|
Re: Measurement of CPU usage
de PaulFXH 05/07/2006 04:42
Al Klein escreveu:
> On 6 May 2006 10:50:26 -0700, "PaulFXH" <paul_hackett2@yahoo.com.br>
> wrote:
>
> >However, I'm NOT looking for a means to see what my CPU usage is--I
> >already have about 6 utilities showing me this.
> >What I want is an explanation as to why these utilities provide such
> >startlingly different results.
> >Which one is right?
> >Or, are they both right with different interpretations?
>
Hi Al
Thanks for your reply.
You make some very interesting points on which I have commented below.
> CPU usage is always 100% - the CPU doesn't (usually) stop. The
> question is which program is using what percentage of the CPU's time.
>
Yes, this is my understanding too. Indeed, the two utilities that gave
the lower results in my test (Process Explorer and WinTop both gave
around 3.5%) do assume 100% total CPU usage with what's not being used
by running processes being attributed to System Idle Processes.
> If you're looking at a program that claims to show actual CPU usage,
> unless it's claiming 100% it's wrong.
This would force you to conclude that the CPU Usage results from
FreeMeter and ACPU are both just WRONG.
However, it's not just these two utilities that show these alarmingly
"high" CPU usage numbers (after discounting System Idle usage). Others
that give results in the same category (on my computer anyway) are
Everest, PCWizard and SysMon1.22.
I find it hard to accept that all of the utilities are just churning
out nonsense results, particularly, when they all give nonsense results
that are practically the same (in tests where I have run many of these
utilitie simultaneously).
It seems to me that these two categories of CPU Usage results derive
from different methods of measurements, or something................
However, a more precise explanation seems to be incredibly elusive.
>
> If you're looking at a program that claims to show the percentage of
> CPU time used by each program, be aware that usage is dynamic - it
> varies microsecond by microsecond. While the system idle process (the
> loop in the OS that waits for something to happen) usually gets most
> of the CPU time, it gives its time up whenever another program needs a
> slice. Unless both programs are reporting the same thing at the same
> time (and that's impossible, due to the nature of digital computers),
> they're going to give you different results. A single program will
> give you different results moment by moment.
I agree but the differences I'm seeing in the CPU Usage numbers
(discounting the System Idle time) is absolutely enormous---3.5% vs 64%
in the test I carried out earlier today.
I don't think anybody can claim this is due to differences in
measurement times.
This was an interesting post but it seems there's still some way to go
before this strange phenomenom gets explained.
Thanks
Paul