[JoeC]

Introduction

Here with this new round of field testing we are introducing our new thermal compound Perihelion. It’s been nearly 2 years since Innovation Cooling’s (IC) test run of IC Diamond here on OCF. While OCF was not the largest of any of the 12 Field tests done to date, it was certainly one of the most thorough of the groups. So with that in mind I selected this group to do the first public tests as it was the most challenging, results from the OCF IC Diamond testing assisted us to effectively trouble shoot a number of issues like proper application amounts and mounting and hardware requirements needed to obtain best performance in addition we were able to refine our marketing message. Forum comments and feedback we received were invaluable. OCF gets first look at our new compound IC Perihelion prior to product release.

So Thanks again all who agreed to participate and those who have helped us in the past.

Background


We are taking a little bit of a different track on things by adding a maximum possible performance metric for consideration rather than just a Paste #1 vs. Paste #2 sort of thing. We are fast approaching material limitations (if not already maxed out) and performance improvements from here on out will be harder to differentiate. More and more we project or expect that future buy decisions will be made on factors such as cost/quality with competitive compounds performing within a few 10th's of a degree of each other with the exception what I call the exotics like diamond and the liquid metals

To demonstrate this we soldered a sink to a test die (one inch Die @120 W) and with a thermocouple we set through a hole in the sink to measure the top of the joint to establish a maximum performance "solder line" for a practical theoretical limit to performance. In our own study there were no significant gains to be had in the thermal cascade of die, compound, sink. We are at the beginning of the end of the performance race with a pile up at the material limit wall.


The Graphics below (Courtesy of Intel) outline how we test.



We test ambient temperature at four points an in addition we measure airflow through the sink drawing a flow/ resistance curve resistance curve and also test the fan by itself. Standard adjustments are made for temperature, barometric readings and altitude.

For thermal measurements we follow Intel's suggested method.



For the synthetic die we just replace the CPU with a 1" copper die. Synthetic test set up provides more information such as to how much power is applied to get the thermal reading and is much easier to control for mounting contact/ pressure repeatability in testing. We then use in-system-tests for a correlation to synthetic tests.


Chart below - samples are the newly released paste compounds which are pretty good performance wise. I tested all the recently released compounds in the last year the top 3 are of those shown here. So in the future it is possible we will have 10 compounds all testing inside of a degree of each other. Future product reviews of compounds will probably start with some statement that the paste has hit the standard performance mark and then move on to other features. Ultra High performance is the new average now.

Test comparison below was done on a synthetic one inch die @ 120 W



We believe this is where thermal compound becomes kind of a ho hum commodity product with cost/quality(long term reliability) becoming more dominant features in the buy decision as noted. The design target for Perihelion fills a gap in our product line for Cost/Quality/Performance in as close to an optimum balance as we could get.

What remains is to see how our lab data translates to in-field use by actual users

Test Requirements


Be sure to benchmark your cpu temps with your current tim and record your results before using perihelion for testing. we need before and after please!

Also, please make sure you don't change any variable other than the thermal compound used for before/after (i.e. same volts/clocks, cooler, and other hardware)

APPLICATION METHOD


For all those participating in the Perihelion giveaway, please follow this link to the Innovation Cooling website for application instructions on the product application is the same as IC Diamond:

Perihelion Application Instructions

Proper application is the key to this product. The nature and consistency of this product is likely quite different from what most of you are used to in a Thermal Compound. We don't recommend use of traditional methods of applying the TIM like razor blades, lines, x's (though some people have reported great success with the X method), etc. Through extensive testing, the best overall application method has been determined to be a pea sized amount (approx. 5.5mm dia.) directly in the center of the CPU. The TIM should be adequately spread via the downforce of the CPU cooler (a sufficient amount of downforce will provide best results; minimum 50psi of downforce should be the goal for optimum results. please see the Innovation Cooling website for more info). Also, for coolers/surfaces that have many peaks and valleys, it may be beneficial to wet the surface(s) with a small amount of Perihelion to initially fill the voids in the mounting surfaces...then proceed with recommended application as usual. Please click on the link provided and shoot me a PM if you have any questions.

Keep in mind, Perihelion has been formulated with the specific goals of providing great thermal conductivity with short set times, but also for long term stability and performance without pump/bakeout through extended thermal cycling or usage. This stuff is very stable over long periods of time.


