What is likely cause of overheating CPU

[Graham]

This is a continuation of the thread "Freeze-ups. Is power supply the
only remaining possibility?"

At this point I'm seriously considering replacing either the
motherboard or the CPU or both.

To summarize:

I acqired this computer with an unknown problem. The symptoms began
with persistent and almost immediate system freezes. This would happen
with any sort of high level activity, for example installing Windows
or runnning Windows. The motherboard had 2 visibly damaged capacitors
which I replaced. This made a major improvement, to the extent that I
thought it was "fixed". I later replaced the third cap in that
cluster.

The freezes returned but now usually take several hours to occur. I
didn't start monitoring the CPU temperature until after these freezes.
The reported CPU temp is consistently 30-40C in BIOS setup and 69-72C
in Windows. I was skeptical of these numbers but I think they may be
accurate. I ran for several minutes with the CPU fan unplugged and the
temp slowly rose from 70 to 92C. During that time the CPU was seen to
throttle back to about 56%. I also (with the fan running) aimed a hair
dryer at the heatsink and saw the temp rise from 70-74C with the fan
increasing rpm from the usual 3000 to 4500. I stopped the hair dryer
and the temp dropped to 70 with the fan coming back to 3000 rpm.

I have also replaced the heatsink/fan.

So, if something is causing the freezes and perhaps as a side effect
is also causing the overheating, would that fault likely lie with the
CPU or on some component of the motherboard?

I have tested the power supply with a multimeter while the computer is
under heavy load and the voltages are well within tolerance. I have
swapped out the RAM and hard drives.

No hardware conflicts reported in Device Manager, nothing obvious in
Event Logs, although I'm honestly not sure what to look for.

System:
Asrock M266A motherboard
Celeron 2.4
512 MB DDR

Thanks,
Graham

 

[Paul]

This is a continuation of the thread "Freeze-ups. Is power supply the
only remaining possibility?"

At this point I'm seriously considering replacing either the
motherboard or the CPU or both.

To summarize:

I acqired this computer with an unknown problem. The symptoms began
with persistent and almost immediate system freezes. This would happen
with any sort of high level activity, for example installing Windows
or runnning Windows. The motherboard had 2 visibly damaged capacitors
which I replaced. This made a major improvement, to the extent that I
thought it was "fixed". I later replaced the third cap in that
cluster.

The freezes returned but now usually take several hours to occur. I
didn't start monitoring the CPU temperature until after these freezes.
The reported CPU temp is consistently 30-40C in BIOS setup and 69-72C
in Windows. I was skeptical of these numbers but I think they may be
accurate. I ran for several minutes with the CPU fan unplugged and the
temp slowly rose from 70 to 92C. During that time the CPU was seen to
throttle back to about 56%. I also (with the fan running) aimed a hair
dryer at the heatsink and saw the temp rise from 70-74C with the fan
increasing rpm from the usual 3000 to 4500. I stopped the hair dryer
and the temp dropped to 70 with the fan coming back to 3000 rpm.

I have also replaced the heatsink/fan.

So, if something is causing the freezes and perhaps as a side effect
is also causing the overheating, would that fault likely lie with the
CPU or on some component of the motherboard?

I have tested the power supply with a multimeter while the computer is
under heavy load and the voltages are well within tolerance. I have
swapped out the RAM and hard drives.

No hardware conflicts reported in Device Manager, nothing obvious in
Event Logs, although I'm honestly not sure what to look for.

System:
Asrock M266A motherboard
Celeron 2.4
512 MB DDR

Thanks,
Graham

Flip to page 24 of this document. See Figure 4.

http://download.intel.com/support/processors/celeron/sb/25174807.pdf

Your processor has a nominal Vcore of around 1.5V. It could be 1.475, 1.5,
or 1.525V. This was typical for the 0.13 micron processors (like my Tualatin
S370).

