Innovation Cooling 7 Carat Diamond TIM Test results thread *NOW ACCEPTING RESULTS!!*

[ICD7]

Originally Posted by CTA

unbelievely -21.5 from "Infiniti/MX-2/-21.5 " i mean huge number will ruin average... (yeah i read his post)

if -21.5 is not including then average would be 3.47...

honestly according this result... i think your IC7 is best TIM...

Usually I try not to second guess what people get, as we go into a higher sample it becomes less and less of a factor. One I had over at OCA was about 14C and ended up changing the overall by .03 of a degree C. So when I factor that into 300 tests it becomes 1/300th of the calculation.

Oddly enough(or not) the final averaged number is almost spot on with what I get in my lab tests within a couple of 10ths of a degree. the errors tend to cancel each other out. All the variables good mount-bad mount, lots of pressure -not enough pressure good ambient reading vs bad ones, hardware problems, different systems, software, literally hundred of variable combination's, if the sample is large enough it seems to work great. All the data is good from my perspective and if you eliminate the 10% of the extremes on either side you end up pretty much back where you started.

A few more tests and I expect the averages will come down a bit, right now the group is off the the 232 by only 8/10 of a degree and by the end of the project the gap should close considerably.

[ICD7]

Originally Posted by Infiniti

Well, yeah I agree that my results were a little strange, but I think the math was done wrong. I'm pretty sure that only the load was taken into the average, so from my calculations, (14+14+13+15)/4 = 14, not 21.

Noted and will edit tomorrow high improvements are not unusual - could be mount another which I favor is that the previous TIM has deteriorated and could have failed which is common.

[CTA]

thats true... only if numbers of sample is large... but i refer to that result of hwc...

[ICD7]

Originally Posted by CTA

thats true... only if numbers of sample is large... but i refer to that result of hwc...

I will post it both ways if you want, not a problem but the sample is small at this point with one zero and only one +1.1 to the negative side. MX-2 only? or the whole group? Kind of makes it awkward to choose the starting point. If we get the numbers up to 50 - 60 user tests it will be easier to take 10% a side.

[Soultribunal]

PRE-IC24 / AMD OPTERON 2.0 / ZALMAN CNPS9000 / OCZ FREEZE

AMBIENT: 23 DEGREES
IDLE TEMPS: 31/33 DEGREES (FOR BOTH CORES)
LOAD TEMPS: 57/59 DEGREES

POST-IC24

AMBIENT: 23 DEGREES
IDLE TEMPS: 27/29 DEGREES
LOAD TEMPS: 55/58 - 54/56 (I NOTICED WITH IC24 THAT AFTER AN HOUR OF LOAD I GOT A LITTLE MORE OF AN IMPROVEMENT)


PRE-IC24 / INTEL CORE 2 DUO E4700 / MASSCOOL 8WA741 / OCZ FREEZE

AMBIENT: 23 DEGREES
IDLE TEMPS: 25/26
LOAD TEMPS: 60/61


POST-IC24

AMBIENT: 23 DEGREES
IDLE TEMPS: 25/27
LOAD TEMPS: 57/58 - 56/57


AMBIENT AIR TEMPERATURE MEASURED WITH A FLUKE THERMOSTAT AT THE SAME LEVEL AS THE COMPUTER AND APPROX 2 FEET AWAY.
THIS UNIT STATES A VARIANCE OF 1 DEGREE, AND HAD BEEN CALIBRATED ONLY 2 WEEKS AGO.
CPU TEMPS WERE TAKEN WITH CORETEMP FOR THE AMD, REALTEMP AND CORETEMP WERE USED FOR THE INTEL.
LOAD WAS APPLIED BY USING PRIME95 SMALL FOR AN HOUR.

