Sign in with
Sign up | Sign in
Your question

Thinkpad cut-out saga continues...

Last response: in Laptops & Notebooks
Share
Anonymous
a b D Laptop
March 3, 2005 1:02:33 AM

Archived from groups: comp.sys.laptops (More info?)

As I related in an earlier thread, my Thinkpad A21p (Win2k) has lately been
showing a disturbing tendency to occasionally suddenly cut out, typically when
I'm doing some graphics-related activity (Photopaint, Adobe Reader). However,
all the usual tests (PCDoctor, Sandra, Prime95, Memtest, DxDiag, etc) - many of
which I've since repeated - claim that everything is fine. (duh!) Lately I've
encountered the problem with standard bundled graphics app Paint, but in this
case it's definitely image-size dependent; everything is fine with smaller
images, but just trying to create a full-colour image a little over 6500 x 6000
in size inevitably causes a cutout. Unlike the situation with the other apps,
however, this one doesn't happen instantaneously. When I select to create a new,
very big image (not yet even saved in a file), with Task Manager running I can
see the memory allocation going steadily up, and at the same time there's a lot
of CPU activity as the kernel presumably backs stuff out to virtual memory.
After several seconds of doing this - clunk - the machine quits. It's a very
dependable fault, and only happens beyond a certain image size. Thinking this
was a simple problem with virtual memory, I wrote a small C command-line test
program that incrementally allocates more and more memory which it then tests in
a very simple way by writing to all of it then reading back from all of it. I
was sure that this would reveal the exact same problem with very large memory
allocations, but - guess what? - once the allocation gets up to the virtual
memory limit, a request for more is just denied - cleanly - and the test program
can then just sit there testing the memory it succeeded in getting without any
problem at all! Windows produces the expected "Virtual memory full" warning
pop-up, and even trying to open up a folder will then be refused - again cleanly
- until I stop the test program, after which the memory is all recovered as it
should be and operation returns to normal.

So what's happening with Paint that's not happening with my test program (or
vice-versa)? Could some problem with the video memory cause this, even though
the size of the on-screen window used by Paint (and not by my command-line
program) is not actually changing? (How do you specifically test video memory
anyway?)

Is the problem due to some post-SP4 hotfix that I've applied to Win2k?

Is it some hardware problem that's peculiar to my machine? Can hard drives or
chipset parts also cause sudden thermal overloads? PCDoctor indicates that
everything is OK with the motherboard heat sensors and the fan, but doesn't give
any values; is there any way I can see the actual sensor temps and the fan
speed? (The normally useful Motherboard Monitor doesn't seem to be able to read
anything other than the current CPU clock rate.)

Any further suggestions, anyone? At the moment I'm totally baffled....
Anonymous
a b D Laptop
March 3, 2005 1:02:34 AM

Archived from groups: comp.sys.laptops (More info?)

R. J. Sutherland <rj.sutherland.removethis@ntlworld.com> wrote in
news:tj9c21lke1mu6027g2c8kl6mf36heosbq1@4ax.com:

> Any further suggestions, anyone? At the moment I'm totally baffled....


Hi, I know this isn't the answer you're looking for, and I'm NOT trying to
be a troll, but MICROSOFT PAINT?!?!?!?!?!?

I can't imagine a worse program on the planet. If you are even .0009%
serious about doing ANY graphics related work, please, find a REAL program
that is built to handle graphics processing.
Anonymous
a b D Laptop
March 3, 2005 1:02:34 AM

Archived from groups: comp.sys.laptops (More info?)