HOW TO POST RESULTS




THE FORMAT IS AS FOLLOWS (ALL TEMPS IN C):

OPTIONAL INFO: PROCESSOR, VOLTS/CLOCKS, HEATSINK/COOLING USED, ETC.

Pre-Perihelion
Compound:
Ambient Temp:
Idle Temp:
Load Temp:
Heat Sink Fan Speed:

Perihelion
Ambient Temp:
Idle Temp:
Load Temp:
Heat Sink Fan Speed:

If you have any problems or need support, please feel free to post here for now. Remember, proper application and good even heat sink pressure (goal should be minimum of 50psi or more) are crucial for peak performance. Too little compound will impact your results so please do visit the Innovation Cooling website for Application Instructions with images and explanations as well as read the review on this site as. For those with direct touch heatpipe coolers, you might want to try wetting the contact surface of the cooler with a tiny amount of Perihelion to fill any small voids before proceeding with the normal application method.

So, I'd just like to thank OCF for this great opportunity and also all the participants and members. If anyone has any questions, please feel free to PM me. Sign ups are still open and should be posted in the sign up thread.

Thanks all !

Application Notes

We strongly recommend using our application procedure out of the gate and no others.

http://www.innovationcooling.com/app...structions.htm

Correct Application Amount

the application should look like this with a 5.5mm bead - This was tested by us and validated on a number of forums (including this one) using less will impact results negatively as shown in multiple tests.

Cauliflower like extensions indicate absorption of air. The more viscous the paste the more you have to use. With thinner or less viscous pastes the "Air Zone" is further extended .

The metric is viscosity, amount of paste to the area being covered.

Our new labels will include a 5.5mm "reference spot" to facilitate proper application.

In this example IC Diamond was applied between two glass slides and stress tested in an oven @ 150C for 20 hours. Applied 5mm bead of ICD.

Note the halo of air bubbles, this is why it is important to apply enough compound, because as the paste thins it spreads it reaches a maximum zone where air is reformed into the paste application between the sink and IHS

In essence you need enough compound to extend the air pocket zone outside the area of IHS contact otherwise you are riding on a cushion air(an insulator) preventing full contact/performance and leading to an early failure of the mount

http://www.innovationcooling.com/app...structions.htm

[Bobnova]

You must have smaller peas around your part of the country, round here peas are more like 8-9mm. Big ole suckers.
5.5mm still seems like an awful lot, but i'll do it exactly by the book for the tests, then test again with a smaller amount.

[muddocktor]

Actually, I have found that a pea sized 5-5.5 mm blob in the middle has always given me the most repeatable results too, Joe. I am presently making my first benchmark run with OCCT/Linpack with Ceramique and a Copper TRUE right now. I plan to make 2 runs like that, shutting down between runs to heat cycle the Ceramique to see if I have any difference between the runs. I will then do that with both the products you sent us to test with also.

EDIT: After seeing Bob's post directly below and not wanting to make another post in this results thread, for the metric challenged folks a 5.5 mm blob corresponds to a 7/32" blob in the American old skool measuring system. I still visualize better in feet and inches than in meters and millimeters. at me.

[Bobnova]

Like i said i'll give it a whirl.
I certainly wouldn't call it pea sized though, not by a long shot.

Just for reference for anybody without a MM ruler, the tip of the cap for both products sent out to test is 5.5mm on the section before the ribs.

Is the 5.5mm blob still the prefered method of HDT heatsinks?

I intend to test with my HDT-1283 as well as my stock e5200 Intel HSF.

I also plan to test it on my GF's laptop, that may not happen though.

[JoeC]

Originally Posted by Bobnova

Like i said i'll give it a whirl.
I certainly wouldn't call it pea sized though, not by a long shot.

Just for reference for anybody without a MM ruler, the tip of the cap for both products sent out to test is 5.5mm on the section before the ribs.

Is the 5.5mm blob still the prefered method of HDT heatsinks?

I intend to test with my HDT-1283 as well as my stock e5200 Intel HSF.

I also plan to test it on my GF's laptop, that may not happen though.

On our label we are going to have an application reference dot so users can just match their application amount to the dot on the tube

[JoeC]

Also have another ongoing at XS for those that like to compare results

http://www.xtremesystems.org/forums/...d.php?t=243824

[Bobnova]

Originally Posted by JoeC

On our label we are going to have an application reference dot so users can just match their application amount to the dot on the tube

Very cool idea, more manufacturers should do that.