Figure 4 is a "load line". When the processor is not busy, it doesn't draw much
current. The operating point would be on the left hand side of the plot, on
one of those lines. When the processor is running Orthos or CPUBurn, it is
somewhere closer to the right hand side. As the current drawn goes up, the
Vcore becomes reduced in value, or "droops".

According the the lines, when the current drawn is zero, the processor is not
supposed to be operating at the specified VID of 1.5V. The processor is supposed
to be run at a little less voltage. And, in fact, since the current does not
drop to zero, you are actually a small distance from the left of the graph, so
the idle Vcore should be less than 1.5V.

You mentioned 1.6V, and if that was measured at idle, that is too high to fit
within the bounds of the upper and lower lines.

Your processor might draw 50 amps under load, in which case the Vcore value
drops by 0.135V below nominal. So, instead of seeing 1.5V, you'd see 1.365V
or so.

So it is important to note the conditions, during which the Vcore was measured.
It is either close to normal (if measured at idle), or grossly off, if it was
measured under load.

I agree, that either your processor or your motherboard is at fault. But the
processor managed to pass several hours of Orthos, which means it is not such
a bad character after all. I'd check out what is going on with the Vcore
voltage, as you vary the load conditions, and see if the voltage is
grossly out of whack. In which case, you might want to replace the motherboard.

If I didn't know which was busted, I'd have to start with the motherboard,
due to the excellent record I've personally witnessed over the years,
regarding defective processors. So if the Vcore remains too high, and there
is no BIOS adjustment to try dropping it, then changing the motherboard
is what I'd do next.

Paul

 

[Not Here]

Your processor might draw 50 amps under load, in which case the Vcore value
drops by 0.135V below nominal. So, instead of seeing 1.5V, you'd see 1.365V
or so.

So it is important to note the conditions, during which the Vcore was measured.
It is either close to normal (if measured at idle), or grossly off, if it was
measured under load.

Thanks again, Paul.

Under load, the lowest I've seen is 1.52V, mostly about 1.6, as high
as 1.66. This is as reported by software. I don't think I'm capable of
getting a multimeter reading of VCore.

Given that the software readings I'm getting for 3.3V, 5V and 12V are
showing somewhat lower than my multimeter says, should I trust the
VCore voltage readings and conclude that VCore is really running high?

Graham

 

[Paul]

Thanks again, Paul.

Under load, the lowest I've seen is 1.52V, mostly about 1.6, as high
as 1.66. This is as reported by software. I don't think I'm capable of
getting a multimeter reading of VCore.

Given that the software readings I'm getting for 3.3V, 5V and 12V are
showing somewhat lower than my multimeter says, should I trust the
VCore voltage readings and conclude that VCore is really running high?

Graham

I'd say it is probably on the high side. But it is hard to believe that would
drive the temperature up that much.

Paul

 

[w_tom]

Given that the software readings I'm getting for 3.3V, 5V and 12V are
showing somewhat lower than my multimeter says, should I trust the
VCore voltage readings and conclude that VCore is really running high?

Because of a voltage monitor may be designed, CPU voltage numbers
may be more accurate. Of course it is easy to verify those numbers
once the difficult part is accomplished - tracing those voltage lines
back from CPU pins.

Meanwhile I still don't see where useful experiments were
performed. For example, as CPU is heated significantly with a
hairdryer on highest heat (or something even warmer), then what
happens to reported CPU temperatures? That temperature rise should be
apparent is less than a minute - or faster.

After put into a condition where problems can be located
(motherboard et al heated as hot as the hairdryer can take it), then
what did diagnostics report? Your previous tests were not definitive
because standard extremes were not implemented.

Minor voltage changes on CPU would have zero effect on CPU
temperature. But voltage variation can affect timing.

Heat is a diagnostic tool. Too often, heat is blamed for failures
when the human does not understand semiconductor and computer
principles. Some are so misinformed as to assume those higher
temperatures create hardware damage. Heat is a diagnostic tool.

