Hi,
My old problem is back.
Some time ago I upgraded motherboard, CPU (AMD 3800+ X2), 2Gb Ram and Nvidia Quadro graphics. The setup was fine but crashed during Norton scan. This was found to be overheating (using Core Temp program). I installed an Akas Evo 120 CPU cooling unit and the problem went away albeit with the case open. The system ran fine for about 3 months.

Now the problem is back but the system will now only run for about 5 minutes before reaching 120 degrees and shutting down. This is without any hard work for the CPU. I've tried blasting the heat exchanger with the air duster but no effect. Any help would be gratefully received

regards

Did you use TIM - thermal interface materials when mounting the CPU heatsink?

How many case fans? Do they all work?

1. What is the ambient temp. in your room?
2. Are your case fans flowing in the right directions (ie. front to back, bottom to top?)
3. It wouldn't hurt to reseat the cpu and reapply thermal grease as well.

If these things are good to go, you should have regular temps (as long as your heatsink is designed for your cpu)

Originally Posted by trs1838

120 degrees

Hmmm, 120°F = 48.9°C - nothing wrong with that! Let's get some clarification before we end up breaking something that may not be broken. Note to TRS - Computer temps are typically provided in Centigrade. Please verify temps and get back with us. Do you have a desk fan you can blast into the side in the meantime? What are you using to measure?

Your motherboard utilities disk should have a monitoring program (or check for a more recent version on your motherboard or PC maker's website). If none, I recommend CoreTemp for newer Intel and AMD64 CPUs. Speedfan is a great and popular alternative, or you can try Motherboard Monitor. Unfortunately, I have found that these programs often have problems properly identifying and labeling the sensor they are reading. The temperatures shown are as accurate as the inexpensive, low-tech sensors will allow, but it may say System Fan instead of CPU Fan. Fortunately, the programs do allow you to edit the labels, so I use Everest to verify the temperatures (as it is able to put sensor to label correctly), then edit the label in the monitoring program. In Everest, look under Computer > Sensor, then wait a couple seconds for the readings to appear. Unfortunately, Everest does not minimize to the system tray to show real-time temperatures, otherwise, you could use Everest instead of the others.
*****

But while I have the soapbox and we're waiting for trs to come back...

It wouldn't hurt to reseat the cpu

Hmmm, no, sorry, I am afraid I (strongly!) disagree and I discourage that practice! Don't do it!

Since the early 90s, ZIF (zero insertion force) sockets have been used for large ICs and CPUs to protect the growing number of fragile pins that snap off too easily. There is no friction to overcome to insert or pull the device from the socket, therefore, you gain nothing by reseating it. You only risk destroying it!

Another advantage to ZIF sockets is once the CPU is properly installed and secured by the locking mechanism (typically engaged by a leverage arm) the CPU module creates an effective dust cover over the electrical contact areas - keeping them clean in spite of the turbulence around it from the CPU's fan.

In Kevin's defense, there is some logic to pulling and reinserting RAM modules and add-in cards, such as graphics or sound cards. PCI, AGP, and PCIe cards and some RAM modules all connect to spring-loaded contacts in the slots, that then use friction to keep the device securely in place. The problem is, tiny contacts (100s in the tiny space of 2 or 3 inches) means there is only a small amount of tension to create the friction on each contact necessary for a good "mechanical" connection, which is necessary to keep dirt and moisture out, and to maintain a resistance-free "electrical" contact.

Every time a card is inserted, the tension becomes weaker, a little bit of dirt and moisture sneaks in (or is carried in on dirty contacts) and over time, resistance between the contact surfaces disrupts the current flow.

In those cases, reseating a card in a friction-type connector slot may scrape the contacts clean enough to hopefully restore circuit connectivity - assuming necessary ESD precautions have not been skipped, and all power, including ATX required +5Vsb standby voltages, removed.

But again, with CPUs and their ZIF sockets, there is no friction to scrape anything clean, therefore, there is never a reason to pull or reseat a CPU after it has been properly installed the first time, except for total replacement.

and reapply thermal grease as well.

I am all for this, but only if there is something wrong with the TIM that is there.

There are only two reasons to replace the TIM.
(1) The heatsink was removed, or
(2) The existing application was done incorrectly.

Okay (2.1) To replace the OEM/generic TIM with the good stuff you should have used in Step 2.

You should never re-use old TIM unless you are certain of it's type because most types cure after a few heat/cooling cycles. They cannot be cured again. So any time you pull the heatsink fan (HSF) assembly (breaking the cured seal), a good cleaning and new application is needed.

Although not a particularly difficult task technically, it requires a steady hand and physical finesse some may not posses or feel comfortable with. It is very easy to totally destroy the CPU, RAM, and motherboard from a single ESD event - without even touching a thing, and without even realizing an ESD event occurred! Other connections can get knocked loose, tools slip, mating surfaces prepared improperly, TIM applied improperly - any can make a bad day worse.

See my sticky on TIM for my methods on preparing for, and properly applying a fresh layer of TIM.

digerati, thanks for the clarifications; by "reseat" I meant take out and reapply TIM. I didn't literally mean try to get it to sit better lol. And your absolutely right, the only reason to reapply TIM is if it wasn't done right the first time. Ive seen some cpus come from the factory with little or no thermal compound at all, so if he is having heat problems this might be a place to eventually check. Just my .02

by "reseat" I meant take out and reapply TIM

I didn't literally mean try to get it to sit better

Well, that's pretty much what it means - to remove, and then put back in the same place with the same item - as opposed to replacing with a like item.

In any event, to replace the TIM between the CPU die and the heatsink, you still do not, and should not remove the CPU.

Ive seen some cpus come from the factory with little or no thermal compound at all

Ummm, most don't. If there is factory supplied TIM, it is typically in the form of a thermal pad on the heatsink, or a little tube of white silicon compound.