Maximum PC

This is the official thread for results for those who received free IC Diamond thermal compound from this thread and agreed to post their results!

(First, a few words from Andrew Lemont of Innovation Cooling)

Hello and thanks to all who have agreed to be a part of this test.

This will be the place to post your benches, test, and results from your IC Diamond 7 Carat thermal compound you received.

Introduction

At Innovation Cooling we have done extensive tests on our own compound as well as all our competitors and I can say from our results that I have never tested a compound that performs better although a couple come close initially on thermal performance due to a more liquid, lower particle bulk loading. This enables a better wetting of the surface to fill imperfections in the adjoining surfaces. They look great after the first couple hundred hours but what happens after 1000 hours? or 2000 hours? It is well documented in numerous published papers that low viscosity TIMs are prone to bake out and pump out. Intel and others in the OEM market recommend particle bulk loadings above 90% to minimize these effects. While this increases reliability in the long run as it is harder to pump a solid than a liquid and performance is measurably impacted as many here will probably have experienced when using thermal pads, gels, phase change, tapes and other OEM type products.

With IC Diamond we believe we have an ideal combination of thermal performance and reliable performance. Our data on a one inch synthetic die with thermocouples on die and sink show only a 1.1C delta @ 100 W. In other words a mechanically thermal perfect contact between the two would only net an improvement of 1.1C @ 100W.

Purified synthetic diamond has a thermal conductivity of 2,000-2,500 W/mK compared to 406-429 W/mK for pure silver.

Go to the company's home page for info on thermal performance/contact and pressure testing/troubleshooting and any other info you would like.

Application

Go to this link for correct application methods.

Every manufacturer has their own prescribed method of grease application and for this test I would use our method out of the gate and no other. The application method has been tested extensively by us and independently by over 127 forum users. Lines, grains of rice, razor film spreads, BB sized, ultrasonically spread with a vibrator, mixed with other pastes have all been tried. Our recommended application is what works. Application is simple and easy.

You may find heating it up a little bit makes it a lot easier to get out of the syringe and helps with the spreading, so mount the heatsink ASAP after application to CPU.

If you have an HDT ("Heatpipe Direct Touch") cooler you should try using the above method first and the method used for HDT coolers later if you wish to see which works best for you.

Make sure the Heatsink and CPUs are well cleaned using Isopropyl Alcohol 90%+ pure, Arcti Clean, or something similar.

Data Required

I appreciate the effort here from all participating members. Any comments pro or con are welcome as well as any suggestions. There is no bad data in this test, it is what it is, just report the result.

The data we would like is:

"Before" test data prior to dismantling your system. Make sure all heatsinks and CPUs are clean, free from dust, securely mounted, etc. We would like good data.

• Ambient temps to be taken at idle and load
• Idle and load temps obviously. Use a program like LINPACK, Prime95 or OCCT and set to max stress the CPU and allow to run for 10-20 minutes or until temps stabilize.
• Use Core Temp or Real Temp as these seem to give the most accurate readings. Real Temp also shows the max temp each core reaches during the load test. If you use another program just note down which one
• Note initial comparison paste and any exceptional conditions as well as heat sink type
• PWM disabled so fan speed is constant for both tests
• Note your system specs and settings
• Remount with IC Diamond 7 Carat Thermal Compound and repeat as above
• Submit results for any TIMs you use as a comparison also so we can compare results

Feel free to make any observations, comments or suggestions. We are trying to refine our message, application and problem troubleshooting so the information you supply to us is valuable to us and much appreciated.

That's basically it. You are pretty much on your own, whatever result you get is what it is. I may comment occasionally or suggest troubleshooting points but not much else. At the end of the exercise we will provide a chart of the results with links to the data for viewing and suggested corrections.

Thanks again all for your time and effort!

Windows Home SP3 32
Asus P5Q Pro
OCZ2P8004GK 2x2GB Dual channel 5-4-4-12
Intel C2D E6550 at 2.33 GHz
Corsair TX750w PSU
EVGA GTX260/216 Core:576 Shader:1512 Memory:999 Fan%:80
Antec Nine Hundred



All Tempuratures in Celsius unless otherwise noted
OEM Stock HSF Unit

Prime95 v25.8 and RealTemp v3.00 Used Screenshot Test 1
Small FFT's setting, 8K through 10K Prime95 tests completed over 30 minutes, temps stabalized.

Ambient/TIM/CPU-Idle
16.55/AS5/24


Ambient/TIM/CPU-load
16.55/AS5/56

Introducing myself here - Andrew from innovation cooling


I thought I would post some raw pressure/contact film image results from another give away along with the thermal result to kick things off. I occasionally provide test kits to a few members to test the integrity of their mounts as a reminder illustration of how important it is to have a secure (tight mount) and good contact to optimize performance.

