• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Thermistor placement question

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Mictlan

Senior OC BOINC User
Joined
Apr 17, 2001
Location
Mexico City, Mexico
After a lot of thinking I think I have a good idea to get my thermistor over the chip die without too much compromise of the heat transfer.

If I use a 1/4" copper cold plate, I could bore a small hole in the middle of the cold plate so that the thermistor could be placed over the die of the chip (in the case of the CPU die, just by the side). I can epoxy the thermistor so thats insoletad of the cold plate. I'll get with all this an insolated thermistor that could (theoretycally) read the actual chip die temperature with minimal compromise to the heat exchange.

All your comments are welcome.
 
That will work. Any time you drill a sensor hole, either straight through or parallel to the face of the baseplate or coldplate, you will impact the thermal characteristics of them. You are right to try to isolate the sensor from contact with the baseplate or coldplate as they contribute to a slightly lower reading by wicking away some of the heat in sensor. One other issue it the addition of a thermal barrier by adding a cold plate. You are going to have a device sitting on top of the core, with thermal compound. You then have to couple that device to your HSF through another layer of thermal compound. If the HSF is Aluminum, you will benefit from better heat spreading by the copper coldplate, but probably give most of that benefit up by the loss of a second thermal barrier. I once added a 1/4 inch 50mm copper coldplate to my old FOP-32. Even with a good coating of AS between them and both having been lapped, I only realized about 1-2C improvement. Improvements in that range can just as easily be explained away as coincident with a better application of grease, better seating upon the core. IE insignificant, so I'm not sure if the coldplate should get the credit for a 1-2C change.

Hoot
 
I'm not installing the coldplate for incresed performance. I'm really aware of the termal resistance increase that a new layer of material will bring to the HSF device. The main objetive of the coldplate will be the ease of installation of the thermistor. It will be easier to install it on the coldplate, and if a mess up it wont be a $30+ heatsink but in a $1-2 scrap of metal. ALso if I upgrade with the installation of a pelier, then I'll be able to use the same coldplate :). If I install it on the heatsink after installing the peltier the reading will be of the hotside of the peltier, not the chip.

Well, I have been thinking also how do I join the heatsink and the coldplate? And if I install the peltier how do I sandwich it? Should I bolt it? Or glue it? I can use AS adhesive, but I think that would be kind of permanet....

Any help will be appreciated. And when I finish I promise to show some pics of the experimient and some results.
 
Believe me, when you get that cold plate slathered up with AS and stick it to the HSF base, the problem will be getting them apart! ;D After you squeeze the two together firmly and the air and excess AS oozes out, they form a very effective vacuum between them. In the case of a peltier device, the recommended clamping pressure is way in excess of the force applied by the retainer clip. In that case, you want to drill and tap the HSF base for four screws, on the corners of the cold plate (oversized) so you can cinch it up tight to the HSF base. Look at how Danger Den does their units, for a good example. Again, I slather AS on both sides of the peltier and wipe off the excess, that oozes out, once you clamp the whole assembly together. Generally, it's a messy, but necessary process to get the peltier operating as efficiently as possible.

Hoot
 
An even better solution is an on die diode like intel uses and the athlon 4 will have :)
Measure core temp with something external is never going to be completely accurate, for a start you aren't measuring the actual silicon temp only the nickel (intel) and what ever it is amd uses, plus differant parts of the chip will be slightly differant temps at differant times, for example I muc doubt the integer unit will be working as hard as the fpu/sse/3dnow units during a game.
 
Yep, you're right. I'll be measuring the die temperature not the core temperature. But I have a lot of uncertanty with my mobo reading. I have 2 internal readings one is named processor and the other motherboard by the BIOS, but I just see 1 thermistor going out of the mobo. As I try to be cautionous with my CPU I want to know the die temperature and try to keep it as low as possible.

In my opinion if the die temperature is low, the average chip temperature will be low.

Well, after all the input I have decided to do the following:
1.-I will solder a extension to the inside bottom of the hestsink.
2.-I'll use an oversized coldplate.
3.-I'll put some AS II (no adhesive just the regular thermal paste) between the HS and the coldplate.
4.-I'll bolt them up.
5.-I think I'll do a hole thru the coldplate to get the clip from the heatsink to the chip in the case of the CPU.
6.-For the GPU, I'll just use the bolt that I have.
7.-This will work also for a peltier.
 
Back