Lets see something recent that claims a 3200 AMD chips outperforms a 3.0-800
P4 much less deserves it's inflated rating..
Some standard winmark benchmark will do that nicely, I think. 6%
faster, actually.
Unreal Tournament also runs faster, faster than the 3.2-800 as well,
FYI. But that is beside the point, as the 3.06 and 3.2 really does
have the edge on many modern apps and most modern benchmarks. Not much
of an edge, but the P4 _IS_ faster. ...on that.
Now for the $64000 question: How many P4-code-optimized apps do you
have, and rely on heavily? How many do you intend to buy before
replacing your current computer?
Because the P4 really sucks bigtime on oldfashioned 386/387 apps!
But Ok, games are the main performance concern, for most of us, and
the P4 does well on that.
http://www.anandtech.com/cpu/showdoc.html?i=1834
"The review community unanimously agreed that the processor was not
deserving of its 3200+ rating, but none were as infuriated by AMD's model
number than the folks at Intel.
That citation is part of an argumentative article that doesn't offer
any testresults that actually support the expressed opinions. Some
claims, in the article, are also quite off mark as well. Such as the
statement that the 'Northwood' core has meant that the P4 is more
competitive. That's largely nonsens. Northwood is only slightly more
efficient than the Willamette.
What has happened since the Northwood release, is that the benchmark
collections, used in comparisions, have changed. Changed to emphasize
streaming instructions, and have also been recompiled for P4 code
optimisation. Since the P4 is quite good at that, of course it shows
up better. Another thing that also has happend, is that Intel has got
into DDR RAM. With a vengeance, certainly. But that only has a real
impact on performance some of the time. Not always.
Religions have little to do with God and the real world. Benchmarks
are like that too. Have people all worked up and full of
righteousness. (Just lok at me now, for example. ;-))
- And have little to do with actual application performance or the
real world.
IMO. In the history of computing there has never been a computer/CPU
architecture that has been as consistently overrated by benchmarks, as
the P4. So I wouldn't have any respect at all for Intel's regards for
others benchmarks. That doesn't mean the P4 is always slow. But it's
performance on exclusively P4-code-optimized,
streaming-instructions-only type of app benchmarks, doesn't really
reflect on its relative performance in other circumstances.
I have been fooled twice. I have acquired two P4's. Both have been
dissapointments (I have Athlons to compare with). I got the second one
after reading benchmarks on extremetech and Toms hardwork. I swallowed
that Northwood number. Since then I've read benchmarks more carefully.
In particular I've noticed that extremetech and Toms hardwork have
gradually removed all benchmarks where the P4 made a poor showing.
Replacing them with others, recompiled by Intel for the P4. They have
made some argument why that is supposed to be 'fair'. I'll not get
into that. My point is that I and most others tend to gloss over such
details, and end up with a 'big picture' that is misleading in terms
of what kind of relative performance we can expect on our usual crop
of software.
Intels P4 seem always to "have the edge" with some testing crowds.
Problem is, about the same time the 3.06 "clearly had an edge" over
the 3000+ at extremetech. PCW threw a bunch of apps at a few systems.
The 3.06HT rated in behind the slowest participating Athlon, 2600+.
Higher end Athlons mostly being 20 -30% faster.
I don't have any experience with fast P4's, my fastest is a 2.4GHz.
But the above does match my experience of extremetechs 'benchmarks',
P4's and Athlons. I don't believe in HT, I don't believe in 800FSB, I
don't believe in extremetech, I don't believe in the P4 anymore. No
matter what some testers cook up. The P4 is lightningfast on some
things. The Athlon is more honest, it crunch code all the time.
I have a relative who works for Intel. He's biased of course - he
works with the Itanium - but he would vomit all over the P4- "folks at
Intel" and their benchmarks. and their claims of P4 performance.
ancra
Dual channel config.