Overheating

[richlevy]

So my desktop computer is shutting off after about 20 minutes of use. I have confirmed that this is a heat issue, since the computer stays on indefinitely if I remove the cover and the amount of time it stays on is shorter when restarted soon after an unexpected shutdown.

The odd thing is that there have been no changes made to the computer. The three fans are all spinning. This includes the fan in the power supply, the 80mm fan I bought a few months ago to place in the back of the case, and the original CPU fan.

I vacuumed all of the dust from the vicinity of the chip and in the cooling block above it. There is some dust on peripheral cards, but the case and motherboard are pretty clean where I can see it.

My current theories are:

1) Dust under the cooling block next to the CPU or between the cooling block and the CPU.

2) The CPU fan speed has diminished. It's a original equipment (maybe five years old) but it's spinning at 2003 RPM. The rear 3-speed fan has a max speed of 2557, so the CPU fan speed is respectable.

3) The power supply is heating up due to age, dust, or it's fan is slowing. Speedfan does not report the power supply temperature or fan speed.

4) I thought the hard drives might be contributing to the heat problem, but they're running at %99F and %102F.

5) I've noticed performance issues with my computer. It's possible the CPU is just heating up due to background processes. I get Windows and VirusScan updates, so the software load continually increases.

I've already removed the CPU fan and cleared off some dust. The next step would be to pull the cooling brick between the CPU fan and the CPU and clean up any dust I found there.

There are no more spots for fans in my case. The rear fan cutout is actually for a 90mm fan but I had could only find an 80mm. It's possible the larger size might make a difference. It's also possible that a slightly faster CPU fan might fix the problem.

Currently we're just running it without the cover.

[Undertoad]

Check temperatures in the BIOS to verify.

[Flint]

Maybe the CPU isn't physically connecting well to the whole cooling mechanism & needs some more of that goop squirted in there.

[ZenGum]

Try lifting the box off the desk with eg a cake cooling tray, to provide better circulation underneath.



You know, it's really all the CO2 in your room that is causing the warming effect....

[Pie]

Careful with the vacuum cleaner. It's basically a big van de graaff generator, and it's pretty easy to blow out some of the more sensitive components on your mobo.


A BIOS update can also fix fan and overheating issues. What is your current chipset/version?

[Spexxvet]

Originally Posted by Flint

Maybe the CPU isn't physically connecting well to the whole cooling mechanism & needs some more of that goop squirted in there.

I agree. I've heard that cleaning out the old goop and reapplying can solve the problem, if that's the issue.

[Crimson Ghost]

Originally Posted by Spexxvet

I agree. I've heard that cleaning out the old goop and reapplying can solve the problem, if that's the issue.

I knew a girl like that in high school......

[tw]

Originally Posted by richlevy

So my desktop computer is shutting off after about 20 minutes of use. I have confirmed that this is a heat issue, since the computer stays on indefinitely if I remove the cover and the amount of time it stays on is shorter when restarted soon after an unexpected shutdown.

Any computer should work normally even in the hottest day of the year. Heat is a symptom; typically not the problem. A computer must work just fine even with large dust balls on that warmest day.

Defective semiconductors (and other problems) are more apparent when warm. Often a failure that will get worse months later are found by simply operating that computer in a 105 degree F room.

What is causing your problem? One method to find the defective part is to selectively heat computer parts until the defective part is identified. Every component inside a computer operates at an ideal temperature even when heated by a hair dryer on highest heat settings.

How did we find problems? We used the guns designed to shrink 'heat shrink tubing'. Those temperatures will burn skin. That heat gun and soldering irons made it easier to find defective parts. One diagnostic tool that anyone has is the hairdryer. Use it.

Also useful are hardware diagnostics. Even Memtst86 (or diagnostics provided by better computer manufacturers) are effective tools if used in conjunction with heat.

This assumes you have established stable voltages exist. Heat (at trivial temperatures) does not cause damage (despite so many popular myths that claim otherwise). But heat does change operational parameters such as voltage thresholds and timing. If supply voltages are marginal, then minor temperature changes (a 30 degree temperature increase) can cause intermittent failures. Just another example of how heat creates symptoms.

Additional cooling would only be curing symptoms. Use heat to find the defect. Do not cure symptoms as so many do by worrying about dust and 'more fans'.