In most cases 80% see a result right out of the gate. In sampling of 4 different forums using the contact film it was found that the 20% who did not approx. 80% of those (Alfrado Defrado's rule?) were found to either have insufficient contact and/or not enough pressure for good thermal transfer.

While the test user had a positive result of -2C he was able to improve his performance by adding more pressure with a "washer mod" to drop temps 5C.



Thanks tkl.hui for the extra effort, much appreciated

Note: The "washer mod" in this case were washers that were used to shim the sink to increase the pressure/contact between the sink and IHS.

Might as well add the full digital analysis this is the 8A result

Not a bad result pressure wise, approx. 1 C under the average, overall force in the 50lb range, average 43lbs. 50lbs is generally/approx. the beginning of the optimal pressure range

On the contact side of things some pressure bias on the left and right sides vs lighter or no contact on the top/bottom axis. A common print with the high spot in the middle making for a kind of "island in the lake" look.

This is the 8B Result. The "After Washer Mod"

Washer Mod improved performance another -3 C for a total of -5C - Increased force 15lbs to 65 lbs and the average force 8lbs to 51lbs.

IHS contact is broad,even with few high spots, pretty much ideal from a performance standpoint

Great contact and good pressure says it all.

Staking my post position here, and will edit as I go through all my own tests. I'm slow, but I'm at least consistently slow. I'm a bit OCD, so I wanna make sure what I see is real or memorex.

My two victims to start are a Xeon 2.66Ghz X3350 and Mark Smith's E2160 (I held it hostage for this Mark!). After all this, I'll use the best application and the best tube on the i7 920. Will also so see what water blocks think of this stuff too!


And I'm going through several compounds since this started, but will only post ICD7 first. After this round, I'll get a few more into the mix. Probably over this coming weekend.

-----------------------------
Long story short, I'm skipping the posting results for the above CPU's for now while I work out an article. I saw nothing between AS5, Zalman, Cooler Master, Shin-Etso, and Skythe. I really tried to find a difference and just couldn't get anything conclusive. I have 8 attempts for each on the Allendale and 13 attempts each on the Xeon quad. If I got picky and gave an average delta, none were out of .2C deltas. I'm pretty shocked actually.

If I compare what I got two years ago with pretty much the same pastes, I saw 2 or 3 degrees. So, it may be a testament of the coolers we have available today where the heat backing up actually dinged the interface material more.

But it's clear that the expected brands start to show differences. Tuniq TX-2, ICD7, and OCZ Freeze are clearly at the top of the pile when it comes to cooling performance. The rest of the opinion comes on price, life, application, and the amount you get for your buck...and that's what I gotta reel in on a bit.
--------------------------------------------
On the i7, I ran all night at 63F controlled environment, and 31.0-33.4% humidity with the shin-etsu; which is the OEM compound on a V-8. Eight workers with 8-8 FFT's.


Ripping apart for the ICD7 run tonight...

I'm going to replace the TIM on my GPU as well as the CPU. What would be the best way to stress the GPU? would GPU folding be enough, or would it be better to load up CoD4 or another intense game?

Edit: no point in making a new post here for my results, so I'll just post them here.

Asus N50:

Stock Thermal paste:
Core 2 Duo T5800:
Idle: 41 C Load: 62 C

Nvidia 9650m GT:
Idle: 51C Load: 66 C


I think I'm going to replace it with ICD7 this weekend.

GPU folding will run it harder than the games will. Plus, you'll be making points at the same time. GO FOLD!!

Putting claim to my post results here. I'm currently modding a case to get my new Folding machine off my test station, then will start in on the ICD7 testing.

Here's the results for my Q6600. Since learning from Chumly and Innovative Cooling that a single dab in the middle of the CPU is proven to be the best method, this is the method I used for all applications of TIM. I plan on testing different methods, the line down the middle, an X, and spreading it in one layer across the whole IHS, but have to wait until the tube of ICD7 I ordered gets here and I have more time. I'm also going to test my E8400 which has a history of having 10°C temp differences between the two cores.