R. J. Sutherland wrote:
> As I related in an earlier thread, my Thinkpad A21p (Win2k) has
> lately been showing a disturbing tendency to occasionally suddenly
> cut out, typically when I'm doing some graphics-related activity
> (Photopaint, Adobe Reader). However, all the usual tests (PCDoctor,
> Sandra, Prime95, Memtest, DxDiag, etc) - many of which I've since
> repeated - claim that everything is fine. (duh!) Lately I've
> encountered the problem with standard bundled graphics app Paint, but
> in this case it's definitely image-size dependent; everything is fine
> with smaller images, but just trying to create a full-colour image a
> little over 6500 x 6000 in size inevitably causes a cutout. Unlike
> the situation with the other apps, however, this one doesn't happen
> instantaneously. When I select to create a new, very big image (not
> yet even saved in a file), with Task Manager running I can see the
> memory allocation going steadily up, and at the same time there's a
> lot of CPU activity as the kernel presumably backs stuff out to
> virtual memory. After several seconds of doing this - clunk - the
> machine quits. It's a very dependable fault, and only happens beyond
> a certain image size. Thinking this was a simple problem with virtual
> memory, I wrote a small C command-line test program that
> incrementally allocates more and more memory which it then tests in a
> very simple way by writing to all of it then reading back from all of
> it. I was sure that this would reveal the exact same problem with
> very large memory allocations, but - guess what? - once the
> allocation gets up to the virtual memory limit, a request for more is
> just denied - cleanly - and the test program can then just sit there
> testing the memory it succeeded in getting without any problem at
> all! Windows produces the expected "Virtual memory full" warning
> pop-up, and even trying to open up a folder will then be refused -
> again cleanly - until I stop the test program, after which the memory
> is all recovered as it should be and operation returns to normal.
>
> So what's happening with Paint that's not happening with my test
> program (or vice-versa)? Could some problem with the video memory
> cause this, even though the size of the on-screen window used by
> Paint (and not by my command-line program) is not actually changing?
> (How do you specifically test video memory anyway?)
>
> Is the problem due to some post-SP4 hotfix that I've applied to Win2k?
>
> Is it some hardware problem that's peculiar to my machine? Can hard
> drives or chipset parts also cause sudden thermal overloads? PCDoctor
> indicates that everything is OK with the motherboard heat sensors and
> the fan, but doesn't give any values; is there any way I can see the
> actual sensor temps and the fan speed? (The normally useful
> Motherboard Monitor doesn't seem to be able to read anything other
> than the current CPU clock rate.)
>
> Any further suggestions, anyone? At the moment I'm totally baffled....

Google hmonitor.exe or Hardware Sensors Monitor. If the laptop has a
temp sensor, it might work where MM doesn't since it is more
laptop-centric. You might try booting to safe mode (Win2K has it?) or
command prompt and delete the page file in the event that it is
corrupted. Run the Win2K equivalent of chkdsk /f (might be the same),
and memtest86.exe. Does Win2K have the Event Viewer? That might offer
a clue if it does.

Sorry if you've done all these already.

Q
Related resources
Anonymous
a b D Laptop
March 3, 2005 1:02:34 AM

Archived from groups: comp.sys.laptops (More info?)

R. J. Sutherland <rj.sutherland.removethis@ntlworld.com> wrote:
>When I select to create a new,
>very big image (not yet even saved in a file), with Task Manager running I can
>see the memory allocation going steadily up, and at the same time there's a lot
>of CPU activity as the kernel presumably backs stuff out to virtual memory.
>After several seconds of doing this - clunk - the machine quits.

At what memory allocation does it quit? I have to wonder if the other
programs bother to check for memory allocation failure, which probably
returns a null pointer, and go ahead and use the 'allocated' memory at
the place the null pointer is pointing (oops!)...

Also, I forget if you had done a manufacturer's diag on your hard
disk, and a full scandisk...

Is Windows automatically handling your virtual memory?
Anonymous
a b D Laptop
March 3, 2005 1:15:29 AM

Archived from groups: comp.sys.laptops (More info?)

In "Re: Thinkpad cut-out saga continues...", Good Man <heyho@letsgo.com> wrote:

>Hi, I know this isn't the answer you're looking for, and I'm NOT trying to
>be a troll, but MICROSOFT PAINT?!?!?!?!?!?
>
>I can't imagine a worse program on the planet. If you are even .0009%
>serious about doing ANY graphics related work, please, find a REAL program
>that is built to handle graphics processing.

Please, I'm trying to find a handle on this problem with my Thinkpad, not engage
in a discussion in the relative merits of various graphics applications. At this
point, I'll take any reliable evidence that helps me sort this out!
Anonymous
a b D Laptop
March 3, 2005 1:15:30 AM

Archived from groups: comp.sys.laptops (More info?)

R. J. Sutherland <rj.sutherland.removethis@ntlworld.com> wrote in
news:bmec21p114stqrkf2s1653te0ptoo1tdlt@4ax.com:

> In "Re: Thinkpad cut-out saga continues...", Good Man
> <heyho@letsgo.com> wrote:
>
>>Hi, I know this isn't the answer you're looking for, and I'm NOT
>>trying to be a troll, but MICROSOFT PAINT?!?!?!?!?!?
>>
>>I can't imagine a worse program on the planet. If you are even .0009%
>>serious about doing ANY graphics related work, please, find a REAL
>>program that is built to handle graphics processing.
>
> Please, I'm trying to find a handle on this problem with my Thinkpad,
> not engage in a discussion in the relative merits of various graphics
> applications. At this point, I'll take any reliable evidence that
> helps me sort this out!