 

[Not Here]

Because of a voltage monitor may be designed, CPU voltage numbers
may be more accurate. Of course it is easy to verify those numbers
once the difficult part is accomplished - tracing those voltage lines
back from CPU pins.

Meanwhile I still don't see where useful experiments were
performed. For example, as CPU is heated significantly with a
hairdryer on highest heat (or something even warmer), then what
happens to reported CPU temperatures? That temperature rise should be
apparent is less than a minute - or faster.

After put into a condition where problems can be located
(motherboard et al heated as hot as the hairdryer can take it), then
what did diagnostics report? Your previous tests were not definitive
because standard extremes were not implemented.

Minor voltage changes on CPU would have zero effect on CPU
temperature. But voltage variation can affect timing.

Heat is a diagnostic tool. Too often, heat is blamed for failures
when the human does not understand semiconductor and computer
principles. Some are so misinformed as to assume those higher
temperatures create hardware damage. Heat is a diagnostic tool.

The reported CPU temp is consistently 30-40C in BIOS setup and 69-72C
in Windows. I was skeptical of these numbers but I think they may be
accurate. I ran for several minutes with the CPU fan unplugged and the
temp slowly rose from 70 to 92C. During that time the CPU was seen to
throttle back to about 56%. I also (with the fan running) aimed a hair
dryer at the heatsink and saw the temp rise from 70-74C with the fan
increasing rpm from the usual 3000 to 4500. I stopped the hair dryer
and the temp dropped to 70 with the fan coming back to 3000 rpm.

I have also replaced the heatsink/fan.

What diagnostics would you recommend I do while the system is being
overheated?

Graham

 

[saturnlee]

This is a continuation of the thread "Freeze-ups. Is power supply the
only remaining possibility?"

At this point I'm seriously considering replacing either the
motherboard or the CPU or both.

To summarize:

I acqired this computer with an unknown problem. The symptoms began
with persistent and almost immediate system freezes. This would happen
with any sort of high level activity, for example installing Windows
or runnning Windows. The motherboard had 2 visibly damaged capacitors
which I replaced. This made a major improvement, to the extent that I
thought it was "fixed". I later replaced the third cap in that
cluster.

The freezes returned but now usually take several hours to occur. I
didn't start monitoring the CPU temperature until after these freezes.
The reported CPU temp is consistently 30-40C in BIOS setup and 69-72C
in Windows. I was skeptical of these numbers but I think they may be
accurate. I ran for several minutes with the CPU fan unplugged and the
temp slowly rose from 70 to 92C. During that time the CPU was seen to
throttle back to about 56%. I also (with the fan running) aimed a hair
dryer at the heatsink and saw the temp rise from 70-74C with the fan
increasing rpm from the usual 3000 to 4500. I stopped the hair dryer
and the temp dropped to 70 with the fan coming back to 3000 rpm.

I have also replaced the heatsink/fan.

So, if something is causing the freezes and perhaps as a side effect
is also causing the overheating, would that fault likely lie with the
CPU or on some component of the motherboard?

I have tested the power supply with a multimeter while the computer is
under heavy load and the voltages are well within tolerance. I have
swapped out the RAM and hard drives.

No hardware conflicts reported in Device Manager, nothing obvious in
Event Logs, although I'm honestly not sure what to look for.

System:
Asrock M266A motherboard
Celeron 2.4
512 MB DDR

Thanks,
Graham

IF i remember correctly, there are 4 green KZG capacitor at the west
side of the board(next to mosfet), 4 brown KZE at the north side of
the board.
When replacing the caps at VRM, i will replace all . I will use
panasonic FM since it has lowest esr(lower than FC, Chemicon KZE and
KZG)
I am just curious what cap you use.
I have the same board before, the temp monitor is kind of off the
scale. So i just use my finger to feel whether the mosfets and caps
are hot or not when running prime95.
The CPU voltage can be measured at the east side of the inductor(if i
remember correctly, it is either the southest coil or middle one)
The lowest voltage at the one of the inductor is the cpu voltage

