OCUK Official IC Diamond/ Perihelion Test Results

I am awaiting contact & pressure paper(s) so I can optimize contact in one go and don't have to keep taking apart my PC. I have some Zalman ZM-STG2 thermal compound aswell which I can use to see how the pattern is turning out and posting pics here,before actually using the papers.

This will also be useful when comparing the IC Diamond pattern which I am using as it has been about 2 weeks now since my last post although both compounds have different density/viscosity.

As I am also approaching my final testing I would want to make sure that my last testing is as accurate as possible hence there are still some factors which I need to discuss before actually carrying out the testing.

I am thinking about my previous tests as a stepping stone and as a learning experience for myself in that some mistakes were pointed out and yielded some further understanding about the whole process from an end user.

However I am surprised the thread has gone quiet now.
 
So mine turned up at last (Big thanks to Innovation cooling for resending :)) and I've done some testing! All temperatures are measured as degrees above ambient temperature during the test, ran Prime95 for 30 minutes with the highest temperatures recorded by Realtemp being used.

Arctic Silver 5 (Old paste): 52 58 60 55
IC Diamond (Pre-curing): 50 56 57 53
IC Diamond (Post-curing): 47 52 54 51

Average temperature drop of about 5 degrees. Very happy. :)
 
WingZero30 said:
I am awaiting contact & pressure paper(s) so I can optimize contact in one go and don't have to keep taking apart my PC. I have some Zalman ZM-STG2 thermal compound aswell which I can use to see how the pattern is turning out and posting pics here,before actually using the papers.

This will also be useful when comparing the IC Diamond pattern which I am using as it has been about 2 weeks now since my last post although both compounds have different density/viscosity.

As I am also approaching my final testing I would want to make sure that my last testing is as accurate as possible hence there are still some factors which I need to discuss before actually carrying out the testing.

I am thinking about my previous tests as a stepping stone and as a learning experience for myself in that some mistakes were pointed out and yielded some further understanding about the whole process from an end user.

However I am surprised the thread has gone quiet now.
Click to expand...

Thread is getting a little long in the tooth. I'll write up the final report in the next couple of weeks and will update it with new info as happens.

This is one for you WingZero30 now you are the resident C/P expert - Below laptop chip 1.3 sq centimeters of area @ 35 W - so what happens to a compound when you lose 25%-50% if the contact area on a chip this size?

laptop_2620m_raw_pressure_image2.png
 
Last edited:
IC Diamond said:
This is one for you WingZero30 now you are the resident C/P expert - Below laptop chip 1.3 sq centimeters of area @ 35 W with 1.3 sq centimeters of surface area - so what happens to a compound when you lose 25%-50% if the contact area?

laptop_2620m_raw_pressure_image2.png
Click to expand...


I can only give my opinion based on what I have observed so far:p

Depending on the pressure applied and surface profile of mating surface between the cpu chip and heatsink, we would expect a fine thin layer of thermal compound or bond line thickness to fully cover the CPU IHS under 100% contact. Ofcourse greater the pressure applied, more thinner the BLT layer will be. However as far as I understand there is a diminishing return when the pressure is applied above a certain point at which BLT doesn't get any thinner and has achieved the desired layer thickness.

However with a reduction of contact area of around 25-50%, we can expect the thermal compound to stay 'wet' in that region or in it's initial state when applied from the tube. It will also be thicker than BLT. So BLT will only be formed in the region of contact area.
Furthermore the 'wet' region of compound will form a barrier or a 'heat dam' as mentioned previously resulting in poor heat transfer.

However reduction of 25-50% contact area doesn't generally result in 25-50% increase in the cpu temps :)
 
No IHS on this one, so direct contact area is smaller so the watt density is higher.

Kind of like here

http://forums.overclockers.co.uk/showpost.php?p=18767784&postcount=285

The compound had an early failure

The C/P image above is from a laptop, our first to date. We are in the early stages of gathering long term data on the laptops and you have real diverse mix of temperature experiences some of the compounds tested here only last from a few weeks to a year. IC Diamond seems OK at a year to about a year and half give or take with only one at 2 years contrasted with a guy over at Anandtech who is 4 years on ICD on a PC.

