ICD 7 Test Results

[RaySun2Be]

This is the official IC Diamond 7 test results thread.

Please post only test results, or testing questions or issues here.

For questions about getting a sample to test with, or delivery issues, please post in this thread: ICD-7 Test


Application

Every manufacturer has their own prescribed method of grease application for this test I would use our method out of the gate, then when you know what the best possible result that be gotten you have a basis to fine tune or experiment with your own application methods if you feel the need to do so. The magic Cross is the best application method I have seen to date but tends to use more paste.

IC Diamond is composed mostly of diamond powder, and as such is quite thick. Proper application is critical to optimum performance. Squeeze onto the center of the CPU an amount of IC Diamond compound about the size of a pea - the center of the CPU is where most heat is concentrated. Place the heat sink on the CPU and press it down , then twist it back and forth a couple of times to spread IC Diamond over the CPU's surface. Clamp the heatsink and power up the PC.

We recommend a pea sized amount on center to increase the likelihood of optimum results for the average user. That amount with the proper pressure will spread to cover the entire area of contact between the sink and IHS 100% of the time. All paste application amounts are subject to a kind of artistic guesswork i.e a grain of rice is marginal with ICD7, it is highly bulk loaded for long term stability, very different than the wet style lightly loaded greases which are subject to bake out or pump out in the short term.

A squirt and squish with maybe a twist to spread it around is just about as easy as it gets and from our experience effective and foolproof.


Requested Test Information

We would like the before load and idle temps along with the age or time in place of the previous mount

The ICD7 load and idle temperature and if possible some care on the ambient temperatures from both before and after readings.

Feel free to make any observations, comments or suggestions we ar trying to refine our message, application and problem trouble shooting 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 trouble shooting 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!

 

[Markfw]

I got mine today. Starting the test

Here is the Setup:

Motherboard : ASUS P5K
CPU E6300
OC = 7 x 480 = 3360. and 7 x 490 = 3430
VID = 1.325 OC vcore = 1.375
HSF = Ninja Sythe using "white paste" probably from Thermalright HSF
Ambient = 73F
Temps taken using coretemp 0.95.4

Before temps:
@3360
IDLE 55/52 C
LOAD 65-66/60-62 C (varies)
@3430
IDLE 56/53 C
LOAD 67-66/64-63 C (varies, still running)

After temps:
@3430
IDLE 50-51 / 47-48 (ICD7 thermal compound installed with glop and squish method described above. Results ~5 min after installation, will wait before load for it to settle)
LOAD 58-59 / 57-56

Note: This CPU was lapped

And this stuff is amazing, 8c drop in temps ?? I may have to try on my B3 problem children next... They run the hottest !

Also, this is after only 10 minutes, I will update if it gets hotter.

 

[covert24]

Motherboard : Abit IP35 Pro
CPU E8400
OC = 9 x 412 =3707.
VID = 1.0500v OC vcore = 1.19v
HSF = Cooler Master Hyper TX2 w/ AC Ceramique
Ambient = 70F
Temps taken using Everest Ultimate

Before temps: (With Ceramique)
@Stock (3.0GHz)
IDLE Core#1- 42 / Core#2- 39
LOAD Core#1- 56 / Core#2-55

@3707
IDLE Core#1- 46 / Core#2- 44
LOAD Core#1- 64 / Core#2- 63


After temps: (With ICD7)
@Stock (3.0GHz)
IDLE Core#1- 42 / Core#2- 38
LOAD Core#1- 54 / Core#2- 52

@3707
IDLE Core#1- 44 / Core#2- 44
LOAD Core#1- 61 / Core#2- 59



Holy Hell this stuff is awesome! my temps actually dropped! haha. LOAD is really where this stuff shines. Dropped my #1 core down a whopping 3*C and my core #2 down 4*C!.

I used Prime95 v25.6 Build 6 on Small FFT for my LOAD temps. All temps were recorded using my CRAP Hyper TX2 heatsink so im usre if I actually had a decent HS like a TRUE or the like, my temps would be that much more amazing.

