May
01 2004:
The end of the Megahertz Myth is definitely upon us.
Volumes have already been written about the infamous Megahertz Myth,
which in short is the propagation of misinformation that computing power
can be judged by Clock Speed alone. Of course the main beneficiary of
the Myth has been Intel, especially with their PIV architecture, which
requires far higher Clockspeeds due to its Longer Instruction pipeline
than the competing chips from both AMD and Apple. I have already covered
the mechanics of how IPC (instructions per clock), and Clock Frequency
both play a part in how efficient or powerful a CPU is at completing
given tasks, and that a CPU running at 3.0 + is not necessarily going
to be any faster than a CPU running at 2.0Ghz. For those that have missed
the earlier articles, you can view them here.
Now before I get into why we can finally bury the fodder once and for
all, first I’ll cover a bit of History in regards to the Myth
itself. The term was first coined by Apple after being stung by criticism
that its PowerPC processor family was slower than Intel's rival Pentium
line. With the G4’s inability to scale as quickly as both AMD’s
and Intel’s X86 variants of the time, the Clock Frequency divide
was growing wider and wider, but the performance delta was not proportional
to the growing discrepancy. This is something that Apple was desperately
trying to get across to the general public, who were more focussed on
the race to the magic 1 GHZ mark that was being staged by the X86 clans.
Now while Intel and AMD shared very similar CPU architectures, the variance
in the competing chips was marginal, so any gain in Mhz, was classed
as a major advantage. The general public was pounded senseless with
the MHZ message from both camps. But alas, Chipzilla was about to change
the rules…
Enter the now infamous Willamette PIV. The new chip
was based on Intel’s then recently developed Netburst architecture,
which due to its extended pipeline relied on Higher CPU cycles to complete
set instructions. This was not however passed onto the unsuspecting
public, who only saw the stampeding clock frequencies and applied that
theory to the past Mhz race. Of course now, not only Apple but also
AMD had a huge problem on their hands as the Intel Marketing Monkeys
milked the MHz Myth to their advantage. And Milk it they did. So what
changed? The introduction of the Banias (Centrino) line of Mobile CPU’s
which are based on the PIII-S architecture, with an extremely High IPC
(Instruction per clock), Huge L2 cache, added SSE2 instructions, and
other architectural improvements, has been the stumbling block. With
the performance of the CPU’s rivalling PIV’s at almost twice
the clockspeed, the cracks were well and truly starting to appear for
the beloved Myth.
|
With Intel’s Netburst architecture
stagnating in the 3.0Ghz range, coupled with the current Prescott
not scaling due to thermal issues, and we now have the situation where
Intel are in damage control, now admitting that CPU performance can
not be judged on clock frequency alone. No Shite Sherlock, Really?
Now to the more technically savvy, that is no great revelation, but
to the Joe public, that is going to be a tough ask to explain. It
is fascinating to see Intel, which for years has benefited from the
Megahertz Myth, suddenly put into the position of having to dispel
it. So what to do now ?
The End of the Megahertz Myth?
Intel has now introduced a new model numbering system that effectively
has eliminated raw clock speeds from its product names. This will
hold them in good stead for the next generation of Desktop CPU’s
those are based on the Banias/ Dothan Core and not on the Dead End
Netburst Megahertz Monsters. (More on that in a future article where
I’ll look into the crystal ball on the future CPU landscape.)
By introducing the model number scheme now, it’ll save them
trying to explain why the future CPU’s will be running a far
lower clock speeds than the Current PIV’s.
The modelling numbers go something like this. Performance: 700 Series.
Banias /Dothan / PIV Extreme Edition. Consumer: 500 Series: PIV Prescott/
Mobile PIV,
Low End. 300 Series: Celeron: The new modelling system implies that
you should be able to move easily from the lower to the top model
within a processor family with a certain discrete step. For instance,
Banias, Dothan and Prescott processors will differ from the next models
by an increment of 10. Celeron processors will be rated with a smaller
increment of 5, because the difference in their performance and clock
frequency is smaller. Now this may help Intel to disengage themselves
from the dreaded myth, and aid the hapless fools in the sales rooms
having to explain to mum and dad why a Centrino laptop at 2.0 Ghz
will make mince meat of a 3.0 Ghz PIV variant, but it isn’t
going to wash away the memory of the fodder they have been ramming
down our throats for the last 4 years or so. It will also be interesting
how AMD will respond, considering their model numbering system is
designed to reflect Intel’s MHZ ratings. – i.e. Athlon
64 3200 – 2GHZ.
Now, hopefully, this will end the Megahertz
Myth once and for all. That’s not to say the transition is going
to be easy or smooth. There is a lot of disinformation to be pried
from the conscience of the general buying public, but at least in
the tech community, we won’t have to put up with the Intel Marketing
Monkeys MHZ ranting any longer. No Fear, there will still be an endless
stream of fodder to plough thru.
Till next time :-)
|