Q6600 – Using CoreTemp to track temperatures. Prime95 v25.9 for loading the cores. CPU @ 2.8GHz and 1.3V.
CNPS9700 @ full speed for all tests
EVGA 780i SLI mobo
Antec 900 with all case fans on Medium

Arctic Silver 5. 74.5°F Ambient temps. Idle temp of 40°C
Lemme find the damned image for this one. It was averaged at 64.5°C


Arctic Cooling MX-2. 74.5°F Ambient temps. Idle temp of 40°C


ICD7 with 72.5°F Ambient temps. Idle temp of 39°C



Once we take into account the Ambient temps, the results are as follows. All I did was add two degrees to the ICD7 result to account for the difference in ambient temps.
ICD7 – 60.5°C
MX-2 – 62°C
AS5 – 64.5°C

Setting up shop here posted current system specs. Will Edit as needed


Gigabyte p35-ds3l
Intel C2D E7300 at 3.4ghz 400x8.5 1.35vcore (2.66ghz is stock)
thermaltake v1 fan @ 100%
4gb patriot ddr2 pc6400 2x2gb 4-4-4-12
antec truepower trio 650watt PSU
Evga 9800gtx stock clock
cosmos 1000


control test with thermaltake v1 with tuniq x2 TIM.

OCCT v3.0.0 cpu:linpack test for 15minutes

realtemp V3.00

Ambient is 67 degrees F

Idle temp is 38c

load temp is 69c



ICD7 Experimental text

OCCT cpu:linpack test idle for 1min full load for 15min

thermaltake v1 with ICD7

ambient 67c

idle temp was 36

full load temp was 67


Edit posted results with ICD7 TIM

On my new setup:

E8400
EVGA 750i FTW mobo
EVGA gtx260 55nm
Zalman CNPS9500 cooler

with arctic silver ceramique I was running nearly 50c at idle.

I lapped the cooler and the proc and here are pics of the process

before:




This is after the first run with the 600 grit. you can see some copper showing through. See just how uneven the nickel surface really in on a CPU heat spreader.




Here is after the 800 grit run. The copper is coming through more. You can tell the lowesrt part of the surface (the surface making the least contact with the heatsink) is the center. the area where the cores are! so this CPU would be a poor thermal performer. In fact, in a short prime95 run, it heated up to over 60c very quickly before the lapping.




Here is after the 1000 grit. You can still see some nickel in the center, this chip's heat spreader was terrible! and you wouldn't know how bad unless you did this.




Here it is pretty much finished. going to do a few more runs with the 2000 grit for sport. Then do the heatsink on the CNPS9500 cooler.




And the final results before reinstall. Not bad eh?



Now with the Diamond, idle is 37c and load with OCCT is ~60c

Finally, got to do the testing on one machine. Room temperature spanned from 60.5-71.1 F during the testing.

This test between a Q9550 OC'd to 3.0 and a TT DuOrb cooler.
In HAF932 Case w/6 intake fans and 6 exhaust (counting vid cards and PSU that all vent out) fans.

Load tests were done using Prime 95 set to run only large FFT's (for consistancy) for no less than 30 minutes prior to recording temps. Temps were then monitored for 5minutes and maximum values posted below.

Idle test was done by killing the Prime 95 and letting the machine cool for no less than 30 minutes (even though it was cooled down in about one minute, WOW) and then monitoring for 5 again.

First I started with the AS5 as it was already in there and had it's chance to set-in/cure/whatever you want to call it.
Loaded up:
Core 0=54
Core 1=55
Core 2=55
Core 3=53
Ambient case temp=51

Then at idle:
Core 0=38
Core 1=37
Core 2=38
Core 3=43
Case temp=34

Then came the ICD7
Loaded up:
Core 0=51
Core 1=53
Core 2=53
Core 3=52
Case temp=49

Idle:
Core 0=36
Core 1=38
Core 2=38
Core 3=45
Case temp=35

All temps are the max value read during a 5 minute run of the temp monitoring utility. One other thing I found of real note with the ICD7 was that while keeping the max temp lower, it also maintained a more consistant temp, fluctuating only 2 degrees during the 5min run, while the AS 5 spanned as much as 6 degrees on some cores.

Folding does not always give consistent load. I like Dreaded's thread:

will edit as I go, doing lots runs at different ambients to see if the delta is constant. IC7 is on. results on runs ongoing.

once Ive done the data runs on this system, Ill swap the temp sensors to my other system (stock Intel cooler with the center copper slug where the only contact point between CPU/HSF is the center of the CPU IHS) and do runs with generic thermal grease (radio shack stuff from years ago), AS5 and IC7. Ill do that system up in another post.