I just want to give a thank you to the people over at ICD for conducting this whole thing. Really speaks a lot about the company and its drive to listen to the consumer. Hey, you don't see the people over at Arctic doing stuff like this right? That being said i think i have just found a new TIM for ALL of my future builds.

Thanks again guys and I hope that the data recorded in this thread helps you guys out.

 

[narzy]

Previous Compound Artic Silver (2 maybe?)
Installed Dec. 2002
Stock HSF
Proc: Intel P4 3.2ghz
Idle 35C (Measured after 10 minutes idle)
Load: 51C
Measured using Speedfan 4.33
Application method Line and smear / smooth using credit card

IC7
Idle 34c
Load 52c
Application method pea sized blob centered.

Video card abandoned

 

[Fardringle]

I don't have the "after" numbers yet, but here is my setup and temperature readings with a roughly 4-month old application of Arctic Silver 5 on the CPU and stock paste (which was dried and flaky) on the video card:

Motherboard: Biostar K8NHA Pro
CPU: Socket 754 Sempron 1600+
Overclock: 8 x 275.5 = 2204 MHz
HSF: Thermalright SI-128 with Arctic Silver 5 paste
Video Card: eVGA 6600GT with stock heat sink
Ambient temperature: 72 Fahrenheit (~22 C)
Temps taken using Speedfan and PC Wizard 2008 (I got the same numbers in both and CoreTemp doesn't like my motherboard for some reason)


Before temps:
CPU:
Idle: 35-36 Celsius
Full Load: 41-42 Celsius
Video Card:
Idle: 56-57 Celsius
Full Load: 79-81 Celsius

I waited about 24 hours after applying the stuff so it could settle in and I could have the same ambient temperature as the initial readings.

After temps:
CPU:
Idle: 32 Celsius
Full Load: 39 Celsius
Video Card:
Idle: 55-56 Celsius
Full Load: 65-68 Celsius

I got a small improvement on my CPU, and a rather large one on the video card. I already had AS5 on the CPU so I wasn't expecting miracles, but any improvement at all is a pleasant surprise. It should be just enough that I can keep my overclock through the hot summer instead of having to dial it down a bit to keep the system stable.

My video card should be much happier now than it was with the dried "toothpaste" garbage that was on it before.



I do want to say that the "squirt and squish" method did not work at all on my CPU. When I did that, the paste only ended up covering the center 1/3 of the CPU heat spreader in a thick layer, and my idle CPU temp hit 43 C just in the BIOS so I immediately shut the system down. I took the heat sink off and spread the paste out in a thinner layer over the whole core (not super-thin like the AS5) and then pressed down hard and twisted the heat sink to get good coverage and to squeeze out any extra paste. After doing that, I ended up with the temperatures that I posted here.

I'll let the system run under full load for a week or so then post back again if I see any changes (good or bad) in my temperature readings with the diamond paste.


Edit: It's been two weeks since the initial application and my CPU is still running happily at 39C with the same 22C ambient temperature in the room. It may be because I already spread the paste out more than the recommended squirt and squish method but it doesn't seem that my temperatures are any different after letting the application cure for two weeks.

I'm not complaining since I got better results immediately after applying the ICD7, but since a request was made to post updates over time, this is my contribution.

