PDA

View Full Version : milling heatsinks for temp sensors???



p27182
08-17-2007, 04:02 AM
intel suggests it in the datasheets for their chipsets, but i didnt see anything like it for the cpu...

has anyone dones this?

if i had a fancy cnc/etching machine i'd do it, but i would expect the inconsistencies from scraping with a dremel or hand tool would decimate the thermal conductivity between devices...

yer thoughts?

thanks
-p

wildcard
08-17-2007, 05:24 AM
it could probably be done with unsophisticated equipment, ie: a sharp drill bit or something similar. There isn't anything that guarantees success. Why not try it on a cheap/stock heatsink?

virtualrain
08-17-2007, 08:28 AM
I'm not sure what the benefit would be? You can easily put a temp probe in your water somewhere and your CPU already has a thermal probe in the core. The temperature you measure with a probe on the bottom of the heat sink will be somewhere between the two, but very close to water temp I suspect.

However, if I were to do it, I would simply use a small file on edge to file out a notch that goes from the edge to the center of the block as deep as your probe/wire.

Waterlogged
08-17-2007, 08:35 AM
Why would you want to put something that doesn't transfer heat too well between the chip and heatsink? ;)

p27182
08-17-2007, 08:56 AM
the following is taken straight from the intel p965 datasheet:

******************************************
Temperature differences between the temperature of a surface and the surrounding
local ambient air can introduce error in the measurements. The measurement errors
could be due to a poor thermal contact between the thermocouple bead and the
surface of the package, heat loss by radiation and/or convection, conduction through
thermocouple leads, or contact between the thermocouple cement and the heatsink
base (if a heatsink is used). To minimize these measurement errors a thermocouple
attach with a zero-degree methodology is recommended.

Thermocouple Attach Methodology
1. Mill a 3.3 mm [0.13 in] diameter hole centered on bottom of the heatsink base.
The milled hole should be approximately 1.5 mm [0.06 in] deep.
2. Mill a 1.3 mm [0.05 in] wide slot, 0.5 mm [0.02 in] deep, from the centered hole
to one edge of the heatsink. The slot should be in the direction parallel to the
heatsink fins (see Figure 3).
3. Attach thermal interface material (TIM) to the bottom of the heatsink base.
4. Cut out portions of the TIM to make room for the thermocouple wire and bead.
The cutouts should match the slot and hole milled into the heatsink base.
5. Attach a 36 gauge or smaller K-type thermocouple bead to the center of the top
surface of the die using cement with high thermal conductivity. During this step,
make sure no contact is present between the thermocouple cement and the
heatsink base because any contact will affect the thermocouple reading. It is
critical that the thermocouple bead makes contact with the die (see
Figure 2).
6. Attach heatsink assembly to the (G)MCH, and route thermocouple wires out
through the milled slot.
************************
(end of quote)

I would like a software-less reading personally, after seeing my computer show me 4 different temperatures from different softwares, all reporting different values more than a full degree off for the same variable.

I was thinking about using a basic stamp or fpga and designing my own monitoring system that would tell me everything about my computer's current state (without consuming ANY of my computer's resources), but at the moment im far to busy with school and work to start a hefty personal project like that...

im really just curious of what some of you guys thought, i was surprised to see it in intel's datasheets.... maybe im not reading em right....

http://www.intel.com/design/chipsets/designex/313055.htm

page 15, take a gander and tell me if im crazy...

-p

wildcard
08-17-2007, 09:14 AM
I say go for it. There is nothing stopping you from doing it and if done right, it shouldn't impact performance very much.

Waterlogged
08-17-2007, 09:49 AM
It looks like your reading that right but again, do you really want to remove material from the most critical place of your heatsink for something that doesn't transfer heat? Doesn't make a whole lot of sense to me.