SUMMARY: If you want to accurately measure heatsink temps, here’s a How-To that can get you going for less than $100.
I’ve had requests from both websites and individuals for more details about how I measure heatsink temps, so I’ve written up a How-To that can get you started for under $100. I’m sure there are other instruments around that will do the job as well, but I’m most familiar with Omega, so I am using their products as an example.
You will need the following items:
- Digital Thermometer
- Thermocouple Wire and Plugs
- Quick Setting Epoxy
- Lapping Materials
The heart of the system is a digital thermometer which uses thermocouples to read temps. Not only is it accurate, but the thermocouples allow you to place probes just about anyplace you can fit a wire. I have used to to measure temps of a heatsink’s surface, the back of a video card, ambient intake air, etc. It’s a very flexible system.
The Omega HH23 is a primo instrument, accurate to within about 0.1 C. This particular model comes with a Calibration Certificate attesting to its accuracy. Cost: $219. Now there is a cheaper model called the
Omega HH11 which goes for $65. It only takes one probe at a time and is not a calibrated instrument, but is accurate enough to give you repeatable readings. Fluke is another company that makes similar instruments.
Since you’re going to epoxy the thermocouples into the heatsink, buying pre-made probes will drive you to the poorhouse. What’s neat about the Omega system is that you buy thermocouple wire and make your own probes; you can make them as long or as short as you need. Once you use it, if you don’t need to measure the heatsink again, you can cut the wire and re-use it in another one.
As shown above, the wire is very thin, so you don’t need big holes; I use a 1/16″ hole. It comes on a 25′ spool which gives you about 12 probes 2′ long. Making the probe is pretty simple:
- The top picture shows the wire as is. Take a razor blade or Xacto knife and split the wire down the middle.
- The middle picture shows the wire split. Note that there is a clear plastic sheath covering the two inside wires – one red, one blue. Discard the plastic excess and strip the two wires about 1/2″.
- Twist them together and solder, as shown in the bottom picture. You just made a thermocouple probe!
With one end as the probe, strip the wires on the other end and mount it on this plug which is inserted into the thermometer. Note that there is a + and – terminal; the copper wire goes to the + terminal and the silver to the other. Don’t mix them up – you will get false readings!
Next step is to prepare the heatsink. You will need to drill into the heatsink to do this. There are three ways to mount the probe, but let’s concentrate on Intel’s recommended method for now: Drill a hole in the center of the heatsink where it contacts the CPU core. If you have a Dremel, I found that the Dremel Drill Press is perfect for this. You can use a regular hand drill also.
When you do this, you may have to drill the hole at an angle, because the clip may be right under the center hole. It helps to have a drill press vise – I found one at Home Depot for about $15 and it works great. I’ve been using 1/16″ drills because they are readily available and robust enough so that they don’t break easily, something of concern when drilling metal.
Next step is mounting the probe into the heatsink. I use 5 minute epoxy. Insert the wire into the hole, coat about 1″ or so with epoxy and pull it into the hole, dragging epoxy into the hole as you do so. You may do this a couple of times to get the hole filled with epoxy. Don’t worry about getting some on the top of the heatsink – you’re going to lap it anyway.
I make sure a little bit of the thermocouple wire sticks up above the hole, so that when I lap it the thermocouple is exposed to the CPU core, as shown above.
And that’s it! Now you’re ready to mount the heatsink, run a stress test like Prime 95, and see how it does. The industry method for measuring heatsinks is to calculate C/Ws. Basically, it’s a measure which accounts for ambient temps and processor heat output, in watts. More details on this can be found HERE and HERE. One of the big advantages of C/Ws is that it enables comparisons among different reviews.
I live very close to Omega, so I can pick up materials right away. If you use Omega products, you can buy on-line or call direct – they are very helpful and you can talk to engineers about the products:
Omega Hand Held Thermometers: Check out models HH23 and HH11.
For thermocouple wire, order TT-T-30-SLE for “T” type;
For thermocouple wire, order TT-K-30-SLE for “K” type.
It runs about $27 for 25′ of wire, enough for 12-18 thermocouples depending on how long you make them. Make them about 24″ long, so when you cut the wire to reuse it, you have enough. You’ll also need male SMP connectors – the blue clips pictured above; they’re $1.75 each and six is enough to get started – you just move them around as needed. If you have a K type thermometer, ask for K type clips; same for T type.
The difference between K and T types is the temperature range; K runs from something like -200 C to 1372 C while T from -200 C to 400 C; for our purposes, T is fine.
Here are three ways to mount the thermocouple:
- Drill a hole in the center of the heatsink (Intel);
- Drill a hole in the side of the base (AMD);
- Mill a groove in the bottom of the heatsink base.
1. Drilling a hole through the center of the base puts a thermocouple on top of the CPU core. First place the heatsink on the CPU with grease, then remove and note where the center of the CPU rests on the base. Drill a hole (I use 1/16″) at this point, thread the thermocouple through it and epoxy in place. Leave a little bit of it above the hole; after the epoxy dries, clip it as close as you can, then lap it smooth.
Problems you might encounter:
- You can’t drill straight up because you’ll hit the clip. In this case, you have to drill sideways, which is more difficult and results in a larger hole. If you don’t do it right the first time, it gets messy very quickly.
