Re: How many times must Apple make such DRASTIC architecture changes?

[Paul Schlyter <pausch@NOSPAMsaaf.se>]

In article <lQw*bgOQq@news.chiark.greenend.org.uk>, 
timc@chiark.greenend.org.uk says...

>Clock speed isn't rising very fast now, but performance and clockspeed
>are only loosely related anyway.  At the bioinformatics applications I
>deal with in my day job, a 1.6 GHz Intel Itanium2 is roughly 50% faster
>than a 2.8 GHz Xeon, running the same suite of programs, and in both cases
>compiled with Intel's compiler.

...and if the run time had been exactly proportional to clock speed, it
ought to have run 75% faster.  And 50% faster isn't that far off....

I've had a piece of software which I wrote back in 1984 which performs
a precisely defined task which was and still is useful:  it outputs a
"calendar" for one calendar year which includes the rise and set time
for the Sun and the Moon, plus start and end times for twilight, as seen 
from one geographical place of your choce.  It was originally written in 
FORTRAN, and later ported to Pascal and then C, where it remains today.  
It's remained a command-line program all those years - after all, you just 
input a year, longitude, latitude, timezone, and output file name and then
the output is written to that file.

The very first version required approx 1 hour of CPU time on an Apple II 
with a Z80 softcard running Apple CP/M - the Z80 there ran at 2 MHz 
effective clock speed (4 MHz for half a microsecond, then halted for the
other half of that microsecond as the 6502 refreshed the Apple's RAM).

Today, the C version of that program runs in perhaps 1/3 second on a 1.8 GHz 
P4.

That's a speedup of 10,000 when the clock frequency was speeded up by a 
factor of 900.  So of that 10,000 times speedup, 900 can be said to be due
to the clock speed, and the remaining 11 to be due to other factors.

So your statement "performance and clockspeed are only loosely related"
is an understatement IMO.  But of course clock speed isn't everything.
And if the clock speed isn't dramatically different, a computer with the
slower clock can outperform another computer with a faster clock.  E.g.
an 80 MHz Cray-1 outperforming a 500 MHz P3 on problems which are suitable
for parallell computation.

-- 
Paul Schlyter,  Grev Turegatan 40,  SE-114 38 Stockholm,  SWEDEN
e-mail:  pausch at saaf dot se
WWW:     http://stjarnhimlen.se/