[onefstsnake]

Test Results on i7 920 D0 w/ Corsair H-50

Testing was done with Prime 95 small and large FFT's. Each compound was applied as a pea sized glob.
Each compound was heat cycled. 5min 100% load, 5min idle, 10min 100% load, 10min idle, 15min 100% load, 15min idle. After 15 min cycle compounds were at 100% load for ~1 hour.

Compound: Shin Etsu X23-7783D
Ambient Temp:21*c
Idle Temp:34*c
Load Temp:69*c
Heat Sink Fan Speed: N/A
______________________________________________

Compound:IC Diamond
Ambient Temp:21*c
Idle Temp:35*c
Load Temp:71*c
Heat Sink Fan Speed: N/A
______________________________________________

Compound:Perihelion
Ambient Temp:21*c
Idle Temp:37
Load Temp:76*c
Heat Sink Fan Speed: N/A

[psionic98]

Results from an i7 920 d0 @ 3.8 OC w/ Thor's Hammer 2x scythe s-flex fans in push/pull

All tests done with Prime 95 with large FFT test only. Each compound done according to manuf. suggestions. The 3 compounds are Arctic Silver 5, IC Diamond, Perihelion. I followed similar testing procedure as onefstsnake: Cycled processor to 100% for 5minutes, idle for 5, 15minutes at 100%, 10 min idle, 15min 100%, 10 min idle, then 2.5 hours at 100% for final temp readouts. Ambients unfortunately ranged during this from 20-21c (68-70F). Results are as follows and is average temps of the 4 cores and fans controlled off fan controller at 100%:


Compound 1: Arctic Silver 5
Ambient Temp:20c
Idle Temp:36.5c
Load Temp:78.2c
Heat Sink Fan Speed: 100%
______________________________________________

Compound 2:IC Diamond
Ambient Temp:21c
Idle Temp:36.8c
Load Temp:77.6c
Heat Sink Fan Speed: 100%
______________________________________________

Compound 3:Perihelion
Ambient Temp:21c
Idle Temp:36.4c
Load Temp:76.8c
Heat Sink Fan Speed: 100%


Afterthoughts: While I did't see vast improvements on idle, only ~0.1 degree, on 100% load my temps dropped ~1.4 degrees. Overall it seems like a good product, but I'll see how it fares in the summer months when the ambients are 30+c daily.

I also testing 2 compounds on my Asus 5770 oc'd to 1000/1375 using atitools 3d-cube to run up temps (its not the best test, but it does raise temps!) I ran this for 20minutes per compound with 10min cleandown/reapply time between..no "heat cycling" The following results are from my gfx card:

Results of ASUS 5770 overclocked/overvolted to 1000/1375 @1.212volts:

Compound: Stock clay-like substance!
Ambient Temp:21c
Idle Temp:32c
Load Temp:73c
Heat Sink Fan Speed: 100%
______________________________________________

Compound:IC Diamond
Ambient Temp:21c
Idle Temp:31c
Load Temp:71c
Heat Sink Fan Speed: 100%
______________________________________________

Compound:Perihelion
Ambient Temp:21c
Idle Temp:29c
Load Temp:68c
Heat Sink Fan Speed: 100%

Seems this worked out alot better on my gfx, but again the testing isn't the best method..but gaining ~5degrees is nice when its overclocked and overvolted

[Bobnova]

Preliminary results (testing right now, posting via netbook) perihelion is ~5*c hotter then Arctic Alumina Ceramique under IBT load. Next up is re-testing with my preferred application method rather then the specified one. After that is IC7, and after that if i don't keel over from lack of sleep, is AACeramique again, to test it fairly rather then when it has had a chance to heat cycle a few times (and boy did it, including to <-15*c). I suspect that the 5.5mm method and HDT coolers don't get along well.

I'm taking pictures of the TIM coverage, too.

EDIT: This is interesting stuff to work with, much thicker then Arctic Alumina Ceramique.
It cleans up very easily with 70% alcohol, which is nice.

Test results to follow later tonight or tomorrow sometime.

[JoeC]

Originally Posted by Bobnova

Preliminary results (testing right now, posting via netbook) perihelion is ~5*c hotter then Arctic Alumina Ceramique under IBT load. Next up is re-testing with my preferred application method rather then the specified one. After that is IC7, and after that if i don't keel over from lack of sleep, is AACeramique again, to test it fairly rather then when it has had a chance to heat cycle a few times (and boy did it, including to <-15*c). I suspect that the 5.5mm method and HDT coolers don't get along well.

