Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: crafty faster on AMD however

Author: Robert Hyatt

Date: 20:38:34 09/27/02

Go up one level in this thread


On September 27, 2002 at 12:12:19, Vincent Diepeveen wrote:

>On September 26, 2002 at 11:16:06, Robert Hyatt wrote:
>
>>On September 26, 2002 at 08:35:19, Vincent Diepeveen wrote:
>>
>>>On September 26, 2002 at 08:22:14, Jorge Pichard wrote:
>>>
>>>>It is customary for intel to compare a higher clock CPU with a much lower clock
>>>>AMD CPU, for Instance, the latest P4 2.8 Ghz vs AMD XP 2200 Ghz. Sure they give
>>>>creidit to a better memory, but this type of comparison is like comparing Apples
>>>>and Oranges.
>>>>
>>>>http://www.pcworld.com/news/article/0,aid,104165,00.asp
>>>
>>>Crafty at AMD XP2600+ 2.133ghz Epox 8KHA+ Motherboard and CL2 ddr ram:
>>>  75.5 seconds base run time
>>>
>>>Crafty at 2.8Ghz P4 533Mhz bus and PC800-ECC RDRAM:
>>>  93.5 seconds base run time
>>
>>
>>
>>Yes...  But the Intel duals are blowing the AMD duals out of the
>>water, totally..
>>
>>AMD appears to win the "single cpu war" at the moment.  But on the
>>duals (and beyond) they are _way_ behind intel's performance.
>
>Speaking for DIEP, Fritz and many other programs which are tuned for
>AMD now, the AMD annihilates the intel things completely.
>
>Of course outdated programs do not run well on dual AMD.

You might find that "outdated" program _very_ difficult to beat, however.
That's all that counts...


>
>>
>>
>>>
>>>you can see the results yourself for amd:
>>>http://www.specbench.org/osg/cpu2000/results/res2002q3/cpu2000-20020812-01551.html
>>>
>>>for intel:
>>>http://www.specbench.org/osg/cpu2000/results/res2002q3/cpu2000-20020909-01639.html
>>>
>>>This is the *official* specbench mark. both manufacturers did their best
>>>to produce optimal versions of each product. Intel even uses its own
>>>compiler. Without this buggy compiler (for DIEP it is buggy, i do
>>>not know for others; it gets a lot of nps that compiler at intel
>>>processors but not giving correct evaluations and it is NOT a bug
>>>in diep, i found out in compiler what is the problem as posted
>>>before) they would be again hell slower.
>>>
>>>I am not sure whether Bob has verified whether that compile from intel
>>>is a bugfree compile; whether it plays as good when using big hashtables
>>>like a default compile of visual c++ or even latest gcc version.
>>
>>
>>I have to provide them with several test sets that must match my node counts
>>_exactly_ for them to "validate" their executable.  Therefore there is little
>>doubt that their compiler is working fine.  I only use the Intel compiler now
>>and it works perfectly.
>>
>>
>>
>>>
>>>Getting a zillion nodes a second doesn't say much about all nodes being
>>>non-random :)
>>>
>>>So reality is that the above result in reality is even more positive for
>>>AMD than it looks like. We simply cannot trust these intel c++ compiles.
>>
>>Sure you can.  I have tested the 6.0 release of their compiler exhaustively,
>>comparing various optimizations with a known good executable from gcc 2.95.2,
>>and the intel compiler is producing perfect code from a comparison of the
>>two...
>>
>>>
>>>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.