CPU Heat Problem!

by thirdeye187 Apr 18, 2005 11:30AM PDT

CPU Heat problem

--------------------------------------------------------------------------------

I've just recently built a new P4 3.0 LGA775 computer. When I first installed the HSF I messed up the thermal compound (was running temp upwards of 90) so I bought some Arctic Silver 5.

The first time I replaced the thermal compound it worked pretty well. I used Isopropyl alcohol (99%) to clean off the HSF and the CPU. I then applied some AS5 to the bottom of the heatsink, rubbed it around with a plastic bag, and then wiped it off with a paper towel (I have been told by many that this is a good way to fill the valleys on the HSF to allow for maximum contact). Then I applied a dab of AS5 to the CPU and mounted the HSF. I started out with an idling temp of 45, then it gradually moved up to about 55 and stayed there. My active temp was around 65-68 for a while, and gradually moved up to 70. When I played Half Life 2, it went up to 75. This was good, but still too high, so I decided to re-do the AS5.

So I did it again tonight, followed the same procedure...and 5 seconds into booting up I had a temp of 90 C. I turned it off right away and redid the AS5. Same thing happened, 5-10 seconds into booting I had a temp of 90.

So...any ideas? I didn't think it was possible for the CPU to get that hot right off the bat, and I definitely can't run at those temperatures.

Discussion is locked

- Collapse -

The CPU gets extremely hot in seconds.

by Ray Harinec Apr 18, 2005 11:51AM PDT

It can be dissipating 80 plus watts immediatly at turn on.

Difficult to see how "messing up" the thermakl compound can cause such drastic effects.

The thermal paste should be a very thin coating on both the heat spreader of the CPU and the base of the HSF. If the HSF is one that came with the CPU in a retail box, it will hold the CPU temp at a safe level [not very cool, as for overclocker types, but safe], that is assuming that the devices are mared together properly and the fan is running at the correct speed.

The other requirement [that should have come on a note with the CPU] is that the internal ambient air of the case in the vicinity of the HSF be 38 C or lower. Shoot for 32C really.

For a better cooler look for posts by someone called ozos [or osoz] where he gives some recco's.

Also try to be certain that the temps you are reading are valid. The CPU should have shut itself/mobo down when you hit 70 C, so maybe your temps are not valid, although I would suspect that somehow the HSF is not adequately clamped to the CPU. That has to be a very tight connection.

Only other thing to think about is that Arctic Siler is an electrical conductor as well as a thermal conductor. Make absolutely certain that none of it gets near any electrical contacts.

- Collapse -

CPU Heat problem (cont'd)

by thirdeye187 Apr 18, 2005 12:02PM PDT

With the original thermal compound, the HSF *slid* a bit when I installed it, and the thermal pad got smudged. When I took it off, I saw that there were air pockets and such in the pad.

None of the AS5 is near any of the contacts. It's coverng a circle in the middle of the heat spreader.

I was told that for AS5, the best way to apply it is to put a small blob in the middle of the heat spreader (over the CPU core) and then allow the pressure of the HSF to spread out the AS5over the surface of the CPU. If I try to spread it around with a plastic bag or something like that it leaves little pockets where there's nothing, so he blob seems to be the best way. As I said, it was working before, I just thought I could reduce the temps some more. It wasn't really going above 75 before, but now its back up to 90.

Are you sure the CPU shuts down at 70? I thought the recommended maxiumum was 72, but that it didn't shut down under 100...

- Collapse -

Even though, the squeezing of a blob

by Ray Harinec Apr 18, 2005 12:55PM PDT

seems to be stated in some sites. What is wanted is a uniformly thin coating over the entire heat spreader and the same area on the heat sink base. From your own observation the squeezing of the blob didn't work. If the blob is too thick, the compound itself gets degraded in its heat transfer ability. Just think of what you observed, if the compound was only a circle in the middle of the heat spreader that means that the rest of the heat spreader was not even touching the base of the HSF, there would be a very thin air gap. That ain't good.

Spread it around with a toothpick [a flat one], it is easy to make it uniform. Yes you may get some on your fingers but you'll live through it. LOL

It's possible that with the latest CPU's Intel raised the temp that they shutdown at, I'll check on that sometime tomorrow night. One must remember though, when they do that it is simply because they had no other choice. I watched as AMD raised the top limit from 85 to 90 then to 95 with the early Athlon's. As with the AMD's, Intel has done nothing that will allow their silicon to run at higher temps, they simply raise the limit and get less margin. I believe that silicon chips start to lose their PN junction function at 125 C [possibly a little higher], which then will not function as a device.

What they do try to do with the smaller line width achitecture is function at lower voltages, thus lower power dissipations. This is not the same thing. Intel shows no power difference beteen their 130 NM and 90 NM chips, but AMD 90 NM 64 bitters show 68 watts, while the same CPU in 130 NM "size" show 89 watts.

- Collapse -

Trouble!

by thirdeye187 Apr 18, 2005 2:32PM PDT

I tried remounting...got a temp of 107 and then it shut itself down. I can't get this thing to work. Will Intel's warranty cover it now that I've removed the thermal pad?

- Collapse -

Yes

by damasta55r Apr 18, 2005 3:43PM PDT

i think your warranty has been voided.
ROger

- Collapse -

If you bought it retail box

by Ray Harinec Apr 19, 2005 1:12AM PDT

they should warranty it, if it is at fault. Even the overclockers get them replaced when they fry them. They just don't mention the o'clocking LOL. In order to remove the HSF from the CPU to return it you screw up the thermal pad.

I would wonder if you somehow got the arctic Silver on a contact or so, else why would the CPU draw so much more power to gert so hot. No need to tell Intel.

- Collapse -

Absolutely hit Intel for a

by Ray Harinec Apr 19, 2005 4:08AM PDT

replacement. The performance that you observed with that CPU seems to show that it had internal problems.

As a separate issue, I think that not shutting down until 107C either says your temp sensor is not at all accurate, or the CPU was defective in the temp diode also.

- Collapse -

Link to an Intel thermal

by Ray Harinec Apr 19, 2005 4:55AM PDT

design document for the family of CPU's that include yours. Note on page 20 the chart clearly shows, that with their heatsink with a 0.28 C/Watt thermal coefficient, they show that they expect the CPU temp to be about 72 C at max power dissipation [obviously then my shut down temp was greatly in error]. This is based on having an ambient temp for the Heatsink to be 38C. Thus getting that ambient temp down will drop the CPU temp down degree for degree.

http://developer.intel.com/design/Pentium4/guides/302553.htm

It's a large document and I haven't had time to find out if they actually give a shutdown temp in simple English rather than requiring an Engineering analysis.

- Collapse -

Another link of summary info

by Ray Harinec Apr 19, 2005 5:22AM PDT

for the P4's. You can see the power increases and temp increases.

http://processorfinder.intel.com/scripts/list.asp

CPU Heat Problem!

CNET Forums

Other Forums

Forum Info