The increased watt density is a killer and I suspect the early failures are low C/P in numbers something like the Pareto principle aka the80-20 rule

I was taking a look at some published information on laptop reliability linked below and a rough analysis might be with their 30% failure rate in 3years approx 1/3 of that due to breakage and about 2/3 or 20% of all laptops is a component failure probably due in large part to high thermal loads - Heat is number one enemy of electronics.


After a re-paste on a PC in the overclockers forums you have an average of 61C after a repaste and the laptops so far average 71 C after a re-paste. I think the extra 10C is the dividing line, in the PC market you may see a few degrees decline and here and there you get the rare 15 -20 C degree improvement but generally most of those are mount related issues unless the compound has been installed a couple of years or more.

The PC failures were generally less than half laptops. The prevalence of early temperature creep in laptops is quite surprising along with the relatively large number of 85 - 90 C systems reported---- 25 to 30 C higher than the PC's, if you have a weak component those kinds of temps have a good chance of killing it. As a kind of general recommendation to avoid being the 1 in 5 failures and as a good maintenance procedure has been suggested here blowing the dust out regularly, I would add a secondary laptop cooler as well as changing the compound every-time it creeps up 10 C past the initial install temps. The cooler it is and the longer time periods it remains cool, the odds that longer system life increases.


http://www.squaretrade.com/htm/pdf/SquareTrade_laptop_reliability_1109.pdf
http://cache.gawker.com/assets/images/4/2009/11/500x_squaretrade3.jpg
Gartner Says Annual Failure Rates of PCs Are Improving, but Manufacturers Can Do Better
Click to expand...
 
Last edited:
Lots of good info obtained. Raw images reveal much as well as pictures of paste impressions

I cannot find the exact full die size but then have not looked too hard yet but initially pressure is very good as I suspected it would be and contact appears to be around 50% - 60%, 9/10 th's of a centimeter

So basically you have half the area for dissipating 35W - +2X the thermal stress on a compound in an environment that is already warmer than most to begin with, little wonder why people are experiencing accelerated compound degradation.

I currently have 3 people on tap for C/P laptop testing it would be nice to get a broader profile of 20 to 30 and to see how things trend.



NOTEBOOKRAWIMAGE1.png

PressureStatistics1.png

imageslaptopCP

PseudocolorRepresentationAndPressureStatistics1.png

PressureHistogram1.png

LineScan.png

LineProfile1.png


3DImage1.png
 
First of all here is IC Diamond pattern after 3 weeks of usage (initial IBT heating, browsing, heavy gaming etc) made by the usage of 'balanced shim' as discussed previously where the left and right surface of shim was balanced.

The shim has been super glued to the washer.

imag0255e.jpg




IC DIAMOND (3 weeks usage)

3weeksicdiamond.png



3weeksicdiamond2.png




Inverted pic of pattern on the heatsink, so top and bottom in both pics align when heatsink is placed on top of cpu

3weeksicdiamond3.png





IC DIAMOND (Applied today)

Applied for 10-12 mins. It can definitely be seen that IC Diamond does require some curing time as it is still more 'wet' than the 3 weeks pattern.

todayicdiamond.png



todayicdiamond2.png





ZALMAN ZM-STG2 (Applied today)

Applied for about 10-12mins. Less viscous than IC Diamond and seems to have reached optimum BLT quicker than IC Diamond

todayzalmanzmstg2.png


todayzalmanzmstg22.png



IC Diamond seems to show again slight bias towards right in terms of more pressure. However Zalman seems to show balanced pattern. Possible difference may be due to compounds viscosity, mounting and application amount of both compounds.


However here is the 'balanced shim' contact and pressure pattern :(

patterns.png



The pattern seems to have become more balanced. However it is strange that everytime I use a new paper, I get different pattern.

I suspect the reason is again the backplate/motherboard thickness which isn't uniform.

The backplate isn't fully tight to the motherboard even though it has been tightly screwed to the bracket mounts. There is still slack which may have been facilitated on purpose.
Backplate only becomes fully tight to the motherboard when IFX-14 is secured to the bracket mounts via mounting plate.

The bracket mounts act as an interface between the heatsink and backplate.

It is exactly the same case with my 775 system as shown with this person's 1366 socket system. He stated that he has fully tightened the screws but still there is slack.



I obtained this pattern by placing the board on my table which I believe is flat enough.

