Innovation Cooling's Diamond 7 TIM test results

[tastymannatees]

I had a quick chat with Andrew over PM. He asked me to post this since I noticed this while mounting my heatsink. The stock Intel heatsink, as we all know, uses push-pins to tighten the cooler to the board. Obviously, no motherboard is perfect. One of these small manufacturing errors can mean that the board has slightly larger mounting holes, or mounting holes that are slightly off-center. From time to time I have seen boards that are bent around the CPU area, obviously because the cooler applied a lot of pressure on the board. This can either be a cooler or a motherboard manufacturing error.
When mounting the stock Intel cooler on this P5AD2-E board, it took me much, much longer to mount than on my Striker Extreme (say 5 minutes vs. 1). On the P5AD2-E mobo, it seems the cooler is much more firmly attached than it was on my Striker Extreme. From the looks of it, it seems the mounting holes on the Striker are slightly bigger as well. Also, when using push-pins rather than a backplate, it could be that the cooler isn't touching the CPU perfectly, something which is less of an issue when using a backplate. This can result in different load temperatures.
__________________

This was a great observation by Martijn. I Have not been getting great results with push pin configurations on the whole, usually equal or +1 or 2C which I was attributing to low mounting pressure due to hardware fatigue and other factors and never relating it to MB mechanical differences that was causing a pressure issue. Something to take note of and see if we can get further confirmation.

This is the kind of stuff that we can use in our troubleshooting guide. You Guys are doing great work, above and beyond what I expected. I'll have to work out a (modest) bonus for the group when we get to completion.

[Emerica]

This was a great observation by Martijn. I Have not been getting great results with push pin configurations on the whole, usually equal or +1 or 2C which I was attributing to low mounting pressure due to hardware fatigue and other factors and never relating it to MB mechanical differences that was causing a pressure issue. Something to take note of and see if we can get further confirmation.

This is the kind of stuff that we can use in our troubleshooting guide. You Guys are doing great work, above and beyond what I expected. I'll have to work out a (modest) bonus for the group when we get to completion.

If you are looking for better mounting pressure with the stock intel coolers with the LGA775 push-pins you can get one of those Thermalright LGA775 bolt-through kits.

[littleowl]

If you are looking for better mounting pressure with the stock intel coolers with the LGA775 push-pins you can get one of those Thermalright LGA775 bolt-through kits.

hum that is a good idea but that don't help with the stock hold down from intel.

[64dragon]

you can't pull the pins out of the stock intel and use the screws and backplate from the thermalright? why wouldn't that work?

[Bail_w]

you can't pull the pins out of the stock intel and use the screws and backplate from the thermalright? why wouldn't that work?

Yes you can.

[jspace]

I got one of those samples a few days ago. Haven't had the time to tear any systems apart due to school, work, etc. I'll see if I can get to it tomorrow or Monday night. I also have 3 more systems I think I will test.

[CyberDruid]

I know I am late to the party but I have been Using the IC Diamond TIM for a coup;le of months now independent of these tests. I purchased mine through Petra's to try out and now highly reccomend and use it in most applications.

I found I needed to use more than I would of the X23 Shin but the results surprised me. I am used to putting a tiny amount of TIM on in order to get good results...I can really blob on the Diamond Paste and still see better than Shin Etsu Performance.

It's good stuff.

[littleowl]

you can't pull the pins out of the stock intel and use the screws and backplate from the thermalright? why wouldn't that work?

yeah you could do that! it would be a good test to see if they get good enough preasure without bending the stock hold downs or something!


EDIT: so has anyone used this stuff with phase yet????

[Jaco]

EDIT: so has anyone used this stuff with phase yet????

Good question. Does ICD7 still work at -50°C ?

[tastymannatees]

The only thing I would consider might be going straight to -50C without time for the paste to spread or level to the optimum bondline thickness it might jut thicken up to much to spread. Maybe let it sit cold for a while, 4 or 5 hrs? mount and put a space heater or hairdryer on it for ten min.?

We ran a test ICD7 in a thermal cycling chamber - from -40 C to +85 C - for up to 250 cycles (with the copper block heated with 130 W of power) and found no change in thermal performance / thermal resistance of the grease

In a separate test we pulled "full vacuum" on ICD7 with a mechanical pump for ca. 65 hours at room temperature. The resulting ... weight loss was almost immeasurable - 0.3 mg out of 6.2222 g. This works out to be 0.0048 wt %.