I know... but isn't this really only happening in Paint? Which brings me
back to what I was trying to suggest, which is that Paint is not designed
to handle large files.
Anonymous
a b D Laptop
March 3, 2005 2:30:43 AM

Archived from groups: comp.sys.laptops (More info?)

In "Re: Thinkpad cut-out saga continues...", Good Man <heyho@letsgo.com> wrote:


>[...]... but isn't this really only happening in Paint? [...]

No - as I said both in my earlier post and again here, it's also happening with
Photopaint and on a couple of occasions with Adobe Acrobat Reader too. It just
happens in a slightly more observable way with Paint, which I thought for a
while at least might be slightly more useful. I'm probably affected most by
Adobe Reader because that's the one of them I use most often, but it worries me
mightily that it can happen at all. With anything. That it seems to involve
graphics-related applications might or might not even be significant,
ultimately.
Anonymous
a b D Laptop
March 3, 2005 2:39:43 PM

Archived from groups: comp.sys.laptops (More info?)

In "Re: Thinkpad cut-out saga continues...", "Quaoar" <quaoar@tenthplanet.net>
wrote:

>Google hmonitor.exe or Hardware Sensors Monitor. If the laptop has a
>temp sensor, it might work where MM doesn't since it is more
>laptop-centric.

Many thanks. I did just that, and it does give a little more info than MBM, even
though hnonitor doesn't formally support the A21p. It apparently shows CPU temp,
fan speed and (after some twiddling) some voltage values. I don't know how much
to trust its readings though, since the fan speed stays planted at 0 even when
the fan is audibly running, and the displayed voltage values are all way over
their nominal values. So do I have a fan with a broken tacho connection, or
faulty motherboard power regulation? Neither is impossible, but if so, I really
am in deep doo-doo.

More usefully perhaps, the CPU temp varies very interestingly. When I persist in
swishing the whole image around within the dreaded Paint, the indicated temp
climbs from a quiescent 47 degC up to as high as 70 degC. If the indicated temps
are accurate, they're somewhat higher than I feel comfortable with, and might
imply some kind of cooling problem, especially since the fan doesn't seem to
kick in at all while this is happening. (Despite PCDoctor's assurance, a fan
problem or fan control problem has to be a prime suspect...)

The plot thickens. I set a 70 degC CPU temp limit in hmonitor, which if reached
causes the CPU clock rate to drop to 25%. It works, as I can easily verify with
the above Paint test. Now, gritting my teeth, I start up Adobe Reader with (oh,
grim irony!) the IBM Thinkpad manual, and scroll it rapidly through sections
that contain a lot of images. (That's the particular set of actions that cause
the cutout to happen with Reader.) The indicated CPU temp lifts a tad from
quiescent to around 56 degC, then - clunk - off goes the machine. So was there a
sudden massive surge of >15 degC in indicated CPU temp that hmonitor couldn't
react to in time, or is there some entirely different kind of overload
happening...?


[As an irrelevant aside, hmonitor offers a week-long free trial, which is fair
enough, but during that period it also keeps stopping after 10min continuous
operation, which for a temp monitor program I personally find just a mite too
mean-spirited... ]


>You might try booting to safe mode (Win2K has it?) or
>command prompt and delete the page file in the event that it is
>corrupted. Run the Win2K equivalent of chkdsk /f (might be the same),
>and memtest86.exe.

In safe mode the system seems, if anything, more likely to cutout. I'll try
deleting the current page file and re-checking the partition it was formerly
located on. Memtest86 I've run on full test for over a day solid, without any
problem. The fan certainly comes on when that's running. All the DOS-based tests
seem to be fine, which is what makes me tend to suspect Win2k.


>Does Win2K have the Event Viewer? That might offer a clue if it does.

Yes it does, but there's nothing relevant in there. The cutout may by its very
nature prevent any error or warning being saved, even if one were in principle
available.
Anonymous
a b D Laptop
March 3, 2005 2:39:44 PM

Archived from groups: comp.sys.laptops (More info?)

