amd .vs intel....
tonynelson at georgeanelson.com
Thu Dec 15 03:41:57 UTC 2005
At 6:35 PM -0600 12/13/05, Jeff Vian wrote:
>On Tue, 2005-12-13 at 11:04 -0500, Tony Nelson wrote:
>> AMD started that fight in the first place, and had to switch to PR
>Chose to utilize performance as the comparison since theirs was
>obviously able to do more at a lower clock speed.
No, AMD wanted to sell their current customers the new chip, but it ran at
a lower clock rate than the old chip, so it looked slower. AMD had trained
them to look only at clock rate. AMD had to fix that in order to sell the
chip at all.
>> (Performance Rating, Public Relations, you pick) when their next faster
>> chips got more performance at an initially lower clock rate. Intel hasn't
>> said a word about AMD PR, because AMD PR compares performance to an
>> original Athlon, not any Intel chip.
>AFAIK it compares to a P4 at the stated speed.
What you know is wrong. I'm sure that AMD doesn't mind your
misunderstanding -- it sold lots of AMD CPUs.
The P4 is so different from AMD's stuff (or even Intel's other stuff) that
the clock speed says almost nothing about how a particular program will
perform. The P4 stepping is more important than the clock speed. (Need
integer multiplies? Shifts? Don't use the earlier versions.)
>> Intel chose the P4 architecture over several alternatives because 1) the
>> architect thought it had legs and 2) it would always trounce AMD in the
>> clock speed race. Intel left the P4 (and the architect left Intel) when
>> the P4 ran out of legs
>I don't think that is fair to Intel. The 32bit processor for all makers
>has reached near EOL. Intel (and AMD) are focusing on their new 64bit
>processors which is a natural replacement for the P4.
What's "fair" is not at issue. Intel pulled the plug on P4 because it
wouldn't go faster anymore. Wider would have made it harder, but it
already wouldn't go faster.
>Clock speed actually means nothing for performance comparisons. I
>believe mips is a much better measure of performance. ...
Ah, yes, the Meaningless Indicator of Performance. MIPS goes up as more
integer pipelines are added, but actual throughput does not, due to the
lack of discoverable parallelism.
>The facts are that AMD has better performance at lower speed and takes
>advantage of that as a marketing tool. It takes PR to let people know
>that clock speed is not a valid measurement. You have to compare
>athalon to athalon or P4 to P4 for clock speed to be valid. Performance
>is the only valid measurement for comparison when comparing different
It took extra PR (Public Relations) after AMD spent a few years convincing
people that clock speed was king.
>NOTE: I am an AMD user, but I think arguments should be fair and factual
>not biased or based on innuendo.
Indeed. I am an AMD user myself for a few years now. I based my post on
following comp.arch for about 5 years. Feel free to research this
yourself, and supply facts.
TonyN.:' <mailto:tonynelson at georgeanelson.com>
More information about the fedora-list