[Jaekus]

This has been going for months and noone's come up with the solution. I'm thinking of just buying a new MB and CPU but I'm the sort of person that has to get to the bottom of these things one way or another.

CPU temp keeps getting up to high temps when using anything graphics based. Basically the readings get up past 70C and climbing if I use anything flash based on facebook. But the heatsink is actually just above room temp and the fan is spinning at full speed (cool and quiet turned off).

I've cleaned off and reapplied thermal paste more times than I count and for a weird reason it fixes the problem for a day or two, then it starts again. Like someone has suggested to me though if it were really constantly getting above 70C my CPU wouldn't even be working by now, so I'm guessing a faulty reading.

Recently updated the bios and that didn't solve anything. I could turn off temp warnings in the bios but I'm not sure if that would be a good idea.

Anyways, I need some suggestions/advice that go beyond the usual "have you applied thermal paste like this... is the fan spinning... is there dust...?" etc cause from a physical perspective I've done everything by the book and still it isn't fixed.

Thanks in advance,

Jake.

[Advertisements]

Like someone has suggested to me though if it were really constantly getting above 70C my CPU wouldn't even be working by now

That would depend on how far past 70°C, and for how long, and the CPU itself. But of course, since we know nothing of your hardware, that would guessing.

so I'm guessing a faulty reading.

Or sensor. It happens. How are you measuring? What is your other sensor readings?

While you may not want to hear about TIM - an improper application can be as bad as none, so see my canned text on applying it properly here.

Do you have good front to back air flow through the case? What do you have for case fans? I prefer one large (120mm or larger) fan in front drawing cool air in, and at least one large fan in back (in addition to any in the PSU) exhausting hot air out.

If all of the above is fine, open the side and blast a desk fan in there and see what happens.

[Digerati]

Like someone has suggested to me though if it were really constantly getting above 70C my CPU wouldn't even be working by now

That would depend on how far past 70°C, and for how long, and the CPU itself. But of course, since we know nothing of your hardware, that would guessing.

so I'm guessing a faulty reading.

Or sensor. It happens. How are you measuring? What is your other sensor readings?

While you may not want to hear about TIM - an improper application can be as bad as none, so see my canned text on applying it properly here.

Do you have good front to back air flow through the case? What do you have for case fans? I prefer one large (120mm or larger) fan in front drawing cool air in, and at least one large fan in back (in addition to any in the PSU) exhausting hot air out.

If all of the above is fine, open the side and blast a desk fan in there and see what happens.

[Jaekus]

Sorry, I should've put my system specs in my first post.

ASUS M2N-MX SE
AMD Athlon 64 x2 4000+ Dual Core
Ultrix 450W PSU
NVIDIA GeForce 8500 GT
2x 1Gb Kingston DDR2 RAM
Windows XP

I just went out and bought some case fans however I can't put one in the front as my case is weird and old, but I got one in the back, also got some metho and cleaned the CPU, some very fine sandpaper (1200g) and polished the heatsink and got some high quality thermal compound (Arctic Cooling) and applied that as thinly and evenly as I could with my finger inside a plastic bag. So far it's reading 47C and slowly rising.

I didn't have case fans so I had the side open. I had a 40cm room fan pointing at it one night on high and even though the air increased the cpu fan from it's max of 3125 up to 3400 it still was rising in temp!

I let the reading (from Everest, which through observation is parallel to the bios temp reading) reach 78C then shut down the computer and immediately pulled off the heatsink and touched the CPU; it was quite warm, almost hot. So I did as above and am monitoring the temp in Everest.

[Digerati]

So far it's reading 47C and slowly rising.

How long does it take to get to 78°C?

If you are overclocking, don't.

The sandpaper idea may not have been a good one. "Lapping", as it's called, should only be done under very precise conditions, typically in a precision machine shop with precision machinery. Lapping by hand to create perfectly flat mating surfaces generally has the opposite effect.