I'm taking pictures of the TIM coverage, too.

I suspect that you are right about that and separate instructions for application are required for the HDT's

[JoeC]

Originally Posted by psionic98

Results from an i7 920 d0 @ 3.8 OC w/ Thor's Hammer 2x scythe s-flex fans in push/pull

All tests done with Prime 95 with large FFT test only. Each compound done according to manuf. suggestions. The 3 compounds are Arctic Silver 5, IC Diamond, Perihelion. I followed similar testing procedure as onefstsnake: Cycled processor to 100% for 5minutes, idle for 5, 15minutes at 100%, 10 min idle, 15min 100%, 10 min idle, then 2.5 hours at 100% for final temp readouts. Ambients unfortunately ranged during this from 20-21c (68-70F). Results are as follows and is average temps of the 4 cores and fans controlled off fan controller at 100%:


Compound 1: Arctic Silver 5
Ambient Temp:20c
Idle Temp:36.5c
Load Temp:78.2c
Heat Sink Fan Speed: 100%
______________________________________________

Compound 2:IC Diamond
Ambient Temp:21c
Idle Temp:36.8c
Load Temp:77.6c
Heat Sink Fan Speed: 100%
______________________________________________

Compound 3:Perihelion
Ambient Temp:21c
Idle Temp:36.4c
Load Temp:76.8c
Heat Sink Fan Speed: 100%


Afterthoughts: While I did't see vast improvements on idle, only ~0.1 degree, on 100% load my temps dropped ~1.4 degrees. Overall it seems like a good product, but I'll see how it fares in the summer months when the ambients are 30+c daily.

I also testing 2 compounds on my Asus 5770 oc'd to 1000/1375 using atitools 3d-cube to run up temps (its not the best test, but it does raise temps!) I ran this for 20minutes per compound with 10min cleandown/reapply time between..no "heat cycling" The following results are from my gfx card:

Results of ASUS 5770 overclocked/overvolted to 1000/1375 @1.212volts:

Compound: Stock clay-like substance!
Ambient Temp:21c
Idle Temp:32c
Load Temp:73c
Heat Sink Fan Speed: 100%
______________________________________________

Compound:IC Diamond
Ambient Temp:21c
Idle Temp:31c
Load Temp:71c
Heat Sink Fan Speed: 100%
______________________________________________

Compound:Perihelion
Ambient Temp:21c
Idle Temp:29c
Load Temp:68c
Heat Sink Fan Speed: 100%

Seems this worked out alot better on my gfx, but again the testing isn't the best method..but gaining ~5degrees is nice when its overclocked and overvolted

Some variability in mounts as I would expect the IC Diamond to outperform the Perihelion.

Just to let everybody know I calculate performance from the temperature increase over ambient.

So ICD -1.6C /Perihelion -2.4C

[Bobnova]

Ok here we go. Pictures will be posted tomorrow, i need to get to bed.
I'm leaving the last mount of Perihelion in place overnight folding@home to see if a few hours of heat help the temps at all, then tomorrow night i will re-test this application, then pull the HSF off and get pics of the coverage.


All tests are done with a Xigmatek HDT-S1283 heatsink and a low speed "turbine" fan, the brand of which i do not remember. The fan speed reporting seems to go in notches of a dozen or two rpm, hence all the fan speed numbers being exactly the same.
Ambient temp was measured just above the fan 1.5" in front of the fan, via an infrared no-touch thermometer scanning a piece of tape strung from fan to case. Not the most conventional of methods, but accurate. Plus i don't have a thermometer to do a more normal method.
Idle tests let the cpu sit at XP32 desktop with all closable programs closed for five minutes (extensive testing has shown me that five minutes is all it takes for this cpu/hsf).
Load testing is done with Intel Burn Test's "standard" test for five loops.
The final "idle" numbers are the lowest recorded by Realtemp.
The final "load" temps are the highest recorded by Realtemp.

The ambient temp of my working area was a slow curve downward, as the house is cooling off for night time. The heater never ran and no doors/windows were open, so it was a very steady decline as seen in the ambient temp readings.
The first ambient temp listed is during the idle test.
The second ambient temp listed is during the final pass of IBT.


Arctic Alumina Ceramique:
Fan: 1016-1054rpm
Ambient: 22*c
Idle: 26/24
Load: 63/63
Ambient: 21.5
This mount had excellent coverage.