Intel Pentium 3.0 Ghz 4C northwood @ 3.2 Ghz
Antec socket 478 cooler (no model #, just lists as "performance CPU cooler)
Intel D865PERL mobo
Antec 3700 AMB case (2x120mm fans, controlled by speedfan, locked at 100% during testing)
Antec Trio 550 W PSU
ThermalTake 4 channel fan controller for slot fan, internal 80mm fan, ATI video card fan and CPU fan
ATI 9600
3 gigs RAM
2x HD, 2x optical
XP SP3
slot exhaust fan (on low for tests)
80mm thermaltake fan directed at RAM (on low for test)
freshly blown out heatsinks, fans and front filter.

used speedfan 4.37 as realtemp doesnt like my oldie P4.

I placed several thermal probes to get ambient temps, one for case intake, one at CPU HS fan intake, one at CPU HS fan exhaust, as shown below.



baseline results with AS5 applied about 4 months ago, case fans at 2000 RPM, CPU 4300 RPM, 80mm internal 700 RPM, all locked at speed.

PSU fan varies under load so I put that in. no way to control it but it shouldnt effect the results as it will act the same with the IC7.

all runs were in pairs, IE run #1 was idle 1st, then load. at least 15 mins cool down time was allowed between runs for idle temps to stabilize.

AS5 - PRIME 95 RUNS:




AS5 - OCCT RUNS:



IC7 runs:

CPU was cleaned with 90% iso alcohol and coffee filters. since I had only done a light lap with 800 grit (previous to these tests, nothing was done extra during these tests) I applied a haze of IC7 to the CPU and HS as per the website instructions. I had done the the same haze bit with the AS5. paste was applied, rubbed in, excess wiped off with coffee filter leaving a very light glaze on each surface. a dollop was applied in the middle of the CPU and the HSF squished down, screws tightened to the stops.



since this application of IC7 is so fresh (less than an hour since applying) Ill separate this run from others in a day or so after its settled a bit.

closest ambient temp match is OCCT run # 1, so I added or subtracted the appropriate ambient temp for this result.

going by case intake temps:
idle: -0.5
load: -1.5

going by CPU intake temps:
idle: +0.5
load: -0.7

I wonder if the discrepancy is just margin of error or if the CPU intake probe shifted slightly and is getting slightly different airflow. or maybe because the system was powered off for an hour or so and the chassis/case had not warmed up to normal 24/7 temps - theres a lot of steel in this case . further runs will probably tell.

IC7 runs after 1 day (or more) after application:



matching to closest ambient AS5 OCCT run and adding/subtracting appropriate ambient temp:

OCCT run 1:

going by case intake temps:
idle: -0.5
load: -0.8

going by CPU intake temps:
idle: -0.4
load: -2.3

Prime 95 runs (separating from OCCT runs and comparing to previous Prime95 runs - even though they seem to load the same temp wise)



matching to closest ambient AS5 Prime95 run and adding/subtracting appropriate ambient temp:

Prime95 run 1 - match to AS5 P95 run 1 and adding/subtracting appropriate ambient temp::

going by case intake temps:
idle: -0.8
load: -0.8

going by CPU intake temps:
idle: -0.4
load: -1.1

Prime95 run 2: - match to AS5 P95 run 4 and adding/subtracting appropriate ambient temp::

going by case intake temps:
idle: -0.4
load: -0.6

going by CPU intake temps:
idle: -1.0
load: -1.1

Prime95 run 3 - match to AS5 P95 run 2 and adding/subtracting appropriate ambient temp::

going by case intake temps:
idle: -0.7
load: -1.0

going by CPU intake temps:
idle: -1.3
load: -1.8

I like this so far

when Im done testing Ill plug all this stuff into a spreadsheet (partly for fun, partly for completeness) and see what kind of variances there are on the data for the various runs. this should give me a good feel for any margin of error induced by the probes or myself and allow a more or less final number in temp differences for the compounds.

casual observations - this is my day to day computer. I fold 24/7 and for the last couple months have run nothing but the gromacs core. normal fan speeds are 650 RPM for case fans, about 1500 RPM for the 80mm and 3000 rpm for the CPU fan for fairly quiet noise levels. so I know more or less what the normal day to day load temps are. my day to day folding temps are about 1-2 C less.. no measurements to back that up, just a casual "knowing your system" type of thing.