Another test we ran ICD7 for 1000 hrs @125 C with no change in thermal performance.


So I think performance would hold at -50. The obstacle might be getting a good bondline at the start

[David_L6]

Here's one that I ran three times (yes, 3 times! That's why it has taken me so long to get these results) because I didn't believe the results the first time. This Diamond 7 is a very good TIM.


Q6600 @ 3.4GHz with a Zalman 9700 heatsink.

with AS5

59, 58, 58, 53

with Diamond 7

54, 54, 54, 49

[SiGfever]

Two Week Cure:

I had posted my Final Results after one week of cure time but I find this interesting that my temps are degrading.

Ambient = 22.7c


Idle:
Core 0: 34
Core 1: 34
Core 2: 32
Core 3: 31

Load:
Core 0: 45
Core 1: 45
Core 2: 46
Core 3: 44


I am going to pull the mount, inspect, and reapply TIM according to coverage pattern.







I am wondering if I have too much pressure forcing your TIM off the mount? I will remount using the kit that came with the FuZion which has "limiters" in the springs to prevent over tightening. I will re-run for one week to test.

[tastymannatees]

SiG,

This is a good candidate for the pressure test Looks like you have more pressure in the upper left on the last sink pic and a kind of pooling of thicker paste on the rounded edge lower right. If the sink was flush you would have no accumulated paste or the rounded edge, it would have squared off under pressure. Hard to tell from the pics, pressure film will indicate whether I am right or wrong on this.

It does seem Like the "Fan" out from one corner type mounts are giving higher temps though.

[SiGfever]

SiG,

This is a good candidate for the pressure test Looks like you have more pressure in the upper left on the last sink pic and a kind of pooling of thicker paste on the rounded edge lower right. If the sink was flush you would have no accumulated paste or the rounded edge, it would have squared off under pressure. Hard to tell from the pics, pressure film will indicate whether I am right or wrong on this.

It does seem Like the "Fan" out from one corner type mounts are giving higher temps though.

Remember that FuZions are "convex" which will force out TIM.

Start up: 04/20/08 16:45:00
Ambient = 23.3c

Idle:
Core 0: 34
Core 1: 34
Core 2: 32
Core 3: 31

Load:
Core 0: 46
Core 1: 47
Core 2: 47
Core 3: 46

[tastymannatees]

Remember that FuZions are "convex" which will force out TIM.

I assume the convex feature is a product of the casting? This was a problem I had when I was trying to source boilers for our thermosyphon I could not get a sample that was not convex so I changed manufacturing methods rather than go to an extra sanding process.

If what Intel says is true that you need 43lbs to make the IHS "flat" this feature would tend to minimize edge contact leaving a contact spot in the center of the IHS instead. This on an initial basis with the light contact in littleowl's post (243) with the pressure film fuzion test, does not seem to be happening - high contact at the edges and much less contact on center with a biased pressure on the right side.

If the convex design is one of intent, you have to help me here as I am not seeing the advantage

[SiGfever]

I assume the convex feature is a product of the casting? This was a problem I had when I was trying to source boilers for our thermosyphon I could not get a sample that was not convex so I changed manufacturing methods rather than go to an extra sanding process.

If what Intel says is true that you need 43lbs to make the IHS "flat" this feature would tend to minimize edge contact leaving a contact spot in the center of the IHS instead. This on an initial basis with the light contact in littleowl's post (243) with the pressure film fuzion test, does not seem to be happening - high contact at the edges and much less contact on center with a biased pressure on the right side.

If the convex design is one of intent, you have to help me here as I am not seeing the advantage

Because a lot of IHS are bowed they started making the better waterblocks convex to overcome this.

[tastymannatees]

Because a lot of IHS are concave they started making the better waterblocks convex to overcome this.

It just does not make sense to me, well mechanically anyway. If you are not on center you have a problem, if IHS and sink have different diameters, sizes, depths not to mention what happens between the IHS and die....I'll have to pick up a couple and run a dial indicator on them along with some pressure film to see what's going on. Maybe I am getting a little thick in my old age.

[SiGfever]

It just does not make sense to me, well mechanically anyway. If you are not on center you have a problem, if IHS and sink have different diameters, sizes, depths not to mention what happens between the IHS and die....I'll have to pick up a couple and run a dial indicator on them along with some pressure film to see what's going on. Maybe I am getting a little thick in my old age.

