480% CPU?

Neil Youngman [ny at youngman.org.uk]

Mon, 22 Oct 2007 13:07:52 +0100

Would anyone care to hazard a guess why top is claiming that the top 3 process are between them using 480% CPU on a dual core box?

Neil

Top Back

Ben Okopnik [ben at linuxgazette.net]

Mon, 22 Oct 2007 13:09:27 -0400

On Mon, Oct 22, 2007 at 01:07:52PM +0100, Neil Youngman wrote:

> Would anyone care to hazard a guess why top is claiming that the top 3 process 
> are between them using 480% CPU on a dual core box?

In my experience, 'top' sometimes goes off into lulu land and claims silly crap. For some reason - I recall that someone explained it to me once, but I've lost the index to the brain cells containing the explanation - it has a problem with some processes (e.g. XFree86) which do some strange thing that 'top' can't cope with - so it just lies.

In other words, put not your complete trust in the reports of 'top'; 'tis a weak reed.

A prudent navigator will never limit himself to a single method,
particularly one requiring ... a device that is subject to mechanical
damage or loss.
 -- Nathaniel Bowditch, compiler of the first American compendium on
	navigation

It would be interesting to go through all the possible combinations of the three processes (if possible) to get a better handle on what triggers it, though.

Does the machine actually bog when you see this report, or is it just showing this stuff without any external effect?

-- 
* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://LinuxGazette.NET *

Top Back

Neil Youngman [Neil.Youngman at youngman.org.uk]

Tue, 23 Oct 2007 13:27:44 +0100

On Monday 22 October 2007 18:09, Ben Okopnik wrote:

> It would be interesting to go through all the possible combinations of
> the three processes (if possible) to get a better handle on what
> triggers it, though.

They're all running today and top is behaving perfectly normally. I did observe that pressing 'I' changed the output, so that 480% became more like 250%. This seems to be an "Irix mode" toggle and I have no idea why it should make any difference. Also I see little point in choosing between 2 different values of wrong, unless it offers any further clues.

> Does the machine actually bog when you see this report, or is it just
> showing this stuff without any external effect?

No, they were all 'nice 19' processes, BOINC clients and the like, and the system was coping just fine.

Neil

Top Back

Karl-Heinz Herrmann [khh at khherrmann.de]

Tue, 23 Oct 2007 23:26:53 +0200

On Tue, 23 Oct 2007 13:27:44 +0100 Neil Youngman <Neil.Youngman at youngman.org.uk> wrote:

> observe that pressing 'I' changed the output, so that 480% became
> more like 250%. This seems to be an "Irix mode" toggle and I have no
> idea why it should

Maybe you could also try to press "1" -- then all CPUs will be shown individually.

K.-H.

Top Back

Neil Youngman [Neil.Youngman at youngman.org.uk]

Wed, 24 Oct 2007 08:58:36 +0100

On Tuesday 23 October 2007 22:26, Karl-Heinz Herrmann wrote:

> On Tue, 23 Oct 2007 13:27:44 +0100
>
> Neil Youngman <Neil.Youngman at youngman.org.uk> wrote:
> > observe that pressing 'I' changed the output, so that 480% became
> > more like 250%. This seems to be an "Irix mode" toggle and I have no
> > idea why it should
>
> Maybe you could also try to press "1" -- then all CPUs will be shown
> individually.

Yeah, that showed individual CPU stats which were phone, but the process stats were still garbage.

Neil

Top Back

Jim Jackson [jj at franjam.org.uk]

Wed, 24 Oct 2007 13:21:15 +0100 (BST)

On Wed, 24 Oct 2007, Neil Youngman wrote:

> On Tuesday 23 October 2007 22:26, Karl-Heinz Herrmann wrote:
> > On Tue, 23 Oct 2007 13:27:44 +0100
> >
> > Neil Youngman <Neil.Youngman at youngman.org.uk> wrote:
> > > observe that pressing 'I' changed the output, so that 480% became
> > > more like 250%. This seems to be an "Irix mode" toggle and I have no
> > > idea why it should
> >
> > Maybe you could also try to press "1" -- then all CPUs will be shown
> > individually.
>
> Yeah, that showed individual CPU stats which were phone, but the process stats
> were still garbage.

I recall a thread on lkml prompted by bizarre TOP figures.

http://lkml.org/lkml/2007/10/3/123

this contribution seems to imply there were problems with dual core AMD64 see...

http://lkml.org/lkml/2007/10/3/276

but there were other issues around an actual kernel bug.

A change of kernel may get rid of the problem?

Top Back