Fwd: [RC5] Why does Pentium 3 process faster than Pentium 4?

Julian Ruhe ruheejih at calvados.zrz.tu-berlin.de
Wed Dec 11 11:45:02 EST 2002


> The problem is that SPEC is only a benchmark for how well a certain 
> processor (not system, just processor) performs when executing code 
> generated by a particular compiler. Just like dnetc (with hand-tuned 
> cores) only tells you how well a processor (not system) performs when 
> using hand-tuned assembler routines. Neither gives you the full picture.

The SPEC link I gave is a pretty good example for unoptimised code, 
because both systems are roughly comparabe: Same compiler, same 
 clockspeed, same memory bandwidth.  Unoptimsed code is the most common 
case and could show that the G4 is more effictive and faster. I does not...

> n general, a complete system contains a mixture of the two (low-level 
> and/or speed critical routines offered by the OS are most likely very 
> optimized and hand-tuned, while the higher levels - such as 
> applications - are most likely just generated by a compiler) and a lot 
> of other factors beside cpu speed come into play (graphics card, hard 
> disk, amount of memory, bus speed, caches, ...).

What makes me angry is, that there are still so many people that claim 
the G4 was the better CPU compared to today's x86 CPU. Except RC5-64 
there is no single benchmark that could prove this theory. All I see is 
G4 losing in general code and and highly hand optimised/Altivec code 
(www.netlib.org/atlas e.g.). So why do people say that, I ask myself. 
Are these nostalgic feelings or compassion...I don't know. To me the G4 
is overrated and outdated since years.

> Anyway, I think the discussion of "which processor is best" (or which 
> system is best) has been fought out already so many times all over the 
> web and usenet (and probably still is being fought out :) that we 
> really shouldn't rehash it yet another time on this list...

Ok, I am new here and here to concentrate on a new RC5 core. I don't 
want to start a new discussion :-)




To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest

More information about the rc5 mailing list