Ok, posting the results from my test.
Sanded down the IHS and heatsink for both tests. Obtained some generic stuff from Newegg. Applied it, and here is for the generic/stock paste:

Stock/Generic Idle
Room Temperature: 20C
Ambient Temperature: 36C
CPU Temperature: 39C

Stock/Generic Load
Room Temperature: 20C
Ambient Temperature: 42C
CPU Temperature: 50C

And for the IC Diamond:

IC Diamond Idle
Room Temperature: 20C
Ambient Temperature: 27C
CPU Temperature: 28C

IC Diamond Load
Room Temperature : 20C
Ambient Temperature: 33C
CPU Temperature: 40C

System Specs
Windows XP SP3
Intel Pentium 4 Northwood (socket 478)
Intel D845GEBV2
Roswell PP550-2 550W
Sapphire ATI Radeon X1950PRO AGP 512Mb
Generic Case

Software Used
SpeedFan, BIOS, and CoreTemp (Didn't work for the IC Diamond tests for some reason, but the temps read the same for the stock/generic stuff as SpeedFan, so I presumed they would be the same for the IC Diamond tests.)
Prime95 for the stress tests, and some folding mixed in.

Bottom Line
The IC Diamond stuff really did some wonders. Brought load temps down 10C and idle temps down 11C, . That's the best stuff that I've tested so far. Fantastic.

what was the source for ambient temp readings? the mobo northbridge? thats down 9 degrees at both idle and load between the generic and IC7 runs .. did you apply IC7 to it (northbridge?) as well?

I received my tube yesterday. I'll be posting my results here in the coming days / weeks.

Test setup:
Athlon XP-M 1800+ 1.53 GHz. oc'ed to 2.4GHz (200Mhz FSB * 12)
2 x 512MB DDR @ 200Mhz
Thermaltake Volcano 7

I have 1 120mm fan pulling in and 2 80mm fans pushing air out of the case.

I'll normalise the temps in my apt. and take temp readings at idle and load over 2 days to ensure a fair reading for the baseline. Then I'll rip the box apart, apply the new TIC, and measure the temps at the same time every day for a couple of weeks. I'm interested to see if there is a temp delta as the TIC cures.

there seems to be in my case. temp results from the 1st hour or two after application are off from others a day later but that was stated by IC7 somewhere and was expected.

but after the 1st day my tests seem to indicate that over the next few days it improves a bit. hard to say for sure as I only have a small data set at this time.

I wonder if thats just my cheapo HS doesnt have enough mounting pressure to squish it flat right away and its working itself flat over the course of several days. be interesting to see your time vs delta results.

Here is the first figures of my triple threat match for the CPU. But first here is the ring where these 3 contender will do battle.

Thermaltake Eureka (fans replaced with Noctua NF-P12-1300)
Gigabyte GA-MA69G-S3H (all cooling aids off)
AMD Athlon X2 BE-2350 (Cool-N-Quiet disabled)
Sunbeamtech Core-Contact cooler (locked at 100% fan speed)
G.Skill 8GB(4x2GB) DDR-800
EVGA Geforce 8800GT (locked at 100% fan speed)
Enermax Liberty 600 power supply
Asus DX audio card (PCI-E)
2x WD Caviar 250GB WD2500KS harddrives running independently
1 Samsung SATA DVD burner

-The test will be conducted using Prime95 as the stress-testing app (five 30 minute burn-in cycles first and then a 1 hour stress) and Core Temp 0.99.4 for temperature readings.
-The room the pc is in is registering 26C at the time of testing (all outside cooling is off, no fan or AC)

First up is the Artic Silver Ceramique:
Core 1- idle - 30C load - 37C Core 2 - idle - 26C load - 37C

Next up is the Tuniq TX-2:
Core 1- idle - 31C load - 40C Core2 - idle - 27C load - 40C

Now comes the moment you've been waiting for, HERE IS the IC Diamond:
Core 1- idle - 34C load - 41C Core2 - idle - 31C load - 41C

Looks like the Ceramique is still on top. Things are not looking good for the IC Diamond 7 . It is in last place with a temperature rise of 4C in idle and load compared to the Artic Silver. Maybe it requires a longer break period than the other two. If so I will host the new result after 2 more days for each treatment and see if things change. For now I have to give the belt to the winner:

Artic Silver Ceramique.

Well the GPU event will have to be canceled. The reason being that as I opened the card up I noticed the culprit of the high temp readings. The thermal tape is coming apart (too much stress I guess). I can replace most, but not all of the points with thermal grease but some vital parts will b left exposed so I will not risk it. Instead I decided to put pretty much everything but the caps under the IC 7 Diamond right away. I did a stress test and everything is working like a charm. Here are the results
- Using Furmark stability test (xtreme burn-in enabled for five 30 minute burn-in cycles and 1 hour stress test) and EVGA Precision for temperature readings.
- disabling fan speed automatic control and keeping it locked at 100%

stock configuration: 60C idle, 90C load

IC 7 Diamond: 49C idle, 81C load

difference: 11C idle, 9C load
Talk about a massive difference, but it may be all due to the poor state of the original thermal paste. Oh well, the CPU test should give us more info on that regard.