The last 'good' pattern I obtained was either due to placing of a folder underneath the motherboard or using a support such as a pen underneath the motherboard near the PCI slots, to ensure motherboard remained as straight as possible.

The washer is around 1.3-1.4mm thick. Shim seems to have added about 0.5-0.8mm on top. So that is 2.1-2.2mm depth of the mod. A stated before ,around 3mm it becomes impossible to secure the IFX-14. So I am guessing there is a space for a further 1 -2 paper cutouts.

My computer is still all open (surgical phase :p) as I am using my brother's PC to type this post. Plus I have two papers still left.

Further suggestions?
 
Looking at the CP image along with the paste patterns the "glazed" top and bottom are the high contact and pressure areas. The larger "globular" compound areas line up as low contact and pressure.

Bottom more pressure than top and middle region ls now the low contact area.

Perhaps with the added pressure there is some flex in the IHS? Sink?

Try the patience of the best mechanic - not easy...

Contact kind of follows a tripod alignment with the contact mostly on the three highest points.

pastevsCPmage.png
 
As seen in the youtube video, the backplate tended to move slightly up and down by few mm. The bracket mounts are screwed to the backplate via hollow pillars. Hence when the backplate moved up and down, this also raised or lower the bracket mounts by few mm.

When the motherboard is placed on any hard object such as table, book etc, this pushed the back plate against the motherboard, raising the bracket mounts. This also meant lower pressure on the mounting plate hence on the Cpu IHS, as the UNC spring screws which connect the mounting plate to the bracket mounts didn't have to go down much further.

When the washer was added between back plate and motherboard with the paper shim facing the motherboard, the slack + movement of back plate was gone:). This also meant that bracket mounts couldn't raise thus forcing the UNC spring screws to go down further, hence exerting greater pressure on the Cpu IHS.

mod5.png





mod3h.png




I also decided to add another shim consisting of 4 rectangular cut outs of A4 printing paper and sticking it on top of the IFX-14 base.


mod4.png


Previously we allowed the backplate slack with the washer above motherboard. So when bracket mounts raised, the washer already compensated for this by raising the mounting plate subjected to washer + shim depth or thickness.
Now washer is below the motherboard with slack gone. So although there is nothing to elevate the mounting plate except for the new rectangular shim, the bracket mounts can't raise either and are subjected to washer + shim depth or thickness.
In both cases, I think the end result would have been same. However the only addition is the rectangular shim in the new case, so slightly increased depth of the mounting system.



When I obtained the second pattern, unfortunately the paper slipped from underneath the heatsink as these papers tend to stick to the bottom of heatsink, after having taken off the heatsink. So I can't be sure of true orientation of pattern, but have taken two screenshots of the orientation which I think are the most likey of the pattern. I can take a pic again in the morning:)


mod1k.png



mod2o.png




Now the thing is tightning the UNC spring screws down fully is becoming a very challenging job as resistive force from the spring on the screws is increasing due to adding more shims. I can may be add one or two more paper cut outs of the rectangular shim and that's it.

Early on I actually tried the rectangular shim with 8 layers of paper cut outs, but it became impossible to force the UNC screws down into the bracket mount hole and was a reminiscent of the 3mm washer scenario.

After this I can't think of any further options. The uniformity of all the materials flatness is definitely somewhat a cause of concern. Even the washer I believe won't be 100% flat at the surface level.

If the flatness of all the crucial variables such as paper shims, washer, Cpu IHS, heatsink base, backplate, motherboard pcb, etc were to be measured with a laser device, I am pretty confident that they would all reveal irregularities perhaps not at mm level but definitely at micrometer level and beyond, and such have been discussed previously.


patternscomparison.png
 
Last edited:
More clearer comparison

patterncomparison2.png





I roughly measured Q6600 IHS (lapped square part that makes contact with heatsink) and it is 2.8cm x 2.8cm = 7.84cm^2 = 1.215 sq.in

q6600ihsdims.png



The rough measurement of the better pattern is 2.7cm x 2.3cm = 6.21cm^2 = 0.963 sq.in

goodcontactdims.png




This would equate to roughly 0.963/1.215 = 79% (almost 80% contact)

Of course these are my rough estimates and may well be wrong.

However I have noticed that in my previous sensor product results, the total area is always quoted around 2.78-2.91 sq.in. How do they calculate this value?
 
