Edit: Okay, internal diode finally calibrated

[Spoooon]

Okay, after much work, I've finally done it and done it right. At 467, my full load temp was 36 and my intake air temp was 23. At 938, my full load temp was 42 and my intake air temp was 23. To measure intake air, I reversed the fan on my Pal and stuck a thermistor about an inch above the fan. To get my full load temp, I used CPUBurn and ran it until temps had peaked.

[(42-7)-23]/[(36-7)-23]=2

 

[phigamman]

The simple answer is this: your mobo holds heat better than the air inside your case. As the core CPU temp goes up, more hot air blows around inside the case. The efficiency of your airflow effects how long the hot air stays inside before new (ambient room temp) air is pumped in. With perfect airflow, the rising core temp will register minimal increases in case temp as hot air is immediately blown out. Since this is impossible, you get a measurable increase in case temp. Hope this helps! ;-)

 

[Mikewarrior2]

Hmm... intersting, maybe you hit hte "point of no return" for your cpu/case cooling. As in, your case temp was hit a point where cooling was best, and anything lower would have minimal effect.

The other thing you may want ot try is to calibrate your internal diode readings.

here's a link: Intel Internal Diode Calibration.

Also, one last thing, which mb are you using?


Mike

 

[Spoooon]

I'm using the CUSL2-C. Well, my case temp is room temp because I have plenty of fans to bring cool air in and exhaust hot air.

edit: Calibrating now

 

[Mikewarrior2]

the only other thing i cna think of is to measure where the cpu fan inlet air is... not sure how to do that on an alpha set to "Suck", though.


Mike

 

[Spoooon]

What I ended up doing was opening the case up and directing a fan at it, which, at least I think it does, causes the inlet air temp to be room temp. Using the temps 41/23(959mhz) and 36/23(467.72mhz), I ended up with a -8 calibration. Would my PIII at 466 idle at room temp? Do you think I calibrated everything correctly? I can't believe that my cpu temp would be that low.

 

[Mikewarrior2]

Hmm.... One second, I would redo the calibration doing one test at 66mhz bus(for 466mhz) and another test at 133mhz bus(for 933 mhz).


Mike

 

[Spoooon]

Doing the test again. At 66.8(the nearest to 66 that will happen apparently), I get 31,23 idle and 36,23 full load. At 134(again, the closest to 133 that will happen), I get 31,23 idle(strange?) and 40,23 full load. This time around, I end up with 9 being the compensation.

938.01/467.72=2.01

[(40-9)-23]/[(36-9)-23]=2.00

So, compensation should be 9. But if I look at the idle temps, that would give me a temp below room temperature. I'm going to change some fans around (inspiration just struck me) and see what happens.

 

[Mikewarrior2]

Hmm....

I would try a different program to load your chip... I would also try measuring ambient air temp at different spot, since you're "technically" supposed to be measuring the temp of air that the heatsink/fan is using to cool itself.

I would recommend CPU burn, available here: CPU Burn.

It bumps up my processor temp a few C(around 2) higher than RC5.


Mike

 

[Spoooon]

Okay, this time I used CPUBurn. I let it run until the temp stabalized. To record the air in temp, I held my other thermistor beside the fins(but not touching) on each side of the HSF. I recorded the high temp as my air in temp.

938.01
45/24(previously with Prime 95 40/25)

467.72
39/24(previously with Prime 95 36/25)

[(45-9)-24]/[(39-9)-24]=2.00

Prime 95 temps
[(40-7)-25]/[(36-7)-25]=2.00

I think that I will take the low compensation just to be on the safe side. Do you think I did everything right? I guess this gives me an excuse to buy a new heatsink just for the heck of it.

 

[Mikewarrior2]

Weird.. the compensation should be teh same for both heatsinks. You compensate your particular cpu-diode to your mb, and that compensation should work for any heatsink.


Mike

 

[Spoooon]

It's probably just because of the way the fan is on the PAL6035. I think I got it about as close as I could without resorting to turning the fan around.

 

[Nevin]

Spoooon,

To get an accurate calibration, you really need to know the heatsink air intake air temperature. This is why the article specifies that the fan must be blowing into the heatsink and also why the procedure does not work on Orb type coolers. Since an Orb type cooler or a cooler with a sucking fan will be pulling in air at different temperatures at different parts of the heatsink, there is no way to come up with an accurate temperature for the fan intake air. Since this value is a critical part of the formula, if you do provide an accurate temperature, the calibration will not be accurate.

Nevin

 

[Spoooon]

Well, what about where I said I opened up the case and had a fan blowing into it? Would that not cause the intake air to be room temperature no matter what? If the goal is to get an accurate measurement of the intake air, I'll just force the intake air to be at room temperature.

edit: One thing: what about voltage? I left the voltage the same when I switched the bus from 133 to 66. Should I have changed it to whatever voltage is required for stability?

 

[Nevin]

Sorry Spoooon, it's not that easy.

There are lots of components in the computer that produce heat, blowing a fan into the computer will move the heat from the hot components inside the case around as well as injecting room air.

The heatsink fan has to be blowing in to get a valid calibration.

edit- And the voltage must remain constant. Changing it would introiduce an additional variable.

Nevin

 

[Spoooon]

Okay guys, I know that resistance is futile. I turned the fan around, I've got my extra thermistor about an inch above the fan. I guess it's time for me to do it the right way.