R. J. Sutherland wrote:
> In "Re: Thinkpad cut-out saga continues...", "Quaoar"
> <quaoar@tenthplanet.net> wrote:
>
>> Google hmonitor.exe or Hardware Sensors Monitor. If the laptop has a
>> temp sensor, it might work where MM doesn't since it is more
>> laptop-centric.
>
> Many thanks. I did just that, and it does give a little more info
> than MBM, even though hnonitor doesn't formally support the A21p. It
> apparently shows CPU temp, fan speed and (after some twiddling) some
> voltage values. I don't know how much to trust its readings though,
> since the fan speed stays planted at 0 even when the fan is audibly
> running, and the displayed voltage values are all way over their
> nominal values. So do I have a fan with a broken tacho connection, or
> faulty motherboard power regulation? Neither is impossible, but if
> so, I really am in deep doo-doo.
>
> More usefully perhaps, the CPU temp varies very interestingly. When I
> persist in swishing the whole image around within the dreaded Paint,
> the indicated temp climbs from a quiescent 47 degC up to as high as
> 70 degC. If the indicated temps are accurate, they're somewhat higher
> than I feel comfortable with, and might imply some kind of cooling
> problem, especially since the fan doesn't seem to kick in at all
> while this is happening. (Despite PCDoctor's assurance, a fan problem
> or fan control problem has to be a prime suspect...)
>
> The plot thickens. I set a 70 degC CPU temp limit in hmonitor, which
> if reached causes the CPU clock rate to drop to 25%. It works, as I
> can easily verify with the above Paint test. Now, gritting my teeth,
> I start up Adobe Reader with (oh, grim irony!) the IBM Thinkpad
> manual, and scroll it rapidly through sections that contain a lot of
> images. (That's the particular set of actions that cause the cutout
> to happen with Reader.) The indicated CPU temp lifts a tad from
> quiescent to around 56 degC, then - clunk - off goes the machine. So
> was there a sudden massive surge of >15 degC in indicated CPU temp
> that hmonitor couldn't react to in time, or is there some entirely
> different kind of overload happening...?
>
>
> [As an irrelevant aside, hmonitor offers a week-long free trial,
> which is fair enough, but during that period it also keeps stopping
> after 10min continuous operation, which for a temp monitor program I
> personally find just a mite too mean-spirited... ]
>
>
>> You might try booting to safe mode (Win2K has it?) or
>> command prompt and delete the page file in the event that it is
>> corrupted. Run the Win2K equivalent of chkdsk /f (might be the
>> same), and memtest86.exe.
>
> In safe mode the system seems, if anything, more likely to cutout.
> I'll try deleting the current page file and re-checking the partition
> it was formerly located on. Memtest86 I've run on full test for over
> a day solid, without any problem. The fan certainly comes on when
> that's running. All the DOS-based tests seem to be fine, which is
> what makes me tend to suspect Win2k.
>
>
>> Does Win2K have the Event Viewer? That might offer a clue if it
>> does.
>
> Yes it does, but there's nothing relevant in there. The cutout may by
> its very nature prevent any error or warning being saved, even if one
> were in principle available.

hmonitor on my laptop only provides the temperature and I've found it to
be directionally consistent with the fan load and the state of plugging
of the heat pipe heat exchanger with cat/dog fur/pizza crust, etc.

You didn't mention deleting the pagefile in safe mode. It will be
rebuilt on the next boot. A corrupt pagefile *might* have something to
do with the problem.

I do not recall from your original post if you removed the display
adapter, rebooted and reinstalled it? If there is nothing in Event
Viewer then it looks certain to be a thermal problem. At this point, if
deleting the pagefile does not achieve anything, I would open the
computer and check the fan for free rotation and if there is heat pipe
cooling, clean it out. If you find nothing in the heat pipe, then it's
time to check the heat sink contact with the CPU and rebond with thermal
paste or new thermal pads. If there is a GPU heat sink, do the same
with it.

Q
Anonymous
a b D Laptop
March 3, 2005 8:08:14 PM

Archived from groups: comp.sys.laptops (More info?)

R. J. Sutherland <rj.sutherland.removethis@ntlworld.com> wrote:
>seems to be some overload aspect to this

I wonder if you can put a DVM on some of the internal voltages and see
what's happening as it falls over...
Anonymous
a b D Laptop
March 15, 2005 11:51:39 PM

Archived from groups: comp.sys.laptops (More info?)

Yes, I'm beginning to fear that a hardware intervention of one kind or another
might be required to nail this one. Unfortunately, right now I need my Thinkpad
more than I need to fix this problem, which doesn't occur too often with my
normal gamut of activities. But many thanks for all your suggestions anyway (and
any further thoughts or experiences are of course most welcome). If I ever sort
the problem out, I'll post the solution here...
!