Perihelion, official 5.5mm ball method:
Fan: 1016-1054rpm
Ambient: 20.5
Idle: 25/24
Load: 67/67
Ambient: 20.5
Note, when load was removed the core temps dropped much faster then with Ceramique, for whatever that is worth.
This mount was found to have lousy coverage.


Perihelion, lots of little dots on heatpipes method (ideally lines down the heatpipes, i found this to be impossible due to Perihelion being very thick and not very sticky), this is the method i've found to work best on this cooler for AA Ceramique:
Fan: 1016-1054
Ambient: 20*c
Idle: 24/23
Load: 61/61
Ambient: 20
This mount had excellent coverage.


Perihelion, Official Method take two (5.5mm wide 5.5mm tall pile):
Fan: 1016-1054
Ambient: 20*c
Idle: 24/23
Load: 63/63
Ambient: 19.5
This mount has unknown coverage at this time.


Conclusion: Perihelion is more difficult to work with then AA Ceramique, but when applied in such a way as to get good coverage on a HDT type HSF, it works better then Ceramique. The official installation method does not seem to work well with this cooler. To be fair, the official method for Ceramique is roughly the same, and it doesn't work well either.
I recommend as close to lines down the middle of the heatpipes as possible, i found that ~4 lots of 3.5mm wide by .5-1mm tall dots down each heatpipe worked well.

Testing with IC Diamond will be done as soon as possible, as well as re-testing Perihelion after time at high temp. I may also do a third install using the official method and a larger pile of TIM, say 7mm, depending on how the second attempt at the official method looks when i remove it.

I'd like to say thanks to JoeC and Innovation Cooling for the chance to test this TIM, this is my first review/test for a company and I've rather enjoyed it

[muddocktor]

My preliminary results:

Test bed system consists of:

• Antec 1040 clone case, with rear exhaust fans modded to 92 mm fans.

• Asus P5Q Pro mobo

• Q6600 at 3600, 1.48-1.49 v vcore under load

• Copper TRUE heatsink with 2 Sanyo Denki San Ace 109R1212H1011's in push-pull, rated at 102.5 cfm and 39 dB. The heatsink is using the TRUE Rev. C LGA775 mount with a Swiftech backplate and the heatsink has a strip of duct tape applied across the top of the base for more clamping pressure.

Procedure:
Processor and heatsink are cleaned off with 50/50 IPA/Xylene, which is an industrial chemical I get from work. This completely cleans the old TIM off as it will cut oil and water based compounds and leaves no residue of it's own. After cleaning, I apply the prescribed amount of TIM, then mount the heatsink and alternate tightening the mounting screws until they bottom out. I then run a 30 minute OCCT run, then shut down the computer for 10 minutes to heat cycle the TIM. I then reboot and run another 30 minute OCCT run then shut down for another 10 minutes to heat cycle. I then reboot again and use the Linpack option in OCCT to run a full hour run. The Linpack option loads the cpu much more than the regular OCCT run option, which should be good for testing the various TIM materials. After the first Linpack run completes, I shut the system down for 10 minutes for one last heat cycle, then reboot and run Linpack once more for an hour. Ambient temps were kept between 20.55-21.11 degrees C (69-70 F) during the tests. Temps were measured in front of the case with a digital thermometer and there is a box fan blowing at the case to keep from having any heat buildup around the case.

Results:
Arctic Silver Ceramique:
CPU1 max temp - 75 C run 1, 73 C run 2
CPU2 max temp - 75 C run 1, 74 C run 2
CPU3 max temp - 69 C run 1, 67 C run 2
CPU4 max temp - 68.75 C run 1, 67 C run 2

Perihelion:
CPU1 max temp - 76.5 C run 1, 76.5 C run 2
CPU2 max temp - 75.5 C run 1, 77 C run 2
CPU3 max temp - 72 C run 1, 73 C run 2
CPU4 max temp - 72.5 C run 1, 73 C run 2

Updated with run #2 results of the Perihelion TIM.

Overall, I am seeing a little worse results so far with the Perihelion product, compared to Ceramique. I did not see any "break in" with the Perihelion, unlike the Ceramique. I will try the IC Diamond this afternoon.