[tw]

Originally Posted by Flint

Maybe the CPU isn't physically connecting well to the whole cooling mechanism & needs some more of that goop squirted in there.

Thermal compound replacement is another example of curing symptoms. That stuff does not go bad. If using better thermal compound from so many heatsink manufacturers, then that thermal compound is perfectly good 20 years later.

Many recommend replacing thermal compound when curing only what they understand. Same people often have no hardware knowledge. Many will install more fans to cure symptoms; not eliminate the hardware defect. A defect that gets worse with age. A defect that has progressed so that even minor temperature increases cause computer crashes.

How to find some defects that may cause computer failure this winter? Execute computer diagnostics on that computer last summer in a room that was 98 degrees.

[richlevy]

I finally tried pushing down very hard on the heat sink. I heard a click and speedfan is reporting temperatures 15F cooler than before. I must have dislodged the heat sink the last time I had to go in and reseat the optical drive cables to the motherboard, something I have to do every month or so.

On the bright side, this has been a learning experience for me about heat sinks, coolant pads, coolant gel, etc.

If I ever do find it cheaper to build my own PC than buy one, I will know what to expect.

[Flint]

Originally Posted by Flint

Maybe the CPU isn't physically connecting well to the whole cooling mechanism & needs some more of that goop squirted in there.

Originally Posted by tw

Thermal compound replacement is another example of curing symptoms. That stuff does not go bad. If using better thermal compound from so many heatsink manufacturers, then that thermal compound is perfectly good 20 years later.

Many recommend replacing thermal compound when curing only what they understand. Same people often have no hardware knowledge. Many will install more fans to cure symptoms; not eliminate the hardware defect. A defect that gets worse with age. A defect that has progressed so that even minor temperature increases cause computer crashes.

How to find some defects that may cause computer failure this winter? Execute computer diagnostics on that computer last summer in a room that was 98 degrees.

But, I was right. It wasn't connecting well physically. Many (most?) problems have simple answers.

Also, I believe that the thermal compound must be spread well and evenly. If the heatsink has become dislodged, that may no longer be the case. Removal and re-application may not replace a faulty substance, but it could improve a botched application (re-establishing optimal physical contact).

Originally Posted by richlevy

If I ever do find it cheaper to build my own PC than buy one, I will know what to expect.

It may not be cheaper, but it is a good excuse to get some hands-on with all those computer guts.

[richlevy]

Actually it just crashed again while I was playing a new game I bought. For now, keeping the case off resolves the issue. I will also be replacing the 80mm case fan with a 92mm I ordered from Amazon. On the bright side Clod gets a commission on the $10 I spent.

So a 92mm fan is 12mm larger than an 80mm fan. But about 1/3 of the diameter of the fan is the hub which does not move any air. So the extra 12mm is a much larger percentage increase than 92/80.

Looking at Antec fans specs here, the 80mm can move 34 cubic feet of air/min and the 92mm moves 38 CFM, mostly becuase the max rpm is lower. I'm hoping that makes a difference.

[tw]

Originally Posted by richlevy

Actually it just crashed again while I was playing a new game I bought. For now, keeping the case off resolves the issue. I will also be replacing the 80mm case fan with a 92mm I ordered from Amazon.

That fan will not solve anything; is a classic example of 'curing symptoms'. A computer must work just fine in a standard room with no chassis fan. One chassis fan - more than sufficient for any properly assembled comptuer - is installed so that the computer will also work in a room above 100 degrees.

The numbers: How many more CFMs does the 90 mm have compared to the 80 mm? 38 CFM into a thermal equation reports about a 2 degree temperature reduction. Your CPU in a 70 degree room must be so cool that when room temperatures increases 30 degrees, the CPU is still in a perfectly ideal temperature. If 2 degrees cooler solves anything, the actual problem remains unsolved.

Use heat (ie a hairdryer) to find the defective part. Heat is a diagnostic tool to locate hardware problems. Any part heated to hairdyer temperatures must work 100% normally.

Why do some install a 120 mm fan? To maintain the same airflow with less noise. Larger fans with the same CFM (more than sufficient for any computer) is a technique for reducing noise.

