High interrupt numbers

  • Thread starter Thread starter RichardM
  • Start date Start date
R

RichardM

Over the past few months my machine has been prone to going IRQ crazy that of
course chews up all the CPU. I am hoping that it is a software problem, but
so far by biggest problem has been how to track it. From the knowledge base
and many other places I found some code tricks but I think they were in C++
or something else of which I have no understanding. The trigger seems to be
high I/O, like when I am running Ghost or doing some other intensive I/O.
What has been equally weird is that it seems to go away on its own. It is
rather like if I walk away from the machine for some period of time, it may
be just fine. Taking a wild guess I am thinking of updating all the device
drivers have anything to do I/O including the chipset. But that road can have
dangers of its own. And I have been in the case checking cables that I am
ready to solder them on. Not that I have the expertise to do that, it is just
a disparate thought. Anybody how there had enough issues with IRQ's that they
have suggestion? Thanks for reading.
 
high cpu would not be
the result of irq's -
far from it.

if you want to track down
the high cpu, then you have
to analyze the processes utilizing
the cpu.

so one method you can utilize
is to open the task manager and
use the information it provides to
you.

there are additional selections
for analyzing the processes via
under the menu option "view".

there is a method that you
can use to pin point the process
hog:

try killing any and all processes
in the task manager that it will
allow you to.

when you kill explorer.exe simply
restart it via the applications tab
with in the task manager.
--

db·´¯`·...¸><)))º>
DatabaseBen, Retired Professional
- Systems Analyst
- Database Developer
- Accountancy
- Veteran of the Armed Forces
 
Excuse me, I managed to miscommunicate my issue in the earlier post. It is
the IRQ's that are driving me crazy. I am using sysinternals Process Explorer
to watch the numbers and in the column titled Process, you will see all sorts
of things listed besides EXE's. Interrupts also fall under that heading so
when the IRQ's go to sixty or eighty, that become part of the count for CPU
usage. And you are most correct about the lack of correlation between the CPU
and the IRQ's. The total CPU usage may be ninety or one-hundred but the IRQ's
will be something like eighty. The rest is made by the other programs which
are use two percent here and four percent there.

If I open up Task Manager in another window it comes up with a completely
different CPU usage count because it doesn't list interrupts anywhere. It has
taken a lot of experimentation but I think one drive is more prone than the
others to cause the IRQ storm. Right now it is a theory and I am gathering
evidence but this all takes me back to the question of shared IRQ's and
whether or not that leads to trouble. It also makes me question if there is
anything an end user can prevent this sort of thing. So, do you happen to
have any suggestions about IRQ's?
 
Er.. there is a correlation between IRQs (interrupt request) and the CPU.
IRQs "interrupt" the processor and then the interrupt handler uses the
processor to perform a task. Until they return, whatever process was
using the CPU is suspended.

I'd suspect hardware, though it could possibly be a bad driver. In your post in
the xp-general group you said it happens more during certain disk
operations. Could be a faulty controller. Have you checked the SMART
values on the drives?

80% CPU usage for IRQ I/O is definetly way off.
 
i know of process explorer,
but i cannot comment on its
accuracy or the purpose of
the data it provides.

it is all too easy to create an
association of varying types
of data and mixing apples and
oranges makes a nice fruit bowl.

-------------

however, i think that if a service
is running that makes a high
demand on a device as seen by
the data on the irq,

then it would be reasonable that the
particular service is also making a high
demand on the cpu.

for example, a service or program
designed to scan or index data
would be a constant burden on
both the cpu and the disk, if that
program is dysfunctional or
poorly configured or malware.

i believe that the best solution is
still killing away the processes running
in the task manager, one by one
until you find which one is
problematic.

however, before you do the above
you can eliminate some of the processes
by opening msconfig and temporarily
disable "all" your startups.

then go to the services tab
and "hide" microsoft services,
then disable what remains.

then reboot and analyze your
system again.

----------
because you disabled most
of the processes via msconfig,
the remains processes will
be evident in the task manager
and the process explorer.

--

db·´¯`·...¸><)))º>
DatabaseBen, Retired Professional
- Systems Analyst
- Database Developer
- Accountancy
- Veteran of the Armed Forces
 
Well, folks I am anxiously waiting to hear what you have to about this. I
did the msconfig and disabled all the startups. Then I hide the MS services
and disabled all the others. Then I picked a couple 18 gb files and moved
them to another drive. The interrupts immediately jumped into the nineties.
The interrupts have stayed in the seventies to a hundred. I will run some
more test with large files using different drives but I suspect the same
results. Being as I am grasping at straws, by any chance could the cables be
loose or something. Of course, there are five drives and so far it appears to
happen on any combination of drives.
 
Interrupts are thrown by hardware devices and low level drivers.

I'll ask again.. have you checked the SMART data on the drives?
HDTune will give you that information on the "Health" tab.
http://www.hdtune.com/

You might try this:
Disconnect all drives (including optical and USB drives) not required
to boot Windows and run your copy test again. That's not going to be
definitive since the problem could be with the drive/controller that you
boot from, but it's a start. If good, reconnect one by one, testing each time.

Failing that,, clone the boot/system drive to a drive on another port.
Disconnect the primary, set the BIOS, boot the clone, ... then test that.
 
The quick answer on SMART is that they are ok. I got that partially from
Norton SysDoc Monitor that suppose to check that on all the drives. And I
also run Western Digital diagnostics. Thinks got so crazy that I may not have
run it on all the drives so that will be repeated. I will also try to get
hdtune.

Your thoughts on disconnecting the drives are exactly on track with what I
have been considering; and trying not to think about because it is or can be
such a long, methodical task. Someone told me that there was a life that did
not have you always in proximity to a computer. What did they mean?
 
Yea but,, what fun would it be if we didn't have to open them up once in a while..

Although your tests do seem to indicate a problem with a hard drive or a HD
controller, there's always the possibility that it's something else. Mouse,
keyboard, sound/video/whatever card. (including any ports) Or it could be a chip
hardwired onto the board.. clock, a timer, or any of the various controllers.
(hopefully not) Some things would be easy to test, others not so easy.


I found the following in Process Explorer's help file. It's probably not much help,
but it is relevant:

<quote>
Interrupts and DPCs

On Windows NT-based systems Process Explorer shows two artificial processes:
Interrupts and DPCs. These processes reflect the amount of time the system
spends servicing hardware interrupts and Deferred Procedure Calls (DPCs),
respectively. High CPU consumption by these activities can indicate a hardware
problem or device driver bug. To see the total number of interrupts and DPCs
executed since the system booted add the Context Switch column. Another
sometimes useful metric is the number of interrupts and DPCs generated per
refresh interval, which you see when you add the CSwitches Delta column.

</quote
 
I was wondering if you got anywhere with this problem.

I have the same issue, HP computer with XP media center edition.

Recently started getting very slow did the check with process explorer and
found 70 - 90 % CPU usage for hardware interrupts.

I purchased a new drive and installed fresh and no problem. Put the old
drive in as secondary and still no problem, so my assumption is that it is a
driver bug on a non critical devcie, but I have not figured out how to
uninstall all the drivers. I know it can be done in the registry, I just
can't remember the right branch to cut off.......

As I recall, if you remove a section of the registry (after backing up of
course) all drivers are removed and y ou can go about doing a hardware
"discovery" for everything on the computer. Anywone know any more about
this????

Let me know how your original problem sorted out..
 
Back
Top