Additional results:

IC Diamond:
CPU1 max temp - 73 C run 1, 74 C run 2, 24 hr burn-in-run3 - 72 C
CPU2 max temp - 73 C run 1, 74 C run 2, 24 hr burn-in-run3 - 72 C
CPU3 max temp - 68 C run 1, 68 C run 2, 24 hr burn-in-run3 - 67 C
CPU4 max temp - 67.5 C run 1, 68 C run 2, 24 hr burn-in-run3 - 67 C

I just finished 2 runs with the IC Diamond, and I am seeing better results with it than with either the Perihelion or the Ceramique. It's not a giant difference but is measurable. Application of the IC Diamond was a little more difficult than with either the Ceramique or the Perihelion but not hard by any means. I found that the Perihelion and the Ceramique to be about even in applying them to the cpu accurately, with the Perihelion being ever so slightly more fluid.

[JoeC]

* Copper TRUE heatsink with 2 Sanyo Denki San Ace

109R1212H1011's in push-pull, rated at 102.5 cfm and 39 dB. The heatsink is using the TRUE Rev. C LGA775 mount with a Swiftech backplate and the heatsink has a strip of duct tape applied across the top of the base for more clamping pressure.

Curious as to the use of duct tape, first this time I have seen it used in a test. Is there a underlying problem with the hardware that prevents full contact contact or pressure? how much does the duct tape improve your temps? you must have seen an advatage or you would not us it.

I use crude mechanical aids like rubber straps or weights sometimes to enhance a mount but I would think that duct tape would be hard pressed to overcome the mechanical stresses in a high pressure mount but I could be wrong.

[psionic98]

if using industrial duct tape, if strapped right, could easily add 15+lbs of pressure.. if its the good stuff.. theres better than that 150+mph stuff sold at lowes/home depots

[JoeC]

True, but the nature of adhesives and adhesive tapes are highly dependent on contact area. To maintain 15 lbs you would need a fair amount of surface area, maybe I am not visualizing it right but I do not see where that contact area comes from in a typical mount or how you would get the leverage in such a short span to manage more than a couple of lbs.

I'll take a look at it later and see if I can find the info to calculate what it would take to add extra pressure maybe I'll learn something, might be technique to it. Point though was it needed to correct a problem?

[rge]

@bobnova, muddocktor,psionic98, onefstsnake,

Nice job and thanks for being so thorough, now if everyone tested and reported like that, I would agree with personal testers giving an accurate picture of tims performance.

[sobe]

Setup:
Antec P182 (2 Scythe S-Flex 1600, 2 S-Flex 1200)
Q9550 E0 @ 3.6GHz (for stock volts) on Xigmatek HDT-S1283.
Gigabyte EP45-UD3R
TIM spread method used as shown on last example from here,
http://benchmarkreviews.com/index.ph...1&limitstart=5


Ambient: 20c
_______________________
Arctic Silver 5:

Idle:
Core #0 - 35c
Core #1 - 35c
Core #2 - 34c
Core #3 - 34c

Load:
Core #0 - 65c
Core #1 - 63c
Core #2 - 56c
Core #3 - 55c
_______________________
IC Diamond 7:

Idle:
Core #0 - 31c
Core #1 - 32c
Core #2 - 30c
Core #3 - 29c

Load:
Core #0 - 61c
Core #1 - 59c
Core #2 - 52c
Core #3 - 54c
_______________________
Perihelion:

Idle:
Core #0 - 38c
Core #1 - 36c
Core #2 - 35c
Core #3 - 35c

Load:
Core #0 - 67c
Core #1 - 65c
Core #2 - 58c
Core #3 - 57c
_______________________


Using ICD7 and I must say I'm highly impressed with the results over Arctic Silver 5.

[muddocktor]

Joe, the duct tape trick is something I learned with the original scissors type TRUE mount. As you know, the original mount had a bad problem holding the heatsink from rotating on top of the processor. I figured that it was due to the bracket having so little contact area between it and the top of the heatsink base. I put the tape originally to give a stickier than metal to metal contact patch between the top of the heatsink base and the scissors bracket and let the bracket "dig in", so to speak. And it works well in that regard. But I also noticed that I started getting lower load and idle temps after doing this too. When you add a strip of the duct tape across the top of the heatsink base (the part that the clamping bracket contacts on the base), the layer of tape also acts to thicken the base that the bracket has to clamp against, giving more clamping pressure. How much more I really can't tell you since I don't have the sophisticated equipment necessary to check the clamping pressure but it does increase it somewhat. And just to double check that the better temps weren't due solely to the heatsink not rotating on top of the processor when clamped, I carefully had mounted the TRUE without having any heatsink movement and ran a stress test, then did the duct tape trick and retested and saw a decrease in temps using the duct tape.