multitasking smoothness, what's he talking about?

[YKhan]

Okay, I realize this guy is just a commentator and I really shouldn't
pay attention to him, but what this guy is talking about seems almost
counter-intuitive. Here's the article:

http://weblog.infoworld.com/yager/2005/06/16.html

Okay, he's commenting about the relative smoothness of multitasking
Apple Macintosh on PPC vs. various x86 chips on Windows. He's saying
that it's very choppy on a single-threaded Pentium-M. Then he says it's
much better, but still a little choppy on dual-core Pentium-D or
dual-processor Xeon processors. He says its extremely smooth on
Hyperthreaded Pentium 4. And then most bafflingly, it's smoother still
on Athlon 64 and Opteron.

Okay assuming that Pentium-M, Athlon 64, Opteron, and even Pentium 4
without Hyperthreading are all single-threaded, then their multitasking
has to be done through the Windows task switcher, and they would all be
in the same boat. So why would Pentium-M be any worse than the others?
Why would Hyperthreading be smoother than dual-core or dual-processor,
when it's supposed to be the bastard child of the various
multithreading technologies? And why would single-threaded Athlon
64/Opteron be smoother than Hyperthreaded P4?

Yousuf Khan

 

[Robert Redelmeier]

http://weblog.infoworld.com/yager/2005/06/16.html

Okay, he's commenting about the relative smoothness
of multitasking Apple Macintosh on PPC vs. various x86
chips on Windows. He's saying that it's very choppy on a
single-threaded Pentium-M. Then he says it's much better, but
still a little choppy on dual-core Pentium-D or dual-processor
Xeon processors. He says its extremely smooth on Hyperthreaded
Pentium 4. And then most bafflingly, it's smoother still on
Athlon 64 and Opteron.

Okay assuming that Pentium-M, Athlon 64, Opteron, and even
Pentium 4 without Hyperthreading are all single-threaded,
then their multitasking has to be done through the Windows
task switcher, and they would all be in the same boat. So
why would Pentium-M be any worse than the others? Why would
Hyperthreading be smoother than dual-core or dual-processor,
when it's supposed to be the bastard child of the various
multithreading technologies? And why would single-threaded
Athlon 64/Opteron be smoother than Hyperthreaded P4?

Good questions, all. But my first question is:
What is this dude calling "smoothness"?

I suspect he is looking a focus-shift time -- the
time it takes to bring a background window to the foreground.
Or maybe the "smoothness" of that repaint.

His laptop Pentium-M almost certain sux at this.
It might even have UMA graphics!

From the order he gives, memory latency seems to
be the important factor.

-- Robert

 

[YKhan]

Good questions, all. But my first question is:
What is this dude calling "smoothness"?

I suspect he is looking a focus-shift time -- the
time it takes to bring a background window to the foreground.
Or maybe the "smoothness" of that repaint.

His laptop Pentium-M almost certain sux at this.
It might even have UMA graphics!

Okay yeah, that would make some sense, if he's using UMA graphics.

From the order he gives, memory latency seems to
be the important factor.

With the Athlon 64/Opterons having the built-in memory controllers,
there would be much less latency there than even a big cache
Hyperthreaded Xeon.

Yousuf Khan