 

[Insidious]

(internal temperatures measured using CoreTemp utility )

Opteron 170 @ 2.6GHz
Stock (4 heat pipe) AMD HSF

First test: (Arctic Silver5, in place ~ 1year)

Idle: Core0 37/38C Core1 38/41C
Loaded: (1 SMP fah client) Core0 63/64C Core1 67/68C

Second Test: (ICD7 thermal compound installed with glop and squish method described above. Results ~2 hr after installation)

Idle: Core0 38/39C Core1 39/41C
Loaded: (1SMP fah client) Core0 64/65C Core1 68/70C

Comments: I'll let it run overnight to see if results improve

Third Test: (ICD7 thermal compound installed with glop and squish method described above. Results ~36 hr after installation)

Idle: Core0 35/36C Core1 38/39C
Loaded: (1SMP fah client) Core0 63/64C Core1 67/68C

Comments: Some reduction in idle temps from original (AS5) setup, but loaded performance is no better. (It wasn't any different so far)

------------------------------------------------------------------------------------------------------

Q6600 @ 3.46GHz
ThermalRight Ultima-90 HSF (90mm Panaflo M1 fan - push) HSF mounting mechanism modified by addition of washer on each tensioner to increase spring compression. Very tight mount.

First test: (Arctic Silver 5, in place ~ 3 months)

Idle: Core0 42/43C Core1 39/40C Core2 41/42C Core3 37/38C
Loaded: (2 SMP fah clients) Core0 62/63C Core1 60/61C Core2 60/61C Core3 57/58C

Second test: (ICD7 thermal compound installed with glop and squish method described above. Results ~ 2 hr after installation)

Idle: Core0 41/42C Core1 38/39C Core2 39/40 Core3 37/38
Loaded: (2 SMP fah clients) Core0 62/63C Core1 60/61C Core2 60/61C Core3 57/58C

Comments: VERY consistant results with other PC. Idle is slightly improved and there is no change in loaded heat removal capability of the HSF. Will run overnight...

Third test: (ICD7 thermal compound installed with glop and squish method described above. Results ~ 36 hr after installation)

Idle: Core0 41/42C Core1 39/40C Core2 39/40 Core3 37/38
Loaded: (2 SMP fah clients) Core0 62/63C Core1 60/61C Core2 60/61C Core3 57/58C

Comments: Both PCs temperatures have not trended much in this short time. Next data in a week or so......

Overall comments so far:

I have not seen anything ground breaking with this product, but it is a good one. I am testing it against meticulously applied Arctic Silver 5 which is certainly no slouch of a thermal coupling medium for PC HSFs.

The pros:
I like the viscosity of this product. It is nice not to have to spend a half an hour spreading and smoothing it to get a good installation. -> I like this aspect of ICD7 better than AS5

The thermal performance of this product seems to be excellent immediately. No waiting for days to see stabilized temperatures in my CPUs

The cons:

I have to use a lot more (a 'pea' as opposed to a 'BB') of the product per installation. I see from casual shopping at frozenCPU and Newegg that this product is about twice the cost on a per gram basis and I have to use 2 to 3 times as much of it. So that comes out to 4 to 6 times as expensive for comporable performance.

What would I buy:

If the price was as comporable as the performance I would choose ICD7 because it is so easy to apply using the recommended "glop and squish" method. But at such a price disadvantage, I would have to go for the AS5 based on what I know right now.

note:

I don't like the way AS5 tends to separate (oil/solids) over time, but love the way it's thermal performance lasts despite that behavior. And, my HSF is still easy to remove, even after a year or so. I am reserving a final opinion of ICD7 until I have it in there for a significant amount of time to see how it ages.... maybe that will put it over the top. Stay tuned

 

[ICD7]

ICD7 is also a great metal polisher

 

[PCTC2]

Haven't gotten it yet. Just reserving...

Motherboard : Abit AW9D 975X
CPU: E6400
OC = 8 * 400 MHz
VID = 1.325 OC vcore = 1.375
HSF = Thermalright Ultra 120 using Swiftech Generic
Ambient = 73F
Temps taken using coretemp 0.97

Before temps:
@3.2GHz
IDLE ??/?? C
LOAD ??/?? C

After temps:
@3.2GHz
IDLE ??/?? C
LOAD ??/?? C

Motherboard : eVGA 680i-A1
CPU: Q6600
OC = 9 * 333
VID = 1.325 OC vcore = 1.375
HSF = Swiftech Apogee GTX and Thermochill PA 120.3 using AS5
Ambient = 73F
Temps taken using coretemp 0.97

Before temps:
@3.0GHz
IDLE ??/?? C
LOAD ??/?? C

After temps:
@3.0GHz
IDLE ??/?? C
LOAD ??/?? C

 

[caferace]

What are the known "hot" DC projects? I know that OGR with the BOINC wrapper at yoyo@home is a killer when it comes to CPU heat. Any others? Peter, I know you keep tabs on this stuff....

