[stearmandriver]

Quick question: I installed the CoreTemp utility today to use while running stress tests for overclocking, and I noticed that CoreTemp seems to read each core's temp 8-10 degrees C below BIOS. For instance, with the system at idle, CoreTemp shows around 21C (pretty much ambient temp), but if I reboot and immediately check system monitor in BIOS, it shows around 31C.

Any thoughts why? Is this a known issue with either Coretemp or the BIOS on a 680iSLI? Is there another temp monitor you guys use for overclocking that's considered accurate?

Thanks...

[Advertisements]

Hi there

I'm unable to give you a definitive answer to this but I can suggest some possible explanations:

1. Some monitoring utilities seem to mix up the sensors i.e. the motherboard temperature (or system temp) is reported as the cpu temp etc.

2. The temp reported in the bios is often slightly higher than the temp reported inside of Windows due to power management etc

3. There are often 2 sensors on the cpu - one that records the temperature in the cpu die, and one that reports the cpu surface temperature. It's possible that the bios is reporting the temperature of a different sensor to the one used by the monitoring utility.

[Samm]

Hi there

I'm unable to give you a definitive answer to this but I can suggest some possible explanations:

1. Some monitoring utilities seem to mix up the sensors i.e. the motherboard temperature (or system temp) is reported as the cpu temp etc.

2. The temp reported in the bios is often slightly higher than the temp reported inside of Windows due to power management etc

3. There are often 2 sensors on the cpu - one that records the temperature in the cpu die, and one that reports the cpu surface temperature. It's possible that the bios is reporting the temperature of a different sensor to the one used by the monitoring utility.

[stettybet0]

CoreTemp has an issue with some Core 2s, in which it reads 15C lower than the actual temp. But then why is the temp only 8-10C lower? That's because CoreTemp reads the actual temp of the core, while BIOS records the surface temp of the chip. In this case the core is 5-7C hotter than the surface temp.

However, what version of CoreTemp are you using? I thought they fixed this problem in version 0.95.4.

Also, what does it list the tjunction temp as?

Edited by stettybet0, 26 October 2007 - 08:56 PM.

[stearmandriver]

I have version 0.95.4.

Tjunction temp is 85c.

Now, I've installed speedfan too. Core 0 and Core 1 temps in speedfan match Coretemp's values. The CPU temp in speed fan is around 5-8c warmer than the core values, so it's jibing with the BIOS values. The "Core" temp in Speedfan is awfully warm - 52c at idle.

So... which value should I take as my cpu temp? And what is that "Core" temp, and why is it so hot? Is that a problem?
Oh, one more interesting thing - under 100% CPU load while running the Orthos stress test, occasionally the Core 1 value will get around 7 degrees warmer than Core 0. Most of the time they match... weird.

Anyway, just curious. All I really care about is which number to watch to keep my chip from smoking. ;-)

Thanks!

[stettybet0]

Alright, then I'm not sure what's up with CoreTemp. I have a feeling that you might be reading the wrong temps, but I'm not too familiar with CoreTemp, so let's stick with SpeedFan for now. The only temp you want to worry about is the CPU temp. To keep your chip safe, you don't want the CPU temp above 65C. It can withstand up to 85C, and will probably shut itself down before that, but you don't want to be messing with stuff like that.

Anyways, a nice way to keep your CPU cooler: go to BIOS and go to System Monitor, then to Fan Settings and turn the CPU fan to manual 100%. This will keep your CPU about 3-7C cooler, depending on your fan. Of course, it might be a little noisy... So if this is supposed to be a quiet PC, then disregard that.

And of course, the #1 way to keep a chip cool: Give it the lowest voltage that it will be stable at. Dropping 0.1V cuts about 5-8C off your temps. And considering how the 680i over-estimates CPU voltage by about 0.1V, this shouldn't be too hard.

[stearmandriver]

Gotcha. I appreciate all the help, Stetty.

One more question for ya... if I plan on the 5:4 ratio you suggested in the other thread, my RAM will be running at 900Mhz. What kind of voltage boost - if any - do you expect I'd need for that? I've read alot about starting points for CPU voltage boosts, but I can't find anything for RAM. All I've seen is that the limit is probably 2.2v.

Thanks... I promise I'll stop buggin ya soon...

[james_8970]

Every chip will act differently, overclock in small increments and stress test the system.
Then repeat.
If your computer fails the stress set (ig crash) then either up the voltages or back off a bit.
Depending on your RAM 2.2V may even be to much, there is 800MHz memory that will run stable at 1.9V. Start at the default recommended voltage for your RAM and raise it accordingly.
Don't raise the RAM voltage over 2.2V as many motherboards cannot do this proporly, it's a known issue with a number of motherboards.
The only way to get an accurate temp reading is by taking a number of programs, running them, and use the temperature that the majority of the programs report. There should be a better way, but at this point in time there really isn't.
James

Edited by james_8970, 28 October 2007 - 07:01 PM.

[stettybet0]

Don't start at the default voltage, start at the recommended voltage. (The 680i defaults my DDR2-800 RAM to 1.85V, the recommended is 2.2V.) Then, if it runs stable, you can try taking the voltage down a bit. My RAM runs stable at stock speeds with 2.1V. Taking my RAM up 100mhz (to 900mhz) requires it to have the full 2.2V.

A quick note about recommended voltage: The recommended voltage can usually be found on the packaging. If you threw it out, try looking up your RAM's specs on newegg, or check the manufacturer's website.

[james_8970]

Yup, I worded that wrong, better edit that. Default on my board is 1.8V, but considering you both have a 680i board, they are probably the same sitting at 1.85V.
James