[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ARSCLIST] Hyperthreading (was Software for Mac)
On Oct 16, 2008, at 10:25 PM, Steven C. Barr wrote:
----- Original Message ----- From: "Richard L. Hess"
<arclists@xxxxxxxxxxxxxxx>
If you're running hyperthreading look at the Windows Task Manager
(ctrl-alt-del or Start>Windows Security if you're remote
desktopping) and check how much activity is in each thread during
some of your intensive processes.
What I saw was one thread was running 50% and the other was
running 0%.
When I turned off hyperthreading, the one thread ran at 100% (or
close to it). That seemed to optimize the PC.
This is why I'm worried about migrating to multi-core processors
as it appears that while the processors may be more efficient in
their use of clock cycles, the individual cores don't have the raw
number of clock cycles as the 3-year-old 3+ GHz P4 chips.
I have been unimpressed by some Core 2 Duo machines that I have
used, and have not been able to trace down why.
If I can get a 3.2 GHz P4 machine from 2005 for $350 or so with an
XPPro license, I'm not going to experiment and spend $2000 or more
for a really top-of-the line quad core at this point. Did I
mention my wife is of Scottish ancestry and it has rubbed off?
At some point, if I could be sure that I could speed up audio
rendering by 2-3 x with a new machine, I'd certainly consider it.
Also, the Nikon scan software did not appear to use the multiple
threads and the display seemed about the same as with the audio
rendering so I've disabled hyperthreading on my two dedicated
machines, the 3.0 GHz audio machine and the "new" used 3.2 GHz
photo scanning machine.
For office work and multiple windows, the multi-core model appears
wonderful, but for single applications it appears that we may not
see the increase in speed that we've seen in the past.
For example, since 1984, here are the clock speeds of my primary PC
4.77 MHz, 66 MHz, 132 MHz (clock doubled in 66),
333 MHz, 2400 MHz, 3000 MHz
Photo work was done on the 66 MHz and up, audio was done starting
with the 333 MHz machine.
The 2400 to 3000 step was caused by diversity and moving functions
rather than for raw speed.
I want an increase like I saw from 333 to 2400 MHz or close to it
<smile>. Other than that the 8 Win XP machines here are 1.8 GHz
Sempron, rest P4s: 2.4, 2.8 (3), 3 (2), 3.2 GHz
For most applications, all of these are fine. The only long waits
are extensive audio rendering operations (and I'm working on
offloading some from one machine to another--the network is not
the bottleneck) and the kids think the Web is sometimes too slow,
but the two family 2.8 GHz machines are lower end than the rest.
The Sempron is my mail/web machine that I remote desktop into --
wish I had bought one of the used Dells instead (all are Dells
except the store-built Sempron).
I'm currently using a new (this spring) Core 2 processor, running
(MUCH TO
MY DISMAY, I might add!) Windows Vista...for the most part, I've
seen NO
increase in speed or capacity; I've been blaming Vista (IMHO, this
abomination
places Bill Gates under the class of "justifiable homicide!!"). But
it MAY be the
machine itself?
Steven C. Barr
Thanks Steven, that's why I asked Richard about the hyperthreading
before. I use a quad cored Mac (each processor is 2.6G) and when I'm
processing audio, I open a widget that displays processor usage. When
I run my restoration software (iZotope Rx) or any other intensive
editing processes, only one of the four processors is doing 90% (just
a visual estimate) of the work. If the one processor is "limiting
out" a second processor will momentarily run up to about 50%.The
other three are barely working (running the display and operating
system?). As a result, my software runs just a little faster than it
would on a single processor. I can't help but think about how much
faster downloading with managers that use "mirrors" to double the
download speed are. Perhaps splitting and rejoining an audio file as
it's processing is completed is more complicated than I realize. I'm
no computer expert to be sure but watching those three other
processors barely working does not make the waiting any easier!
So you're not alone Steven. I just hope that it's the typical long
time lag for either the firmware or the software to take advantage of
the multi-core machines.
Steve Koto