-jim

 

[PCTC2]

Originally posted by: caferace
What are the known "hot" DC projects? I know that OGR with the BOINC wrapper at yoyo@home is a killer when it comes to CPU heat. Any others? Peter, I know you keep tabs on this stuff....

-jim

Sorry, but just to keep it straight, could you start a new thread and keep this for ICD7 results only?

--Not a Mod. Just a DCer.

 

[Fardringle]

Originally posted by: PCTC2

Originally posted by: caferace
What are the known "hot" DC projects? I know that OGR with the BOINC wrapper at yoyo@home is a killer when it comes to CPU heat. Any others? Peter, I know you keep tabs on this stuff....

-jim

Sorry, but just to keep it straight, could you start a new thread and keep this for ICD7 results only?

--Not a Mod. Just a DCer.

Although not specifically ICD7 results, I think caferace's post is relevant since some DC projects don't heat up the CPU nearly as much as others, so knowing which project(s) are more likely to really test the effectiveness of the paste can give us more accurate results.

 

[narzy]

I'm getting a few degrees warmer right off the bat, I'm sure it is an installation issue, this stuff is a big pain to work with, I'm sure I didn't use enough/spread it properly/used too much / the intel hold downs aren't applying the right amount of pressure, but it shouldn't be working less then 5 year old artic silver...which was crusted on like concrete...so I will re-evaluate tomorrow and see if I can't do better.

note posting time PDT

 

[Assimilator1]

Originally posted by: Fardringle

Originally posted by: PCTC2

Originally posted by: caferace
What are the known "hot" DC projects? I know that OGR with the BOINC wrapper at yoyo@home is a killer when it comes to CPU heat. Any others? Peter, I know you keep tabs on this stuff....

-jim

Sorry, but just to keep it straight, could you start a new thread and keep this for ICD7 results only?

--Not a Mod. Just a DCer.

Although not specifically ICD7 results, I think caferace's post is relevant since some DC projects don't heat up the CPU nearly as much as others, so knowing which project(s) are more likely to really test the effectiveness of the paste can give us more accurate results.

Agreed.
I've found Prime 95 running small FFTs to raise temps far higher than SETI, DPAD or F@H (by about 8-10C IIRC).

Btw folks be careful using speedfan, some versions don't read internal temps correctly on AMD CPUs & (unless theirs been a recent update) it doesn't read C2Q temps accuratley by default either. Confirm temps using CoreTemp linked above.

 

[ICD7]

Troubleshooting note - There are three main issues that that we have Identified That can affect performance.

1. Too little grease applied- Most people are taking care now that we have focused more attention on it and that problem has become minimized if full pea size is applied.

2. Mounting pressure with push pin type mounts or overweighted sinks for the manufacturers supplied hardware. Some retention hardware does fatigue over time and may have to be replaced or reworked. You people are experienced users and are likely to identify any issues there on your own.

3. Temp measurement accuracy - In the expanded application notes That was emailed to you I provided a link to an Intel Thermal management Doc. In there they have a subsection (4.2.7.2) on the Internal diode accuracy where it points out that depending on the components used by the vendors accuracy can range from +\- 3C to +\- 1C. So in some cases it may just not be possible to get a reasonable measurement or even a result depending on your components as your error may be greater than performance range tested. It is interesting to note that in any thermal testing or component validation of consequence they recommend cutting a groove in the IHS and mounting a thermocouple for temperature readings rather than use the diode.

