[Ti] A real world comparison

Dennis Fazio dfz at mac.com
Thu Jun 9 16:08:08 PDT 2005


--On June 9, 2005 12:34:50 PM -0500 Shawn King <shawn at yourmaclife.com> wrote:

[...]
> And, before you lose your mind again, I'll be the first one to say, as I did
> on our show last night, that if Apple *doesn't* soundly beat Windows in
> those tests, it's over for Mac OS X.

That may not be possible, at least not "soundly." OSX will not significantly 
beat Longhorn (or vice-versa) in performance tests. It will have to beat it 
in overall functionality and user experience.

We went through this extensively in the '80s with supercomputers, when there 
were all kinds of competitive performance claims, by commercial vendors and 
by University computer architecture pretenders who invented many theoretical 
better-performing machines that had no chance of success at manufacturing in 
quantity.

Anyway, a precept that a sage in the industry always reminded us of was: 
"there ain't no magic."  By that, he meant that all of the manufacturers had 
the same kind of smart people with the same access to silicon fab technology 
and techniques, the same software engineering technology and expertise, etc. 
and nobody had any magic "thing" at their disposal that could make their 
product clearly faster than the competition.

Sure, each would perform better in specific applications, depending upon its 
architectural optimizations, but, by and large they were pretty equivalent 
within generations. You bought what did best in your set of applications 
whether it be large vector-intensive weather computations or highly scalar 
Monte-Carlo simulations.

So, the best Apple OSX and Microsoft Longhorn software engineers will be 
using the same software engineering techniques, compiler optimization 
techniques, etc, etc,  on nearly identical hardware and will likely come out 
with about the same performance overall.

Even between OSX on PowerPC and WinXP on Pentium, there is not today, and 
never really was in the past, nor would there be in the future (even if IBM 
produced super-duper G5's and G6's) much of an overall performance difference 
if you could average the entire gamut of popular applications.

So, when you say Apple has to beat Windows, I'm not sure what that means. The 
application binaries will run pretty much identically on both because it's 
pretty much the same binary. Their system calls may be a major part or a 
minor part of their overall process time, so that may change things a little. 
The file system managers should be pretty equivalent, memory management 
subroutines about the same, I/O drivers similar. The graphical interface 
should be pretty responsive for both and the vast majority of users probably 
won't be able to tell much of a difference in that area, or if they do, it 
won't matter enough to make them switch platforms or change their choice in a 
new purchase compared to the many other criteria that are much more 
important.

With a few exceptional cases that won't affect mainstream purchasing volumes, 
running applications under OSX on a Mac or under Longhorn on a PC should not 
be significantly different in performance for equivalent hardware. There 
ain't no magic.

-- 
Dennis Fazio
dfz at mac.com


More information about the Titanium mailing list