Innovation Cooling's Diamond 7 TIM test results

[123bob]

Farm-09 AS5 vs D7 - re-test after HW shuffle from previous test. Like SiG's machine, this one seems to be an outlier on the test. Looking at the spread patterns you can see that the CPU looks convex. It begs to be lapped, which I will do only when Tasty has had his way with this setup.

Farm-09 is my bench rig, sitting open on the bench. It has a Q6600 on an Abit IP35 Pro. It is running at stock for the last couple of weeks for this test. (It will get clocked back up as soon as the last AS5 cures.... I can't STAND stock speed anymore.... )

It has a TRUE with a Scythe fan and a 120 mm 110 CFM fan blowing from left to right WRT to the front of the mobo.

Short story results up front, detailed data below.....I tried four different runs.

1. The first was AS5, pea blob, which was allowed to cure.
2. The second was D7, pea blob, Also allowed time to cure, but never did in the time allowed.
3. The third test was the cross pattern, run for 5.5 days. Again, never changed with the exception of a modification my lab assistant and her partner made during an intense game of hide-and-seek. Noted below.... This may perhaps turn out to be a "eureka" event.....
4. The fourth test was back to AS5, pea blob, to re-validate my results, which so far look consistent with run #1. The last piece of data I will add to this post is the cure temps of AS5 on test #4.
EDIT 4/15: I added cure Data to the below with spread pics. Any further edited text is in italics through this post. I also added the start of test #5 Shin Etsu X23-7783....

5. Shin Etsu X23-7783, pea blob

Bottom line - The cured AS5 from test #1 outperformed the D7, at load, in this particular case, over the time tested. The second AS5 test #4 was on par with D7. The I don't know why, other than the D7 is so much thicker, and perhaps with the convex CPU, that made the difference. A theory is that due to the convex shape, and the big difference in consistency, the AS5 was better able to form a thinner interface on the smaller, critical contact point. OK, now to the details.....

Side note - I cleaned the be-jeebers out of the TRUE and CPU on changing pastes. I used acetone and cleaned the AS5 and D7 so hard, it's near impossible to read the chip label now. As noted before, D7 also makes good polish!!


Test #1 was with AS5, healthy pea blob. Pic below.



I'll save results for below. Here's the spread from the AS5 on breakdown. this is after cure time. It took about 24 hours for a significant temp drop from curing to show up with AS5. I think cure time was accelerated using the 50% profile with EIST enabled. I'll also note that AS5 is much thinner than D7. Shin Etsu is closer to D7 in consistency.



Note the fine spot in the middle. It appears to be where the AS5 had the closest contact after the cure. This will get the pressure film test when it gets here.


Test #2 - Here's the intial pea blob of D7



Here's the spread pics






Test #3 - For this next test I used the cross pattern. I found I got good coverage with this in prior passes and figured I'd give it a whirl on this pass.



I also opted to put a little more weight on the matter. The juice jug provides a few more pounds. (I don't think it matters what flavor, but I may be mistaken... )


Here's the breakdown spread of the cross mount after 5.5 days.





Test #4 - This is initial pea blob for the AS5 re-test


EDIT: Here's the spread pics of test #4 AS5





EDIT: Here's the beginning of Test #5, Shin Etsu X23-7783


OK, the results……..


Test #1 AS5, Pea blob
AS5 – Temps

Idle: 15min time running
Ambient 23.9 C
CoreTemp – 33,34,31,31

Load:
BOINC as load – 15min time running
Ambient 24.2 C
CoreTemp – 47,49,44,45

Load: CURED
BOINC as load – Approx 22 hours time running
Ambient 23.2 C
CoreTemp – 39,42,37,38 (Restarted machine to verify numbers. They were the same about 10 minutes into restart. Verified results…)



Test #2 D7, Pea blob
D7 - Temps

Idle: 10min time running
Ambient 23.9 C
CoreTemp – 30,31,27,27

Load:
BOINC as load – 15min time running
Ambient 23.9 C
CoreTemp – 44,47,42,43

Load: at 21 hours
BOINC as load – Approx 21 hours time running
Ambient 24.7 C
CoreTemp – 44,47,43,43
(Pulled early since I thought my mount was bad. Turns out it was good….)



Test #3 D7, cross pattern
D7 - Temps

Idle: 10min time running
Ambient 24.8 C
CoreTemp – 32,33,30,30

Load:
BOINC as load – 60min time running
Ambient 25.4 C
CoreTemp – 45,48,43,44

Load: at 3 days (Before Bottle knock off…)
BOINC as load – Approx 3 days time running
Ambient 21.6 C
CoreTemp – 42,44,39,40

NOTE – What occurred between these two data points is that “someone” bumped the bench and knocked the bottle off the cooler. They didn’t put it back on, otherwise I would not have known it happened…..Note that the temps got about 2 degrees worse. This is like I noted in a prior post about “getting gorilla” on the mount and blowing the temps. On a convex mount like this, is it perhaps that in rocking the mount, it removes the paste from the outer edges, thereby reducing overall TIM contact? Seems logical….