Optionally you might try another system if available.Otherwise do not waste too much effort after review of the first 2 troubleshooting recommended solutions just report your results. The overall averaging we do at the end tends to factor out errors, + or -


http://download.intel.com/desi...m4/guides/30255304.pdf

 

[Insidious]

Originally posted by: ICD7
3. Temp measurement accuracy - In the expanded application notes That was emailed to you I provided a link to an Intel Thermal management Doc. In there they have a subsection (4.2.7.2) on the Internal diode accuracy where it points out that depending on the components used by the vendors accuracy can range from +\- 3C to +\- 1C. So in some cases it may just not be possible to get a reasonable measurement or even a result depending on your components as your error may be greater than performance range tested.

ICD7, you need to take a look at CoreTemp utility
The key points are the accuracy and the fact that the motherboard venders cannot ferk up this measurement.

I think you will find it a much more practical method of reliable, repeatable core temperature measurements than heat sink temperature measurements with embedded probes.

-Sid

(YMMV )

 

[ICD7]

Insidious

Use what works for you, it works for me. I am interested in the total picture. I was just making note.

 

[Assimilator1]

I believe the point was that coretemp could get rid of some of the inaccuracy.

Just FYI I've not got my paste yet.

 

[RaySun2Be]

I've got CoreTemp, SpeedFan, and Sissoft Sandra, and all three seem to be reporting the same or similar temps, on a quick baseline test. I will do more testing this weekend.

 

[ICD7]

On application size

We came up with the pea thing when we analyzed some of the first results we got on our first giveaway and on the subsequent HardwareLogic giveaway overall user results improved almost .75 C with the use of a pea sized call out.

Smaller than a breadbasket but larger than a grain of rice...We realize it probably is a little overkill but want to insure that people get a good result, We are trying to remove the subjective guesswork out of application.

One Idea we had is to put a window on a graduated syringe and call out something like 1/10 of a CC per application.

Another is we have this nifty screw type syringe which we can mark a specified amount by 1/4 turn 1/2 turn whatever it works out to be.

Anybody else have a method that might work? let me know we are open to suggestions.

Also to note
Diamond weight vs silver weight-It is easy to forget weights and volumes. AS5 is 3.5 grams silver and fills 1/2 cc on a syringe. ICD7 is 1.5 grams and fills 1/2 cc on syringe - - a carat is .2 gram. For OEM and system builders we are going to be selling a 24 carat tube in the near future.

 

[Markfw]

Originally posted by: ICD7
On application size

We came up with the pea thing when we analyzed some of the first results we got on our first giveaway and on the subsequent HardwareLogic giveaway overall user results improved almost .75 C with the use of a pea sized call out.

Smaller than a breadbasket but larger than a grain of rice...We realize it probably is a little overkill but want to insure that people get a good result, We are trying to remove the subjective guesswork out of application.

One Idea we had is to put a window on a graduated syringe and call out something like 1/10 of a CC per application.

Another is we have this nifty screw type syringe which we can mark a specified amount by 1/4 turn 1/2 turn whatever it works out to be.

Anybody else have a method that might work? let me know we are open to suggestions.

Also to note
Diamond weight vs silver weight-It is easy to forget weights and volumes. AS5 is 3.5 grams silver and fills 1/2 cc on a syringe. ICD7 is 1.5 grams and fills 1/2 cc on syringe - - a carat is .2 gram. For OEM and system builders we are going to be selling a 24 carat tube in the near future.

Here is my idea for the perfect amount:

Another is we have this nifty screw type syringe which we can mark a specified amount by 1/4 turn 1/2 turn whatever it works out to be.

Using that:
"turn screw until paste appears, then 2 more full turns, in a blob and stop. Then press and twist heatsink using small rotations, many times. Then secure HSF using manufacturers retention device. Boot and watch temps".

