[Ti] A real world comparison

[Chris Olson]

One of my engineers has an IBM ThinkPad T40 with an Intel Centrino 
processor running Windows XP Professional SP2.  We compared it to my 
15" Titanium PowerBook with a G4/1.0 by ripping and encoding a test CD 
with 4 audio tracks to 128 Kbps AAC in iTunes on both machines, and 
timed the task.  My PowerBook beats the ThinkPad on average of ~11 
seconds on that task, replicated three times.

I decided to really step back in time because I have an old eMachines 
T1090 desktop with an Intel Celeron 900 running Windows 2000 
Professional SP4 with AutoCAD.  So I did the same "benchmark" comparing 
to my first generation PowerBook G4/400.  The old PowerBook soundly 
trounced the Celeron - time to completion for the PowerBook was 4:01, 
time to completion on the Celeron was 4:16.  I only did that 
"benchmark" one time where we replicated it three times on the newer 
machines.

We decided to one more using ESRI's ArcExplorer rendering a Tiger point 
dataset in ESRI shapefile format.  This was done on the newer machines 
because the older generation machines are too slow for this one.  The 
PowerBook soundly kicked the ThinkPad's butt - time to completion 6:33. 
  The ThinkPad, 9:51.

Several general observations:
.)  iTunes' AAC encoder is heavily AltiVec optimized, but ESRI's 
ArcExplorer is not.

2.) iTunes may be more optimized for the Mac OS/PowerPC than for 
Windows/x86, but I believe ESRI ArcExplorer is more heavily optimized 
for Windows/x86

3.) The Intel cpu's may be faster on other tasks using optimized 
benchmarking software.

4.) We used real world tasks using real world software that's not 
optimized for benchmarks, and both the PowerBooks were faster, one of 
them clocking only 44% the clock speed of its opponent, and being a 
first generation G4 portable vs an Intel-powered desktop

I'm not convinced the Intel-based machines are even superior to the 
aging G4 on real world tasks.  Most every benchmark you see on the 
internet is either using optimized benchmarking software, one 
particular filter in Photoshop or other software, or doesn't compare 
real-world workflow tasks.  What they fail to tell you is the little 
details such as what we noticed on the older generation machines.   The 
old Power G4/400 stayed responsive and allowed me to do other tasks 
while encoding the CD to AAC.  The Celeron-based desktop wouldn't even 
allow me to resize an already open Firefox browser window while 
encoding the AAC files.  A PC that allows you to apply a Photoshop 
filter 3.2 seconds faster than a comparable G4, but doesn't allow you 
to do anything else because of inferior cpu threading or whatever, 
while that filter is being applied, ends up being the slower machine at 
the end of the work day.

Read between the lines.
--
Chris