Here is some good reading...

http://www.xtremesystems.org/forums/...&highlight=ihs

http://www.xtremesystems.org/forums/...&highlight=ihs

[littleowl]

Here is some good reading...

http://www.xtremesystems.org/forums/...&highlight=ihs

http://www.xtremesystems.org/forums/...&highlight=ihs

but the fuzion is not convex untill you put in a bigger O ring!?

[SiGfever]

but the fuzion is not convex untill you put in a bigger O ring!?

Yes it has a slight convex shape, the o-ring just bows it more.

[littleowl]

Yes it has a slight convex shape, the o-ring just bows it more.

really? I was under the impreasion that the fuzion was flat unless you put the O ring to convex it. I will have to do some looking at mine when I get the time to do the next set of tests.

[tastymannatees]

Here is some good reading...

http://www.xtremesystems.org/forums/...&highlight=ihs

http://www.xtremesystems.org/forums/...&highlight=ihs

Ok, I have the concept down, essentially you are putting the same amount of force, overall increasing the psi in a smaller area over the die. It is quite clear in the grease pattern a single broad line down the middle with a thicker coat on either side. Looks like it would be tricky to get it right.

The paste has hit the bondline as you can see the IHS through it.


Looking at Martijn's post with the stock Intel cooler it looks like to me that intel does the same type of thing. Smaller available contact area higher pressure. I think that locating the contact area inside of the IHS edge is key to capitalizing on performance. The top will flex, the edges will not as much due to the reinforcing bend on the edges.

I bet you could make a better water block by just mimicking Intel's base contact

It will be interesting to see what the pressure film kicks out of all this.

[Das Capitolin]

I will begin this post by stating that I have not and probably will not read every single other post that preceeds it... so forgive me if my information is redundant.

After the myriad of opinions surrounding my 33-Way Thermal Interface Material Comparison article, we went back to work with a larger and more improved follow-up with over 50 TIM's tested (or will be tested eventually).

But to the point... we have tested all of fifteen products thus far. It has taken almost a month, thanks largely in part to Arctic Silvers unreasonable cure time of 200+ hours and variations in relative humidity (we test in 10-15%). Here are my test results averages thus far of some popular products:

Temp Difference / Product Name / Cure Time
46.65 Arctic Silver 5 200+ Hours
48.06 OCZ Freeze 1 Hour
48.18 Noctua NT-H1 1 Hour
48.45 IC 7 Carat Diamond 4 Hours
48.73 Tuniq TX-2 1 Hour
48.80 AC MX-2 1 Hour

All of the products listed above have been tested after their respective cure time. If no cure time was suggested, the product received one hour of thermal cycling via timed restart command batch files and F@H clients.

There were not less than eight temperature samples recorded during the test period, which often times would take days to match environmental variables.

For the purpose of confirming results each material was cleaned from the surfaces and purified after confirming that speard coverage was consistant and complete. A second application was applied and then tested, using the same method as above. The results were collected by subtracting the ambient temperature from the processors stated temperature, and then averaged with the other results.

On an additional note, the IC7 TIM was tested a third time for the simple purpose of making sure our results were over-validated. There seems to be a lot of people out there making some very bold claims, and it's difficult to argue with controlled test results.

[Das Capitolin]

I forgot to mention that these new tests are performed using using a Intel P4 Northwood 3.0 GHz CPU overclocked to 3.61 GHz (1.648V). To ensure uniform distribution of thermal interface material, we are using Intel socket 478 stock reference coolers which utilize lever arms to compress the unit onto the processor. So far, this is proving to be a very successful configuration. Additionally, we are going to make our entire test results available (hundreds of entries) and screenshots in a package download.

[tastymannatees]

For the pressure film testers- I sent out the stuff today. Got everybody from all forums except 2 as I was a little overzealous on my own testing I ran short and had to order more, so some are getting 5 and some are getting 4 test strips.

2 or 1 can be sent back for digital analysis and the balance you can use on your own account, troubleshooting, comparison etc.

I will email you guys with generalized instructions in the next day or two.

I included a couple extra tubes to compensate for the remounts.

I know it sounds stupid but this was the most complicated mailing I have ever done. You each get six envelopes. 2 donor/print, 2 sample , 1 return, 1 to hold it all...