[tw]

Originally Posted by Flint

Also, I believe that the thermal compound must be spread well and evenly. If the heatsink has become dislodged, that may no longer be the case.

Thermal compound must be applied so sparingly so that most of a CPU makes direct contact with the heatsink. Thermal compound has higher thermal resistance - means less heat transfer.

Best thermal interface is when most of the CPU makes direct contact with a heatsink. Some microscopic holes will exist. A CPU to air pocket to heatsink has even higher thermal resistance (also called an insulator). Only enough thermal compound must be applies to fill these fewer air pockets. Thermal resistance that is higher than the direct 'CPU to heatsink' interface, but more conductive than air.

How sparingly? No thermal compound should spread out to the outer edges of the CPU to heatsink interface for numerous reasons. One is because most all heat transfers occur in the center. Another is that thermal compound outside that interface can create electrical problems and computer instability. Apply only enough thermal compound so that is spreads out in the center half of that interface.

How to discover the quality of a heatsink. All computers must work just fine with the heatsink attached using no thermal compound. Some manufacturers make heatsinks so poorly machined that thermal compound is necessary. The experiment is quite simple. First operate the computer with a heatsink and no thermal compound. Then repeat the experiment with thermal compound. If the heatsink was properly machines, then thermal compound should only create a less than 10 degree temperature reduction. Trivial. But some heatsinks (especially those without the necessary degree C per watt parameter) are not properly machined.

We tell everyone to always use thermal compound. Trying to explain reality to a majority of computer assemblers is virtually impossible. Trying to get them to even ask for the degree per watt numbers is futile. They only understand sound bytes. So they are told to always use thermal compound.

Then Arctic Silver gets them to spend massively for the same compound. People who fear numbers love to be scammed; even believed that Saddam had WMDs.

Bottom line – use as little thermal compound as possible so that only the center half of a CPU to heatsink interface is coated. If thermal compound squeezes out to the edge, then a *worse* (less) thermal conductivity exists.

[Flint]

Originally Posted by tw

Thermal compound must be applied so sparingly so that most of a CPU makes direct contact with the heatsink. Thermal compound has higher thermal resistance - means less heat transfer.

Best thermal interface is when most of the CPU makes direct contact with a heatsink. Some microscopic holes will exist. A CPU to air pocket to heatsink has even higher thermal resistance (also called an insulator). Only enough thermal compound must be applies to fill these fewer air pockets. Thermal resistance that is higher than the direct 'CPU to heatsink' interface, but more conductive than air.

How sparingly? No thermal compound should spread out to the outer edges of the CPU to heatsink interface for numerous reasons. One is because most all heat transfers occur in the center. Another is that thermal compound outside that interface can create electrical problems and computer instability. Apply only enough thermal compound so that is spreads out in the center half of that interface.

How to discover the quality of a heatsink. All computers must work just fine with the heatsink attached using no thermal compound. Some manufacturers make heatsinks so poorly machined that thermal compound is necessary. The experiment is quite simple. First operate the computer with a heatsink and no thermal compound. Then repeat the experiment with thermal compound. If the heatsink was properly machines, then thermal compound should only create a less than 10 degree temperature reduction. Trivial. But some heatsinks (especially those without the necessary degree C per watt parameter) are not properly machined.

We tell everyone to always use thermal compound. Trying to explain reality to a majority of computer assemblers is virtually impossible. Trying to get them to even ask for the degree per watt numbers is futile. They only understand sound bytes. So they are told to always use thermal compound.

Then Arctic Silver gets them to spend massively for the same compound. People who fear numbers love to be scammed; even believed that Saddam had WMDs.

Bottom line – use as little thermal compound as possible so that only the center half of a CPU to heatsink interface is coated. If thermal compound squeezes out to the edge, then a *worse* (less) thermal conductivity exists.

You're right. I didn't intend to say "more is better" I just thought it could have become, for lack of a more specific description, "messed-up" somehow once it was confirmed that the heatsick wasn't actually making a good physical connection.

Incidentally,

Originally Posted by richlevy

I finally tried pushing down very hard on the heat sink. I heard a click...

If you have Socket 775 you should be confirming FOUR CLICKS. Furthermore: each damn click could possibly UN-CLICK one of the previous clicks. I've had a hell of a time getting one of these bastards to get all the way attached. It can take two sets of hands.