My own thoughts:

Intels setup for how they mount heatsinks helps a lot more in applying this compound evenly. For the AMD I found that making a pattern (I made 5 very small dots like the '5' side of a die) then pressing it firmly down with a little 'wiggle' gave me the best results.
It is a unique TIM to have the pleasure of working with , not what I am used to for sure.
I don't know if its perhaps because I got the IC24, but I personally found that cycling it thermally for a few hours dropped my temps a little bit more. I don't know if its because this compound has to cure, but it was noted and noticed in my CoreTemp Logs (I say hour, it was more like 1:07 but who's counting).

Thank you for the chance to give this stuff a test.

Regards,
ST

[ICD7]

Originally Posted by CTA

thats true... only if numbers of sample is large... but i refer to that result of hwc...

Infiniti - If I plug in his corrected number of -14C I get -3.92 for MX-2 as an average
overall -3.79 C

If I take out the highest and lowest for MX-2 the average is -3.62
overall the average is now 3.59.

Corrected for MX-2 the difference is 3/10ths of a degree C
Corrected overall the difference is 2/10ths of a degree C

As we are about halfway through the exercise my point is that as the sample size increases the difference between corrected and uncorrected will shrink to about 1/10th of a degree and out of 300 it'll be in the low 100ths of a degree. The larger the sample the less the outliers influence the final result, if the sample size was only 10 then the outliers have more influence and the reason for cutting them make much more sense.

+/- a 1/10 degree C is not a lot to quibble about. The ongoing data is preliminary and subject to changes and corrections and will smooth out some as we go. In the final results I will post a correction chart also and all can be compared at that point.

This is user data, not my data and it stands on it's own. I generally resist the temptation for corrections because then that line becomes blurred and I have to explain and defend the corrections - in effect it loses some validity in the user's eyes because the data is being manipulated by us and we now own the results.

[Soultribunal]

ICD7 , I would like to state, that although I was happy with my results, I would not count the AMD setup in that field. The only reason I say this is because the nature of the Mounting setup on the CNPS9000. I got numbers both better and worse for the OCZ and IC24. It is not the ideal system to make these tests for sure, and in the end I took what I felt was the best Mount of both TIM's and posted as opposed to averaging the results over each application.

I am however extremely confident in the LGA775 system setup, as it is a backplate mounting system with spring screw Bracket. Much easier to get consistent runs using that.

Regards,
ST

[Jonwall]

I think that as this is user results it makes sense to leave all results in. Obviously not everyone is going to be able to apply this stuff perfectly, and its happened with the other icd7 giveaways ive seen as well. Basically a consumer can take these results and know that on average, out of all tests, thats what they can expect to get. But if they spend the time getting a perfect application they can achieve more. Personally i havent seen any TIM compound with such in depth instructions and explanations on how to apply it, and why it works best, along with the images to prove it. Thats probably one of the reasons this stuff outperforms pretty much everything else, if someone went to as much work figuring out pressure patterns etc with mx2 im sure the gap would be smaller, but they dont. The fact that ic goes to this much work to help you get the most for your money is what will make me buy icd7 in the future whenever i need tim.

[SKYMTL]

My turn:

i7 920 @ 4Ghz on CoolIT Boreas

Load = 20 minutes Prime

Ambient temps constant @ 20.2C at an average humidity of around 17% (yes, electric baseboards make it DRY)

MX-2:
Idle: 15 / 14 / 18 / 16
Load: 38 / 38 / 41 / 40


IC24:
Idle: 15 / 14 / 17 / 16
Load: 37 / 36 / 40 / 38

[Sagath]

Well, finally got around to slapping some IC7 on things. Still waiting for the watercooling setup...that will be a while for the build.

But so far on my 88gt overclocked I went from full load no fan 70c to 64c, still with no fan. Full load with fan the difference was actually larger (maybe due to the better ability to pull heat away?) at nearly 8c. 'stock' goop vs IC7

Pritty impressive for a GPU!

On my 120 Extreme 6600@3.4 I load out at 54/54/49/51, with IC7 I'm down to 52/52/47/49. So a solid 2c drop across all 4 cores!!! This is over AS5 TIM.

Pritty impressive to say the least!!!