For future reference, when you seek help on multiple forums (not normally considered good etiquette, BTW), please reference those threads, or a least indicate what was suggested, and done. This is to avoid confusion for all concerned. You asked the same question on the AMD Forums and was told to check your power supply, but I don't see where you have done that - although I see you have discussed replacing the PSU here, but in reference to replacing the motherboard and CPU. It is always essential to ensure you have clean, stable power when dealing with hardware issues. See my canned text below on testing PSUs. And note you should decide on a the size of a new PSU after you have decided what hardware you are buying - especially the graphics card, which often consume more power than CPUs. And while I also like Corsairs, Cooler Masters are good too.

I have never heard of Ultrix PSUs and cannot find them on any recommended lists. That does not mean it is bad, but it does suggest it does not have a good reputation either.

***
Testing PSUs canned text:

To properly and conclusively test a power supply unit (PSU), it must be tested under various realistic "loads" then analyzed for excessive [url="http://en.wikipedia.org/wiki/Ripple_(electrical)""]ripple[/url] and other anomalies. This is done by a qualified technician using an oscilloscope or power analyzer - sophisticated (and expensive) electronic test equipment requiring special training to operate, and a basic knowledge of electronics theory to understand the results. Therefore, conclusively testing a power supply is done in properly equipped electronic repair facilities.

Fortunately, there are other options that are almost as good. I keep a FrozenCPU Ultimate PSU Tester in my tool bag when I am "in the field" and don't have a good spare power supply to swap in. While not a certain test, they are better than nothing. The advantage of this model is that it has an LCD readout of the voltage. With an actual voltage readout, you have a better chance of detecting a "failing" PSU, or one barely within specified ATX Form Factor Standard tolerances. Lesser models use LEDs to indicate the voltage is just within some "range". These are less informative, considerably cheaper, but still useful for detecting PSUs that have already "failed". Newegg has several testers to choose from. All these testers contain a "dummy load" to fool the PSU into thinking it is connected to a motherboard, and therefore allows the PSU to power on, if able, without being attached to a motherboard - great for testing fans, but again, it is not a true load or suitable for conclusive testing.

As mentioned, swapping in a known good supply is a tried and trued method of troubleshooting used for years, even by pros. Remove the "suspect" part and replace with a "known good" part and see if the problem goes away.

I do not recommend using a multimeter to test power supplies. To do it properly, that is, under a realistic load, the voltages on all the pins must be measured while the PSU is attached to the motherboard and the computer powered on. This requires poking (with some considerable force) two hard and sharp, highly conductive meter probes into the main power connector, deep in the heart of the computer. One tiny slip can destroy the motherboard, and everything plugged into it. It is not worth the risk considering most multimeters, like plug-in testers, do not measure, or reveal any unwanted and potentially disruptive AC components to the DC voltages.

And remember, anything that plugs into the wall can kill. Do not open the power supply's case unless you are a qualified electronics technician. There are NO user serviceable parts inside a power supply.

[Jaekus]

Sorry, I'v just been frustrated and haven't had the money to take it to a computer tech to find out what the problem is, hence also not buying a new PSU etc yet. My thought process in all this is to find the answer and then close the threads in various forums with the answer and thank you's to those who have helped.

I didn't know about the lapping, it was suggested by a couple people and I've already done it. Maybe I need to get a better CPU cooler anyway, so when I get a chance I'll give that a go. Bit annoying with all these expenses coming up at this time of year!

[Digerati]

I understand your frustration. This problem is definitely not an obvious one.

I would not spend any money on a new cooler until I was sure I had good power - even if I had to swap it out of another machine that was working. But buying a new PSU is not necessarily a waste as it will carry over to your next computer.

Lapping works, but only if done right. Typically, the grit is placed on a precision, machined, perfectly flat surface. A kitchen table or counter does not cut it. And if you used your thumb then you may have hallowed a deep cavity to trap all kinds of insulating air in - the worse thing you want between the heat sink and CPU die mating surfaces. I hope it was only done to the heat sink, as that can be replaced. The CPU die cannot.

AMD and Intel machine the CPU die surfaces to be pretty darn flat. HSF surfaces are too. The TIM compensates for any microscopic imperfections. Even if lapping is done right, it only amounts to a few degrees improvement, if that.