Last edited:
Phenomenologica said:
So mine turned up at last (Big thanks to Innovation cooling for resending :)) and I've done some testing! All temperatures are measured as degrees above ambient temperature during the test, ran Prime95 for 30 minutes with the highest temperatures recorded by Realtemp being used.

Arctic Silver 5 (Old paste): 52 58 60 55
IC Diamond (Pre-curing): 50 56 57 53
IC Diamond (Post-curing): 47 52 54 51

Average temperature drop of about 5 degrees. Very happy. :)
Click to expand...

Hmm... I should try this on the PlayStation3 40GB model. I currently have Arctic Silver 5 on it. If I could shave off a few more degrees for it, it might help not kicking the fan up in this warmer weather. Normally it's silent in a cool room but with the warmer weather lately it's kicking in quite a bit.

Might try and lap the copper plates while I'm at it. Quite heavy machine grooves on it.
 
However I have noticed that in my previous sensor product results, the total area is always quoted around 2.78-2.91 sq.in. How do they calculate this value?
Click to expand...

They just take the paper size assuming that the person submitting the sample is aware of the total contact area.

As as general rule I usually check the Itel supplied dim. which I believe is 30mm X 30mm

I expect from your print you are in the 60%+ range - Send back your best print and I will get a lab analysis done.

I think right now you are kind of on the ambient noise of the mechanicals little tighter, little looser, little to the right or left.

Without more lapping you are where you are give or take with pressure probably the more dominant factor
 
IC Diamond said:
They just take the paper size assuming that the person submitting the sample is aware of the total contact area.

As as general rule I usually check the Itel supplied dim. which I believe is 30mm X 30mm

I expect from your print you are in the 60%+ range - Send back your best print and I will get a lab analysis done.

I think right now you are kind of on the ambient noise of the mechanicals little tighter, little looser, little to the right or left.

Without more lapping you are where you are give or take with pressure probably the more dominant factor
Click to expand...


As I am tightning the heatsink, mounting pressure is increasing. I have noticed that once the heatsink has fully tightened, the backplate has slight tendency to warp around the corners plus the motherboard also warps slightly. These factors tend to slightly produce bias in the contact pattern where highest pressure is around the edges.

If only motherboard and backplate (plastic) were made out of steel then pattern may well have been more accurate. Unfortunately these factors are beyond my control and as stated before; variables are more difficult to control with my setup as an end-user.

This coupled with the fact that Q6600 IHS is slightly concave in the middle. I think as seen in the lapping post, when I started lapping Cpu, the edges got lapped first suggesting that they were at a slightly elevated level.

Even with hand lapping some bias still remains with regards to the surface of both Cpu IHS and heatsink base. However belt sander may well have given much finer results. But since hand lapping is widely known in the overclocking community, I think it still does a very good lapping job:p.

As for dimensions of Q6600 IHS, I found them to be definitely between 2.8-2.9cm each side, so I guess this is close enough to 3cm x 3cm intel spec.

I will post the pattern hopefully by tomorrow :)
 
Finally!!! Some results from me :-) Many thanks to IC Diamond for bearing with me... I have to fully remove my motherboard to get my cooler on and off, so these are just initial results with no 're-torquing' just yet, after I applied it last night.

C2Q Q6600 (VID 1.212V) O/C 3.4GHz @ 1.325V VCore (1.40V BIOS/ 1.36V @ idle in windows / 1.325V @ full load in windows
Cooler Master 212 with 2 sickleflow 2000RPM fans at full speed.

Prime 95 Small FFT's for max heat (and it's what I used before)

Previous Paste: A/C MX-2

A/C MX-2
Ambient: 20C
Idle: 47 (Core 1)
Load: 73C (73, 72, 70, 69)

New Paste: I/C Diamond 24 Carat.
I/C Diamond 24 Carat
Ambient: 21C
Idle: 45C (Core 1)
Load: 69 (69, 69, 65, 65)


So on average so far, taking the delta into account: 5 degree C drop. And that's without a re-torque.
 
Just back from vacation -WingZero30 I did get your CP results and will have processed next week currently battening down for a little tropical squall this weekend

Lost my home PC due to lightening strike(s) So a couple days reconfiguring all and should wrap this up test results next week.
 
You must log in or register to reply here.
Back
Top Bottom