Load: at 5.5 days (After Bottle knock off…)
BOINC as load – Approx 5.5 days time running
Ambient 24.7 C
CoreTemp – 46,49,44,45




Test #4 AS5, Pea blob
AS5 - Temps

Idle: 15min time running
Ambient 24.8 C
CoreTemp – 31,32,28,28

Load:
BOINC as load – 2 hours time running
Ambient 23.5 C
CoreTemp – 44,47,43,44

EDIT:
Load: at 1.5 days cured (Before pushing TRUE)
BOINC as load – Approx 1.5 days time running
Ambient 20.4 C
CoreTemp – 39,43,38,39

NOTE – What occurred between these two data points is that I pushed on the top of the TRUE with only about 10 pounds of pressure while trying not to rock it. It was enough to blow the mount such that the temps never recovered....Delicate stuff this AS5.....I cranked much harder on the D7 in order to blow it's mount...

Load: at 1.6 days (blown Mount)
BOINC as load – Approx 1.6 days time running
Ambient 21.1 C
CoreTemp – 42,44,39,40



Test #5 Shin Etsu X23-7783, Pea blob
Shin - Temps

Idle: 15min time running
Ambient 23.7 C
CoreTemp – 27,30,27,27

Load:
BOINC as load – 15 min time running
Ambient 23.7 C
CoreTemp – 43,46,42,43




Figuring for ambient changes, the first cured AS5, at load, was cooler than either D7 test. I must have had a great mount on the AS5. I’ll add the cure data for the second AS5 test when it comes in then I’ll update this post. D7 does very well on the idle tests, but not against the cured and loaded AS5 run.

Based on all the data others have provided, as well as my own, I'm starting to form an opinion. (Note that this is an opinion only, not a scientific conclusion fully supported by data....) I think D7 is a very good long term TIM. I've come to trust it over time, much like I trust Shin Etsu. I must say I was surprised to see it beat my last Farm 07 Shin Etsu post. I will be re-visiting that when the new tubes come from Qouc at Petra's. I'd like to know about this outlier. I don't think the AS5 mount would last over 6 months, but who knows. I may take it that long on the final test pass....

I still find it interesting that rocking the mount on a convex case causes that much trouble. I guess it makes sense when you think about it. I would guess any paste would have this problem. When I get ready to break down this AS5 mount, I'll do the same thing and see what happens.

EDIT: The second AS5 mount confirmed a couple of things. The mount I had on the first test was great....There is variance in the AS5 mounts though. The D7 mounts are more consistent. The other thing of note was how easy it was to blow the AS5 mount as compared to the D7 mount. It really took much less pressure to do it. I intentionally rocked the D7 mounts in order to blow them...

The Shin Etsu (test #5) is looking good so far. This paste has a distinct cure time and almost triples it's thermal conductivity in that time. That's according to the Shin data sheets. We'll see real world results with this mount, one I consider a problem child.... If I can blow this mount as easy as the AS5, I must say I'm nearly done testing for myself, I'll be a D7 fan. Any further testing will be for Tasty (Andrew).

SOAPBOX: Like my rather conservative crunching clocks, I want something I don't have to dork with, worry about, or not have confidence that it will crunch without me for months....The Tortoise and the Hare, a lesson well learned over the years.....:END SOAPBOX


This CPU and cooler are definite candidates for a lap job, and will get it as soon as I’m done testing this configuration. EDIT: Look at the mill marks on that TRUE!! Disgusting...

Tasty, this will definitely get the film test. I’ll hold this configuration as long as you want me to try stuff on it. As stated above, the clock will go back up, but I’ll baseline that. If needed for a particular test, I can downclock again if we are trying to match an earlier data point. This is my bench machine, so it’s easy to dork with it….
One issue I was thinking about is how we execute the film test. I could see in mounting we will have to be careful not to give false reads on the film by cranking the cooler. On the other hand, this may provide evidence of different coolers pulling stuff astray by the mount conditions. The TRUE maybe particularly so…..We should discuss this a bit among those who are receiving film…..This is new ground for me, you paid for the film, so you tell me your thoughts, and maybe the folks here will be able to refine this some.
Send SiG some film, whether he asked for it or not. Now I have to know why we are "outliers"…..I’m sure the good Doctor will indulge one of his best patients….

I'd like to say thanks to Tasty (Andrew) for providing this test opportunity to us. I've learned a LOT more about how I approach TIM application and treatment.

Now, when can I have some Nehalems to test this knowledge out on????

Regards,
Bob