Just reading back over your problem - have these high temps ever caused the computer to shutdown, freeze or reboot?

[Jaekus]

Yes, once the CPU reading gets to 80C the computer shutsdown without a shutdown sequence.

I did use my thumb and immediately realised my error, tried to compensate by using something to subtitute as a sanding block but it wasn't a completely flat surface. So I may have ruined the heatsink but I definitely didn't even touch the CPU with anything other than methylated spirits and a tissue, rubbed carefully over the surface. So I'll probably need to replace the heatsink I guess.

My flatmate has a computer he said I can use for spares so I'll grab the PSU out of it tomorrow (at work till then) and see if that helps. I've noticed in Everest the CPU voltage fluctuates between 1.3 - 1.36v - is this unusual behaviour?

Thanks,
Jake.

[Digerati]

1.3 - 1.36v is this unusual behaviour?

That is not a voltage you need to worry about. You need to be concerned with the +12V, +5V, and the +3.3V. Those need to be fairly close to those values, and stable. BUT - a software based monitor uses your hardware's sensors which may not be accurate either. They would still show if unstable, however.

Since the system is shutting down, I fear something is truly getting too hot. If blasting a desk fan in there does not stop the overheating, then that suggests to me an over-current situation, which, in turn, suggests a part is failing, and drawing too much current, perhaps in the motherboard's regulator circuits. Not good. Do make sure you have backed up all your important data.

[Jaekus]

Hmm, well since the CPU is overheating is it probably that? In Core Temp one core more often than not reads at 10-15C higher than the other no matter how hot it gets. Typically one would read 59 whilst the other is about 47 - even when it apparently gets up to insane temps past 100C one core will be 115, the other 102C - but surely it's not getting this hot or the CPU wouldn't even work anymore, would it?

I'll still try swapping in a different PSU and see what happens.

[Digerati]

Hmm, well since the CPU is overheating is it probably that?

It's a "possibility" but I'm not going to say "probably" without knowing if it is being fed good power. But one core being 10-15° higher than another seems like a big difference to me. On my Intel 5300 Dual core, the cores are typically just 1 or 2° apart. On my Core i7 quad, there is typically a 3 to 4° spread. I don't have an AMD here to check.

I don't know which 4000+ you have. There are 3 as seen here. In any case, 72°C seems like the top of the normal operating range.

BTW - you mentioned the heatsink does not get hot - that is not really surprising since the heat is spread out so much across a large surface area, and the materials are chosen for their conductive properties.

[Jaekus]

The CPU I have is the ADO4000IAA5DD, as per the link you provided (number read from the top of the CPU).

I swapped PSUs and it seems to be running a little cooler, but when I ran a test using OCCT for 5 mins the temps steadily increased and didn't drop at all until after the test.

Another weird thing is I'm getting different readings. Core Temp says my current CPU cores are running at 46 and 57C, whilst OCCT says they're only 36 and 25C respectively. And Everest says 51C... I don't understand how this is.

My flatmate's computer has the same socket as my current CPU (AM2) so I can swap it and see what the results are. But I'll hold off for a bit first.

[Jaekus]

Update:

I swapped CPUs, ran OCCT for 15 minutes and the highest it got to was 54C for once second, but remained pretty stable between 50-53C. The CPU I swapped to is AMD Athlon 64 3200+. I used the same heatsink and fan as the previous CPU which was giving me problems. Would it be leaning towards a faulty CPU? Incidently AMD emailed me today asking for the nature and details of my overheating problems. Only thing is I don't have proof of purchase. If that avenue doesn't work I'll just buy a new CPU, probably an AMD Athlon 64 X2 5050e that I've seen advertised for $79 AUD.

[Digerati]

It looks like a failing CPU then.

[Jaekus]

Good news is AMD have decided to honour their 3 year warranty as the CPU was made in 2007 (serial number includes 0745, which the staffer said it means it was made in the 45th week of '07) so it should be all good from what I can determine at this stage.

Thanks for your help Digerati.

[Digerati]

Super! Thanks for the follow up.