I just upgraded an old machine, to use on my desktop. Not needing a lot of power, I chose an inexpensive Gigabyte GA-K8U motherboard (ULi1689 chipset) that got decent reviews on Newegg and a Sempron 2600 CPU. Figuring to play with overclocking a little (even though this is a lousy mobo for that purpose), I gave it a huge heatsink - A Zalman CNPS-7000B AlCu job with 92mm fan.

I'm currently runing this at stock speed (main clock @200MHz, CPU at 1600 MHz), as that's the fastest it seems to want to work at. Wondering why it wouldn't overclock at *all*, I downloaded a temp-checker program (SpeedFan 4.25) and was astonished to see it reporting idle temps for the CPU of roughly 50C. (!) Distrusting SpeedFan, I tracked down Everest Home Edition (cool program, BTW), only to find that it's reporting 50C too.

What's weird is that the base of the HS/Fan is only barely warm to the touch. Even more strange, if I cook the CPU by running CPU Burn, getting the temp up to ~56C, then *immediately* reboot the computer, the BIOS reports a CPU temp of only 34C. (Total delay from end of CPU Burn to checking temp in BIOS is about 40 seconds.)

So here's my question: Is there any possibility that the mobo is just not reporting temps to apps like SpeedFan and Everest properly, but *is* working correctly when talking to the BIOS? Am I in danger of cooking the CPU, or is this all just a case of bad data of some sort. - Has anyone else ever heard of a mobo reporting incorrect temps?

I'm pretty certain that what's being reported is indeed some sort of CPU temp though, because it does go up and down pretty rapidly, depending on CPU load.

I also am pretty confident that the HSF is attached properly, because I wrestled it off, confirmed that the thermal paste (Artic Silver Ceramique) was evenly distributed. It looked like I had a bit too much on, so I wiped a little off (now leaving a pretty darn thin layer, I hope it's enough), but the results after bolting everything back together are virtually identical.

Any ideas?

TIA,
Dave E.

Try a bios update?

http://www.giga-byte.com/Motherboard/Support/BIOS/BIOS_List.htm

It is the mark of an educated mind to be able to entertain a thought without accepting it. --Aristotle--

There might not be any mystery to it, as CPU heatsinks are designed to dissipate heat very quickly, and as you mentioned you have a large heatsink. If the running temp worries you, try improving the air circulation in your case, by putting extra case fans in, moving devices etc. Things like untidy cables can make a big difference to CPU temp. I found with my Duron the PSU used to put a lot of heat out directly over the CPU until i replaced it. Might be worth getting a more powerful PSU?

Galileo - Thanks for the suggestions. I'll look into a BIOS update. I'm on version F5, the latest is F6, although the notes for it don't mention anything about temperature reporting. I'd been thinking of that as well, am just a little nervous about screwing up the bios flash somehow and rendering the mobo unusable. (NewB)

prayersonfire - Good thoughts, but I don't think air circulation is the issue. The HSF runs quite cool to the touch, and airflow in the cabinet is quite good. (Very few cables, two case fans blowing air in, PSU right above the HSF is sucking air out.) Good suggestion though.

The temperature reporting diode is usually UNDER the CPU, therefore it will be hotter than your heatsink, since the heatsink's job is to dissipate heat.

If you are overclocking using the front-side bus, remmember that your memory should be rated higher than the speed you are setting it.

Rule #1 Good computers don't go down.
Rule #2 There is no such thing as a good computer.

In an ideal situation, one would want the heat-sink to be "hot to the touch". This indicates, that in fact, it is doing its job by absorbing the heat being generated from the CPU. On the contrary, if the heat-sink is "cold to the touch" this implies insufficient contact between the two metals, thus resulting in inadequate thermal transfer. I found this to be the case when I first installed my heat-pipe cooler, the base had a concave in the middle of the whats supposed to be flat base. This formed a gap between the CPU and heatsink causing my temps to sky rocket.

And any rate, in your case I believe the temps are simply just not being reported correctly.

Here is an interesting read, if you are curious and still concerned about your temps.

http://www.heatsink-guide.com/content.php?content=maxtemp.shtml

People demand freedom of speech as a compensation for the freedom of thought which they seldom use. . --Kierkegaard--

Galileo & Rick -

Thanks for the info! Interesting point about the location of the temp-sensing diode. It'd certainly be hotter than the surface of the case, OTOH, isn't it actually just a diode integrated on the IC itself, so it'd be actually showing the temp of the silicon? (Or is it a separate component, built into the chip package?) More and more though, I suspect that there's something just being interpreted wrong with the temperature information. - When I do a quick reboot, the BIOS screen always reports dramatically lower temps than do any windows-based temperature-display programs (Speed Fan or Everest). In the 40 seconds or so that it takes to reboot, I'm thinking it unlikely that the CPU temp would drop by nearly 20 degrees C.

It is definitely worth checking to see if my cooler or the chip package itself isn't flat. I've heard of people lapping the surfaces to get them perfectly flat, but I always assumed that was just anal-retentive obsessivenes. Galileo's experience indicates that isn't the case though. Maybe I'll try lapping the chip case and cooler this weekend, see what that does. The HSF is definitely barely warm to the touch when the system is running. (OTOH, the temp is 50C at idle, and only goes up to 56C when running 100%. If there really was bad contact between the chip and HSF, I'd think that the idle/100% temp differential would be a lot greater.)

As it happens, I've been bitten by the OC bug, so may be upgrading the whole thing again anyway. - I've been so impressed by what this lowly little Sempron is doing in a cheap mobo, that I'm seriously considering tearing it all out and installing an Athlon 3400+ in a DFI LanParty board, so I can OC more. (My current mobo has no access to HT multipliers, RAM dividers, etc.) Having worked with a 2.4 GHz Pentium for quite a while now, I'm flat-out amazed at how noticeably faster a 1.6 GHz Sempron is...

Anyway, thanks for the great info. I may try lapping the Sempron and HS this weekend, to see if it makes a difference.

It looks like the problem is chip-specific, some AMD chips are consistently underreporting temperatures. There's a long thead over on dfi-party with a lot of users' experience with various chips:

http://www.dfi-street.com/forum/showthread.php?t=2305&page=1&pp=15

For whatever reason, a variety of AMD chips seem to just report much higher temps in apps like SpeedFan than they are actually running at.

- So I don't think I'm going to take the time to lap the HSF this weekend, AFAICT, the chip is actually running fairly cool.

Thanks for everyone's help and suggestions!