the data we currently >>>have is saying exactly the opposite. >>> >>>So I'd like to know what you based that statement on. >>> >>>Magic 8-ball? >> >>Actually I based it on data that Dr. Hyatt posted previously. The data Vincent >>has for his program doesn't show such wonderful gains. That's nice. It's not >>representative of everything, and it's not even conclusive. Truth is, Vincent's >>data doesn't show jack. You need multiple tests and multiple machines to show >>anything conclusively. >> >>>>but it's been optimized for HT. >>> >>>It's not - even Robert will tell you this. >> >>Ok, it's been optimized for Pentium 4, which is -almost- the same thing. If it >>runs well on P4, it runs well with HT because it will fall into an I/O burst >>cycle. He has stated a number of times that his interests are in the P4 because >>he has one on his desk. Eugene's interests are obviously in the P4. It's not >>rocket science to figure out what that means. >> >>-Matt
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.