CPU usage mystery

[PaulFXH]

Hi
I have already posted this question to two other groups without getting
anything like a clear explanation.
In addition, I have googled extensively for an answer with a similar
lack of success although I did get the impression that understanding
this matter may be somewhat trying for the average home computer user.

In any event, can anyone explain to me why there seem to be two
different interpretations of CPU usage?
Thus, while Process Explorer and WinTop (my OS is WinMe and I don't
have Task manager) typically indicate the CPU usage on my machine to be
in the 4-6% range, other utility programs suggest that it is
considerably higher, being typically in the 60-70% region. The others
include FreeMeter, SysMon and PCWizard.

My CPU is an Intel Pentium 4 1800MHz and does not support
HyperThreading.

I would be grateful if anyone can point me to a link that might help me
get to the bottom of this mystery.
TIA
Paul

 

[Clark]

I thought I remembered WinMe having a Taskman.exe program.

Anyway, have you run both the low usage and high usage programs at the same
time to see what they say? For all I know the high usage software is
actually making it high.

According to Task Manager on my WinXP machine, it is normally the lower
range unless something is going on. I have one program, that is a little
messed up, that drives it to 100% for several seconds at a time.

Try loading several busy work applications and see what the different
applications show.

Clark

 

[PaulFXH]

Clark escreveu:

I thought I remembered WinMe having a Taskman.exe program.

Anyway, have you run both the low usage and high usage programs at the same
time to see what they say? For all I know the high usage software is
actually making it high.

According to Task Manager on my WinXP machine, it is normally the lower
range unless something is going on. I have one program, that is a little
messed up, that drives it to 100% for several seconds at a time.

Try loading several busy work applications and see what the different
applications show.

Clark

Hi Clark
Thanks for your reply.
Yes, I have run up to five of these CPU monitoring utilities at the
same time with the same results that I already mentioned. That is, the
numbers for CPU Usage fall into two distinct categories; either in the
4-6% range or more than 10 times as high.
You are correct in suggesting that these utilities do add to the total
CPU usage but only in the region of 0.5-1.0% per utility. So, this does
not explain why some numbers are dramatically higher than others.
You are also correct in surmising that WinMe has a Task Manager.
However, it is a considerably watered-down version of that in WinXP and
does not show CPU usage. For that reason, I use WinTop for this
purpose.
As I have already noticed the exact same phenomenon on my other machine
running on WinXP, I would be astonished if no-one else has experienced
this odd behaviour.

Paul

 

[Mike Walsh]

Since Win95 Sysmon has at times shown much higher CPU usage than Wintop. I have always had more confidence in Wintop's CPU numbers that Sysmon. It has been my experience with WinNT / WinXP that when Perfmon shows high CPU usage it is because an application is actually hogging CPU cycles.

 

[PaulFXH]

Mike Walsh escreveu:

Since Win95 Sysmon has at times shown much higher CPU usage than Wintop. I have always had more confidence in Wintop's CPU numbers that Sysmon. It has been my experience with WinNT / WinXP that when Perfmon shows high CPU usage it is because an application is actually hogging CPU cycles.

Hi Mike
Thanks for your comments.
Actually, in my situation, SYSMON, not just at times, but ALWAYS
indicates dramatically higher CPU usage numbers (and also PCWizard and
FreeMeter) than do WinTop or Process Explorer.
I've just carried out the test suggested earlier by Clark where I ran
all five of the above utilities while carrying out a virus scan.
Interestingly, SYSMON/PCWIZARD/FreeMeter all indicated 100% CPU usage
throughout the whole period of the test. Not once did any of these
indications dip below 100%.
The other two utilities (WinTop/Process Explorer) while still
exhibiting high usage numbers (95-100%), did show usage figures of as
low as 75% for short (20-30s) periods during the scan.
When the scan ended and the scanner was turned off, WinTop/Process
Explorer returned immediately to their normal indication range of 3-5%
while the other three dropped only to about 90%.
After the computer was rebooted, however, the CPU usage indication for
FreeMeter/PCWizard/SYSMON dropped to its more customary 60-65%.