 

[w_tom]

The reported CPU temp is consistently 30-40C in BIOS setup and 69-72C
in Windows. I was skeptical of these numbers but I think they may be
accurate. I ran for several minutes with the CPU fan unplugged and the
temp slowly rose from 70 to 92C. During that time the CPU was seen to
throttle back to about 56%. I also (with the fan running) aimed a hair
dryer at the heatsink and saw the temp rise from 70-74C with the fan
increasing rpm from the usual 3000 to 4500. I stopped the hair dryer
and the temp dropped to 70 with the fan coming back to 3000 rpm.
...
What diagnostics would you recommend I do while the system is being
overheated?

Temperature sensor does measure temperature change. If that number
is also accurate, it is rather excessive (assuming this is not a CPU
rated for 90 degree C).

Only certain hardware can create an OS failure. Diagnostics,
especially when hardware is heated, should be performed on video
controller, sound card, memory, CPU, and motherboard hardware
interfacing with those functions. Devices such as keyboard, mouse,
floppy and hard drives, CD-Rom, etc would not cause a system freeze
and need not be tested with diagnostics.

If I understand (remember) what is (was) posted including CPU
manufacturer, it does appear higher CPU temperature (created by
hairdryer) is causing CPU to throttle. Minor changes in CPU voltages
would not account for overheating. But minor CPU voltage changes can
change CPU timings which (if large enough) could cause a CPU crash
(freeze). However that temperature number for a CPU in room
temperature (70 degree F) is troubling; may be too high depending on
temp specs for that particular CPU.

CPU heatsink - does the manufacturer provide a 'degree C per watt'
spec? Suspicious is any heatsink assembly that does not provide that
most fundamental number. Obviously, CPU performance should not be
decreased by hairdryer heat. That is a suspicious and telling number.

 

[PCBONEZ]

IF i remember correctly, there are 4 green KZG capacitor at the west
side of the board(next to mosfet), 4 brown KZE at the north side of
the board.
When replacing the caps at VRM, i will replace all . I will use
panasonic FM since it has lowest esr(lower than FC, Chemicon KZE and
KZG)
I am just curious what cap you use.
I have the same board before, the temp monitor is kind of off the
scale. So i just use my finger to feel whether the mosfets and caps
are hot or not when running prime95.
The CPU voltage can be measured at the east side of the inductor(if i
remember correctly, it is either the southest coil or middle one)
The lowest voltage at the one of the inductor is the cpu voltage- Hide quoted text -

- Show quoted text -

KZG series caps sometimes have problems.
They fail with no visible signs of a problem.
I've repaired 4 motherboards (2 MSI, 1 Asus, 1 Gigabyte) by replacing
only the KZG caps.
It's not consistant and Chemicon isn't talking.

Asrock is NOT Asus. (As some people seem to think.)
Asrock is basically an Asus design that is physically manufactured by
PcChips.
(More correctly Hsing Tech who does the fabrication for PcChips and
many other low end boards.)
..

..

 

[Bob M]

KZG series caps sometimes have problems.
They fail with no visible signs of a problem.
I've repaired 4 motherboards (2 MSI, 1 Asus, 1 Gigabyte) by replacing
only the KZG caps.
It's not consistant and Chemicon isn't talking.

Asrock is NOT Asus. (As some people seem to think.)
Asrock is basically an Asus design that is physically manufactured by
PcChips.
(More correctly Hsing Tech who does the fabrication for PcChips and
many other low end boards.)
.

.

Are you sure? I thought Asrock was the budget arm of Asus. And ECS is
PcChips. At least that's what I've always read and have been told.

Bob

 

[wiseolman]

--
--------------------------------- --- -- -
Posted with NewsLeecher v3.9 Beta 2
Web @ http://www.newsleecher.com/?usenet
------------------- ----- ---- -- -