thermal diode calibration

[skymaster]

I'm getting high temps on a 2500+. 48C idle at stock, 43 when underclocked to 1000MhZ, 54 under load at stock, more when oc'd. Mobo is NF7 rev1.2. HSF is Glacialtech silent breeze 462, which is by the way completely inaudible. Now here's the question: would you figure this may be due to bad calibration of the diode? I.e. the increase in temp (load - idle) is a very standard 6C, so you wouldn't think the heatsink isn't up to the job; I therefore suspect that the displayed temperature is simply off by some 10C due to badly calibrated diode. Comments?

 

[phsstpok]

Actually 54 degrees sounds about right.

Digit-Life.com tested a bunch of heatsinks and found the GlacialTech Igloo 462 Silent Breeze to have a thermal resistance of 0.52 C/W. Since the Barton XP2500+ has a nominal heat dissipation of 53.7 watts (more if overclocked). This means you should get a deltaT of 27.9 (53.7 x 0.52) degrees C or about 28 degrees above your case temperature. This implies your case temperature is about 26 degrees (54 degrees load temp minus 28 degrees deltaT = 26 degrees).

54 degrees is not bad for a true core temperature reading using a nearly silent cooler.

<b>99% is great, unless you are talking about system stability</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 05/05/03 05:42 PM.</EM></FONT></P>

 

[skymaster]

Thanks for the reply; it is much appreciated. The calculation explains it all, and as you say the digit life review has higher temps even. Then I wonder why some people in the forums claim they have temps in the 30s on air. Maybe their diode isn't quite well calibrated then.
Now that I'm at it, what would you figure is a stable overclock I could get out of my set-up? Or basically, what kind of temps should get me worried?
Tried 185x11 but crashes after a few minutes under full load. May be the board rather than the chip though. HardOCP reviewers had similar issues with their board at first; they claim burn-in increases oc potential. Do you believe that?

 

[phsstpok]

Thanks for the reply; it is much appreciated. The calculation explains it all, and as you say the digit life review has higher temps even. Then I wonder why some people in the forums claim they have temps in the 30s on air. Maybe their diode isn't quite well calibrated then.

Many people have motherboards that don't read the XP's thermal diode instead they have a thermister in or near the CPU socket. Other mobos can read the thermal diode but this is only available in BIOS while some monitoring software can only read the motherboard sensors. Just a few reasons for varying readings. However, mostly people with really low readings have HSF's that are more suited to overclocking.

Generally, overclockers' HSFs will have a more power fan than your Silent Breeze. Instead of the 22-25 CFM fan that you have they will have 40 CFM, 60 CFM, even deafening 85 CFM fans.

Now that I'm at it, what would you figure is a stable overclock I could get out of my set-up? Or basically, what kind of temps should get me worried?

I wouldn't worry about temps at stock speeds but as I indicated if you want to overclock you should have chosen another heatsink. Silent and low powered isn't ideal for overclocking.

I can't tell you what temps are safe because it varies from system to system, overclock to overclock. Too many variables; CPU model, amount of overclock, core voltage, case cooling, differences in sensors, HSF, etc.

To give you an example, on my own rig, I needed to keep my CPU below 50 degrees (socket sensor). That was my old Tbird 1.0 @1.5 Ghz, 1.85 volt. Since my load temp was normally about 48 degrees I had very little headroom. If the room gets to warm, my CPU temps go up, and my system becomes unstable Summer was a problem. I usually reduce my overclock and use a room fan to blow extra air into my case.

Cutting it that close is probably not what you would want to do.

My current CPU is a Tbred B XP1700+ @2.1 Ghz. My overclock seems to be limited by voltage and not so much by temperature. I simulated summer and increased my room temp by 10 degrees. Core temps reached 56 degrees. System remained stable yet I can't overclock any higher (normal room temp) without using extreme voltage. 2.1 Ghz needs 1.775 volt but 2.25 Ghz needs 2.15 volts just to POST.

Hope this example makes my point.




<b>99% is great, unless you are talking about system stability</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 05/06/03 08:11 PM.</EM></FONT></P>

 

[skymaster]

Mm, I see. So that doesn't look too good for me on the overclocking front then; if it's the cpu that's holding me back, upping the voltage isn't really an option as I guess it'll increase heat generation quite a bit. I'll see if it really is the cpu by oc'ing with the multiplier only.
Stability and silence is my main aim though; my stuff runs for several days on end anyway and 5 hours more or less isn't going to make a difference. Just going to replace that amazingly annoying northbridge fan (it's the loudest component in my system by a margin) by one of these zalman jobbies.
Thanks for the examples; it's really helpful!