Author: Robert Hyatt
Date: 13:16:48 10/01/02
Go up one level in this thread
On October 01, 2002 at 09:34:16, Vincent Diepeveen wrote: >On September 30, 2002 at 15:05:48, Robert Hyatt wrote: > >>On September 30, 2002 at 02:13:54, Aaron Gordon wrote: >> >>>On September 29, 2002 at 23:43:26, Robert Hyatt wrote: >>> >>>>1. I didn't see anyone post _any_ 1.7x number for AMD when I asked for them >>>>a week or two back. >>> >>>It was in a seperate post. Slate benched a number of the latest binaries I >>>compiled with automatic parallelization. >>> >>>>2. I don't see how the binary is going to affect this at all. You should >>>>get the same ratio of single to dual whether you use a fully-optimized binary >>>>or one with no optimizing at all. Since the dual speed is relative to the >>>>single cpu version, the base NPS is unimportant. >>> >>>Perhaps you should test this yourself if you can. Slate got 1.4x with your >>>binary, 1.7x with mine. >>> >>>>No, actually I am using a quad intel machine. Where are the quad AMDs? Why >>>>do you think there are none? Think about "scaling"... >>> >>>Clawhammers & Opterons will be out in a few months and there has already been >>>pictures posted of dual/quad Hammers. Also if I recall correctly >>>my single Athlon is faster than your Quad. >> >>I wouldn't argue that point. My quad 700 is getting around 1.6M nodes per >>second using the intel compiler. However, a quad itanium-2 shows a lot more >>promise, if raw speed is the issue. >> >>I'm more interested in a slower quad than a faster dual, because the 4 processor >>machine is more difficult to use efficiently, and that is what the parallel >>search is all about. > >I saved this great statement to harddisk :) > >> >>> :) My board + chip now days costs >>>$154 together. I'm sure it would still cost over $500 to build that quad you >>>have which is slower. :) The gap would be huge if you drop a 2600+ in here and >>>even more so with a 2600+ at 2.5GHz. A quad may 'sound' nice but if all the cpus >>>are slow then whats the point? >> >>There are some fast quads out. I've seen linux output from a quad 2.2 intel >>machine (xeon-based). There are plenty of 1.5-1.6ghz quads around, but the >>processors are not compatible with my older MB. > >It is known that intel tried to produce a few quads for experimental reasons, >but i do not know a single quad P4 which i can buy in a shop. Even if it is >a 'slow' 1.6ghz quad :) I found several. I just had a quad 1.6 machine shipped to me by mistake and sent it on to the right end user. I have _seen_ actual output from a quad 2.2 but I have not seen the machine advertised. > >Also i would be pretty amazed if a quad P4 runs stable for more than >5 minutes after booting, not to mention what happens if you write down >'mt 4' in crafty :) Why? Every quad intel box I have owned has run perfectly, reliably, and for extended periods of time. I have 9 quad 550 xeon boxes, my quad 700 xeon box, a quad 400 xeon box, and all were reliable from day 1. I also have an old quad p6-200 that _still_ runs perfectly. > >Best regards, >Vincent > >> >> >> >>> >>>>Except that I can buy a quad or 8-way P4 system, but not an AMD. And now >>>>they get left in the dust... Not cheap of course.. But not even doable with >>>>AMD. >>> >>>See above.
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.