Author: Vincent Diepeveen
Date: 02:34:22 10/24/02
Go up one level in this thread
On October 24, 2002 at 05:11:19, Jorge Pichard wrote:
>If you compare the architecture advantages of a Dual AMD MP such as: The
>operation per clock cycle, the floating point pipelines and the L1 Cache size
>you can immediately see why AMD is the King.
>http://www.amd.com/gb-uk/Processors/TechnicalResources/0,,30_182_865_4362,00.html
Most definitely. Interesting is:
AMD P3 P4
Full x86 decoders 3 1 1
Based upon this table the P4 would be by definition 3 times slower than
the AMD K7. Well it isn't!
Reason is that the trace cache can also deliver up to 3 iops a clock.
But that means that if the part of the program that gets planned to get
executed, that if it is NOT in the trace cache, that you have a major
problem as you can see.
DIEP is one of the many big programs that has such a bad big luck usually.
However the processor is not 3 times slower for me but 1.5 times slower
so i guess that the effective throughput in the P4 is a bit higher than
1 and in the K7 it is a bit lower than 2
Of course the 8KB L1 datacache is a big weak spot too.
Completely on paper seen, simple small programs should be way faster
at the P4 than big complex programs. Especially small programs fitting
within 12000 microops.
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.