 

[ICD7]

Originally posted by: Fardringle
I don't have the "after" numbers yet, but here is my setup and temperature readings with a roughly 4-month old application of Arctic Silver 5 on the CPU and stock paste (which was dried and flaky) on the video card:

Motherboard: Biostar K8NHA Pro
CPU: Socket 754 Sempron 1600+
Overclock: 8 x 275.5 = 2204 MHz
HSF: Thermalright SI-128 with Arctic Silver 5 paste
Video Card: eVGA 6600GT with stock heat sink
Ambient temperature: 72 Fahrenheit (~22 C)
Temps taken using Speedfan and PC Wizard 2008 (I got the same numbers in both and CoreTemp doesn't like my motherboard for some reason)


Before temps:
CPU:
Idle: 35-36 Celsius
Full Load: 41-42 Celsius
Video Card:
Idle: 56-57 Celsius
Full Load: 79-81 Celsius

I waited about 24 hours after applying the stuff so it could settle in and I could have the same ambient temperature as the initial readings.

After temps:
CPU:
Idle: 32 Celsius
Full Load: 39 Celsius
Video Card:
Idle: 55-56 Celsius
Full Load: 65-68 Celsius

I got a small improvement on my CPU, and a rather large one on the video card. I already had AS5 on the CPU so I wasn't expecting miracles, but any improvement at all is a pleasant surprise. It should be just enough that I can keep my overclock through the hot summer instead of having to dial it down a bit to keep the system stable.

My video card should be much happier now than it was with the dried "toothpaste" garbage that was on it before.



I do want to say that the "squirt and squish" method did not work at all on my CPU. When I did that, the paste only ended up covering the center 1/3 of the CPU heat spreader in a thick layer, and my idle CPU temp hit 43 C just in the BIOS so I immediately shut the system down. I took the heat sink off and spread the paste out in a thinner layer over the whole core (not super-thin like the AS5) and then pressed down hard and twisted the heat sink to get good coverage and to squeeze out any extra paste. After doing that, I ended up with the temperatures that I posted here.

I'll let the system run under full load for a week or so then post back again if I see any changes (good or bad) in my temperature readings with the diamond paste.

I appreciate the detail on the application stuff as we would like to make this application as simple and fail safe as possible and is a work in process. How did you spread the paste after the initial squirt and squish. you think it is a feasible step to add to the application?

I am very interested in the long term stress you guys put on you systems, Nobody has run these kind of tests to date. great job.

 

[ICD7]

Originally posted by: RaySun2Be
I've got CoreTemp, SpeedFan, and Sissoft Sandra, and all three seem to be reporting the same or similar temps, on a quick baseline test. I will do more testing this weekend.

I would not stress to much about the temp programs, what I am looking for is consistency. I am already getting more professional feedback than I have from the other groups and am very happy with the reported results. You are, as we say a high quality sampling group

 

[ICD7]

"""Using that:
"turn screw until paste appears, then 2 more full turns, in a blob and stop. Then press and twist heatsink using small rotations, many times. Then secure HSF using manufacturers retention device. Boot and watch temps". """

Referring to Fardringle's post the application procedure needs a little "something" after the squish part. Alot of users complain if they try to spread the paste with a razor blade or finger in a baggie, you think it's enough to put in extra twists?

 

[caferace]

No numbers yet, but I received my ICD7 today. Looking forward to replacing a lot of aging material this weekend on some of the hotter boxes.

Carry on.

-jim

 

[RaySun2Be]

Originally posted by: ICD7

Originally posted by: RaySun2Be
I've got CoreTemp, SpeedFan, and Sissoft Sandra, and all three seem to be reporting the same or similar temps, on a quick baseline test. I will do more testing this weekend.

I would not stress to much about the temp programs, what I am looking for is consistency. I am already getting more professional feedback than I have from the other groups and am very happy with the reported results. You are, as we say a high quality sampling group

Oh, I'm not stressing about the temp monitoring programs, several I haven't used before, and I just want to do a validation that I'm getting consistent readings before running the official tests.