- You make a mistake and drill the hole way off-center. If you hit the extreme edge of the CPU, you may get a lower reading than at the center, throwing results off.
- It’s impossible to drill a hole straight up or sideways due to the heatsink’s design – a good example is the ThermoEngine. In that case, I had to drill a hole into the side, then drill a hole straight up to meet the side hole. If you miss, gets real messy and you may ruin the heatsink for testing purposes.
- Lapping means that you alter the heatsink’s base. A lot of times, you actually make it slightly better, as many heatsinks have extrusion marks on them. For heatsinks that precision lap the base (e.g, Swiftech), lapping makes it worse. The effect is most likely minor, but it’s there.
2. Drilling a hole in the base parallel to the CPU avoids the issues listed above; however, drilling a small, deep hole is more difficult than one that goes just through the base. You have to use a LOT of thin oil to lubricate the bit as it gets deeper. I use Marvel Mystery Oil and it works great (BTW: Use it for any hole you drill). I drill for about 1/4″, remove the bit, put a drop in the hole and continue. So far, I have not had a drill bit stick.
Problems you might encounter:
- The drill bit sticks in the hole – somehow you have to get it out. You could start drilling from the other side, then you have removed even more material from the base; gets messy very quickly.
- On one heatsink, you’re 1/8″ above the CPU. On another, due to heatsink design, you have to be 3/8″ above the CPU. As you increase the distance from the CPU, you begin to skew results. How much? I can’t tell you without testing each heatsink, because there are too many factors you can’t control that come into play.
3. Milling a groove in the heatsink’s base means you remove more material directly over the CPU than any of the others. You have to make the thermocouple as thin as possible to keep the groove small and shallow. There’s no doubt in my mind that this is the most “sensitive” of the options listed – it’s action reminds me of what I see using Intel’s thermal diode.
Problems you might encounter:
- Lapping issues identified above.
- The amount of base surface area impacted is far greater than the other options. The more you alter the heatsink, the more potentially you influence results.
The main objective is that the measurement method should not effect results. Whatever method you choose, consistency in methodology is very important – doing the same thing every time makes you better at whichever method you choose and makes your comparative results meaningful. I use a notebook to log all tests so I can always refer back to my notes if needed.
I plan to do some further testing by using each method on three of the same design heatsinks and comparing results. Based on some preliminary work, I think AMD’s method is the most likely to give results that are furthest from the CPU’s core temp, due to the distance from the core.
220.127.116.11 Processor Case Temperature Measurements
To ensure functionality and reliability, the Pentium ® 4 processor is specified for proper operation when T CASE is maintained at or below the value listed in the Pentium 4 processor in the 423-pin Package Datasheet. The measurement location for T CASE is the geometric center of the IHS. Figure 8 shows the location for T CASE measurement.
Special care is required when measuring the T CASE to ensure an accurate temperature measurement. Thermocouples are often used to measure T CASE . Before any temperature measurements are made, the thermocouples must be calibrated. When measuring the temperature of a surface, which is at a different temperature from the surrounding local ambient air, errors could be introduced in the measurements. The measurement errors could be due to having a poor thermal contact between the thermocouple junction and the surface of the integrated heat spreader, heat loss by radiation, convection, by conduction through thermocouple leads, or by contact between the thermocouple cement and the heat sink base. To minimize these measurement errors, the following approach is recommended:
1. Prepare 36 gauge or finer diameter K, T, or J type insulated thermocouples.
2. Ensure that the thermocouple has been properly calibrated.
3. The thermocouple should be attached at a 90° angle to the integrated heat spreader and the heat sink covers the location specified for T case measurement.
4. Drill a hole 0.150 inches (3.8mm) maximum diameter through the heat sink base. This hole must be positioned on the heat sink base so that it matches with the center of the IHS when assembled. This hole will reduce the heat sink performance by approximately 0.02 °C/W.
5. Create a small depression, approximately 1/16 inch (1.5mm) in diameter by 1/64 inch (.4mm) deep at the center of the IHS (see Figure 8). This will facilitate the attach procedure by keeping the thermocouple centered and hosting the adhesive.
6. Route the thermocouple wires through the hole in the heat sink base and attach it to the processor IHS. The use of more viscous adhesives and minimizing the use of drying accelerators will prevent problems with the adhesive spreading.
7. A small fixture may be required to hold the thermocouple and apply a steady force during the curing process to ensure the thermocouple is making contact with the IHS. A Digital Multi-Meter can be used to check continuity between the IHS and the connector as the adhesive cures.
8. Make sure there is no contact between the thermocouple adhesive and heat sink base. Contact will affect the thermocouple reading.
9. Verify the cured adhesive bead is smaller than 0.15 inches (3.8mm) in diameter and height so as to fit in the hole drilled in the heat sink base. Trim as necessary.
10. Place the TIM on the heat sink base. If it is a semi-liquid type apply it on the IHS around the thermocouple. The clamping force will spread the TIM. If the TIM is a solid type, punch a 0.15inch (3.8mm) diameter hole in the center of the TIM pad and cut a line from a side to the hole. This will allow the installation of the TIM to the IHS with the thermocouple already attached to the IHS