Author: Vincent Diepeveen
Date: 08:01:05 11/18/02
Go up one level in this thread
On November 18, 2002 at 10:58:35, Robert Hyatt wrote: >On November 18, 2002 at 10:15:53, Vincent Diepeveen wrote: > >>On November 18, 2002 at 01:28:17, Terry McCracken wrote: >> >>>On November 17, 2002 at 13:58:43, Vincent Diepeveen wrote: >>> >>>>Hello, >>>> >>>>If i read the below section of the intel programming guide very well, >>>>then i interpret that as: >>>> >>>> a) only the Xeon MP has hyperthreading and not the normal P4s, >>>> not even P4s sold as Northwoods. >>>> >>>> b) even the Xeon MP you should not assume that it supports hyper >>>> threading. In fact something that gets sold as Xeon MP you should >>>> still not assume that it supports hyper threading, but instead >>>> one must assume that only after specifically questionning the cpu >>>> whether it can do it. >>>> >>>>Intel IA-32 programming manual. Page 247 (of 780): >>>>" >>>>7.6.1 Implementation of Hyper-Threading Technology in IA-32 Processors >>>> >>>>Intel's Hyper-threading Technology was introduced into the IA-32 architecture >>>>in the Intel Xeon processor MP and in the later steppings of the Intel Xeon >>>>processor. Both of these processors implement Hyper-Threading Technology with 2 >>>>logical processors per physical processor. When reading the following sections >>>>(and other sections in this book that refer to this technology), assume that all >>>>references to the implementation of Hyper-Threading Technology in the Intel Xeon >>>>processor MP also apply to those later steppings of the Intel Xeon processor >>>>that implement Hyper-Threading Technology. >>>> >>>>Intel recommends that software not rely on IA-32 processor names to determine >>>>whether a processor supports Hyper-Threading Technology; but instead, it should >>>>use the CPUID instruction as described in Section 7.6.5, "Detecting >>>>Hyper-Threading Technology". >>>>" >>>> >>>>Vincent >>> >>>Careful, Intel sues! >> >>Not at all. They now claim only the > 3Ghz P4s will have hyperthreading >>enabled, but that's an easy claim, because i cannot check it as the >>fastest P4 i can test now is 2.53Ghz. With some luck soon also 2.8Ghz P4. >> >>At the time the > 3 Ghz P4s are there within say a 6 months or so, >>of course those statements will have been forgotten. >> >>The only official statement is that the group of processors which >>*might* support SMT is the Xeon MPs and only >>after you ask the cpu a CPUID and when that response is positive >>*then* it supports it. >> >>So they give ZERO garantuee simply yet. >> >>Best regards, >>Vincent > > >It isn't only the xeon MT. The MTs are certified for 4-way and 8-way, but the >other xeons are also doing hyperthreading. The 2.8's in particular. This is >from experience, not from guessing, btw... Not a single P4 i tested so far with the 'smt' turned on is doing it Bob, of course i do not do a CPUID in DIEP to check whether the chip is supporting it. The statement from intel is very clear. *nowhere* they garantuee it works and at the Xeon MP it might work after you do a CPUID. The results posted by Nalimov here are all from dual Xeon MPs. Best regards, Vincent
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.