Does this elucidate matters for anyone?
TIA
Paul

 

[Paul]

Clark escreveu:


Hi Clark
Thanks for your reply.
Yes, I have run up to five of these CPU monitoring utilities at the
same time with the same results that I already mentioned. That is, the
numbers for CPU Usage fall into two distinct categories; either in the
4-6% range or more than 10 times as high.
You are correct in suggesting that these utilities do add to the total
CPU usage but only in the region of 0.5-1.0% per utility. So, this does
not explain why some numbers are dramatically higher than others.
You are also correct in surmising that WinMe has a Task Manager.
However, it is a considerably watered-down version of that in WinXP and
does not show CPU usage. For that reason, I use WinTop for this
purpose.
As I have already noticed the exact same phenomenon on my other machine
running on WinXP, I would be astonished if no-one else has experienced
this odd behaviour.

Paul

I tried a search like this one:

http://www.altavista.com/web/results?q=idle+task+winme

The System Idle task is explained here a bit.

http://forum.osnn.net/showthread.php?p=710528

And the metrics generated may not always be accurate...

"Task Manager may not calculate CPU usage correctly on Windows 2000..."
http://support.microsoft.com/kb/873289/en-us

I'm sure each MS OS does something different, and I bet there
aren't nearly as many articles discussing the foibles of WinME.

The real question would be how the performance monitoring tools
are obtaining their information. If they consult the system, and
a performance counter controlled by a System Idle task, then they
would not "pollute" the measurement, except when they update the
screen. If a monitoring tool inserts its own version of the idle
task, sets the priority low, then it can function as the idle task,
but the real System Idle task will now reflect how much of the
system is left after the utility has done its bit. The utility
might also have the side effect, of not allowing the system to
enter a lower power state while sitting in the desktop. There are
a lot of variables here to play with, and the quality of the
utility implementation could make a big difference to the
answers you get.

In any event, the designers of your system monitoring utilities,
may not have considered the test case of running a bunch of
them at the same time

Paul

 

[PaulFXH]

Paul escreveu:

I tried a search like this one:

http://www.altavista.com/web/results?q=idle+task+winme

The System Idle task is explained here a bit.

http://forum.osnn.net/showthread.php?p=710528

And the metrics generated may not always be accurate...

"Task Manager may not calculate CPU usage correctly on Windows 2000..."
http://support.microsoft.com/kb/873289/en-us

I'm sure each MS OS does something different, and I bet there
aren't nearly as many articles discussing the foibles of WinME.

The real question would be how the performance monitoring tools
are obtaining their information. If they consult the system, and
a performance counter controlled by a System Idle task, then they
would not "pollute" the measurement, except when they update the
screen. If a monitoring tool inserts its own version of the idle
task, sets the priority low, then it can function as the idle task,
but the real System Idle task will now reflect how much of the
system is left after the utility has done its bit. The utility
might also have the side effect, of not allowing the system to
enter a lower power state while sitting in the desktop. There are
a lot of variables here to play with, and the quality of the
utility implementation could make a big difference to the
answers you get.

In any event, the designers of your system monitoring utilities,
may not have considered the test case of running a bunch of
them at the same time

Paul

Hi Paul
Thanks for your informative post which I found very useful (even if I
didn't quite understand all of the System Idle Process explanation).
As I had already mentioned in my first post, this topic is not one than
lends itself to a ready explanation for amateur home computer users.
Nevertheless, even having read this stuff, it is still not clear to me
why there should be two quite distinct measurements entitled CPU Usage
on the one computer.
I should point out that, as far as I can tell, the numbers churned out
by the three "high cpu usage" utilities are all exactly the same as
indeed are those generated by the two "low cpu usage" programs.
I should also mention that I had understood from some MS literature
that System Idle cpu usage is actually exploited by various
low-priority housekeeping tasks rather than leaving that percentage of
the cpu availability with absolutely nothing to do as is implied in the
osnn.net thread.
I would welcome any further clarifying comments
TIA
Paul