IC Diamond 24 Giveaway/ Reliability Survey

3 years later wooow this is amazing!.

Ok, late but here is it . I sold the laptop a few years ago BUT the system remains the same and my friend let me take a benchmark . I've got here the new tests under heavy overclocks:

Original temps: http://forum.notebookreview.com/gateway-emachines/372333-official-icd7-thread-12.html#post4854660

Current temps with strong overclocks:


As I have said before this thing is GREAT. I only use this paste on my laptops, nothing else .
Ambient temps 26C.

 

Much appreciated for the updates MexicanSnake, ribtits and others - Been out of circulation due to some health issues but now back on track and will review progress and update over the next few days

 

I believe the reliability issue is going to resolve down to fundamentals of mechanical of Contact and Pressure (C/P)

Before I got side tracked I did manage to get two C/P tests - One had only approx 50% area in contact and the other the pressure was < the range of the paper, under 28 PSI. While it was a no result test it still was a significant result.

C/P are closely related, as in the charts below if your contact area is only 50% your Watt density will be 2X vs a full contact mount. So full contact mount might look like MexicanSnake 's 3 years where temps remain stable over a long period of time and those that have a lesser contact say for example 10 or 20% will over stress the compound and cause the compound to fail perhaps in just a couple of months.

I would not underestimate the effects of heat density, I recall a post/ picture back in the old Duron days where the overclocked chip had welded itself to the base of of an aluminium heat sink - it had melted the aluminium so the chip was actually embedded 20-30 thousandths into the base!!

What I would like to do is get more samples to generate some charts similar to the ones below - C or P vs Reliability etc.

I need an increased sample size of around 20 -30 tests and will supply a free 24 carat tube along with some C/P test kits in exchange for the information - any takers?

 

OK, I'm in Andrew .

Many of my classmates need a repaste... I could provide more info!. I also need the kit to test the pressures.

 

I have a secondary computer so I can take apart my laptop more willingly now, I wouldn't mind doing a contact pressure test for you

I have a CPU die and a GPU die though, if that helps you out, I have to do both anyway when I remove it.

 

Blaze-Senpai - Thanks for participating - PM Me your address and I will get it off to you this week.

 

Mr. Snake How many test kits to start? I have your mailing address

 

I'm in can do a CP test on both of my laptops. My classmates' laptops are also in need of repasting so the more the merrier

 

If I ever repaste again for whatever reason, then I'll do the CP test.

 

So Andrew you are finally back.

Which of the C/P pressure test is mine? I sent you one last year.

Right now I have two dies ready for C/P testing - one CPU and one GPU die from a laptop.

If you don't remember my mailing address then PM me.

 

Still not much of a change. 74.0C after 25 IBT runs.

 

PM me your address and how many laptops you are planning to do

Anybody else want to try a test? Mailing first test group on Monday.

 

New laptop.

Idle -

CPU:
49°C
48°C
40°C
43°C

GPU:
45°C


Load -

CPU:
90°C
90°C
86°C
83°C

GPU:
83°C

Results taken from my latest Vantage run:

Spoiler

Ambient temp is a constant 20°C.

 

I am updating results to date -

Thanks to all that have taken the time to post their results. Should have it up in a day or so just have to recheck some points.

Reminder to be as complete as you can with Idle and loads and please note AMBIENT OR ROOM TEMPS In HVAC environments room temp can easily fluctuate 5- 8 degrees if no HVAC daily temps can easily fluctuate 20 degrees morning to night - or July vs January can be significantly different as another example.

In processing or normalizing the data we subtract ambient from the CPU temp to get the temperature rise above ambient - fairly easy to be 5 -10 degrees higher summer vs a spring/fall temperature and 5 -10C cooler vs spring/fall in winter with a 10-20 degree fluctuation year round we will still pick up a trend without it but it will be hard to nail on a time time scale so please be as complete as possible.

 

Preliminary Update - any errors or omissions on my part let me know for corrections.

Pretty good, some go the distance - my money is on contact issues for those that have had temperature increases, C/P tests will be interesting

People need to update their results many are past due so a reminder here. Much thanks to those that took the time to update their results it is appreciated.

Also I am putting together a Wall of Shame for several who failed to keep their end of the bargain to post results......

 

I did not post the idle temps, but if you want, i can start posting those too. Anyways, i got the results for up to this month, but forgot to post them yesterday, will upload them sometime today when i'm on my JH including screenshots. I'll have to look at my data a bit carefully, but i think i saw a trend of the GPU and CPU having an influence of about 1C on each other when i took a quick look at my Excel spreadsheet.

The verdict so far after 9 months is that the temps are still quite similar with a regular dusting of the heatsinks.

 

Just want to remind Andrew in case he missed this.

 

Since I have alot of Extra ICD including the free one provided I decided to show just how well the install holds up. Now I can run Prime95 on my Q9200 at 2.95 GHz but I will regularly see 95c under load with 50c idle at 72F ambient. So of course I only run it to check upon the system. I have fairly heavily modified my original 7805 and this particular install is 9 months old.

So Here is a disassembly done this evening. Since the images are large I will just link to them. I had to pry the shims from all chips as ICD had this thing almost like it was welded together. Please pay particilar attention to the cleaned shims as you can see the defined heat transfer areas.

In the end if you want the best heat transfer and for the installation to last, go with ICD. And a disclaimer yes I have recieved one free tube of ICD but have paid and used several others.......................

www.pcambrosia.com/ICD/ICD_0000.jpg
www.pcambrosia.com/ICD/ICD_0001.jpg
www.pcambrosia.com/ICD/ICD_0002.jpg
www.pcambrosia.com/ICD/ICD_0004.jpg
www.pcambrosia.com/ICD/ICD_0006.jpg
www.pcambrosia.com/ICD/ICD_0008.jpg
www.pcambrosia.com/ICD/ICD_0010.jpg
www.pcambrosia.com/ICD/ICD_0000.jpg
www.pcambrosia.com/ICD/ICD_0012.jpg
www.pcambrosia.com/ICD/ICD_0014.jpg
www.pcambrosia.com/ICD/ICD_0016.jpg
www.pcambrosia.com/ICD/ICD_0018.jpg

Edit;
I should note in looking at the Shims the HS side did not look like there was 100% high heat transfer and it blended off from where the actual chips on the other side of the shim contacted. This time I assemble in a different order.

I first worked the shims into the HS with ICD to be sure that the 3 shims slightly squeezed out TIM on all sides. I then dabbed the chips and worked the HS into the chips. I did the work ins with a gentle small circular motion.

In all it seems to cool slightly better, about 2c under 2.95 GHz and Prime 95 and 3-5c better under normal work loads. I should note this is not for the faint of heart as ICD is abrasive and can end up scoring the CPU and/or heatsink...................

 

Nice to see life in the thread again.
Il do some measurements in the weekend, gotta pick up a can of air to clean the fans before the readings.

Im guessing the results will end up the same as my other readings.

 

Yeah, this is an awesome TIM. Probably overkill for most but some people live for overkill. I hadn't yet done an intentional pull yet to show one of the best things about ICD, how well it holds up over time under extremes.

My system is up 24/7 and when I am using it normally see's temps in the 80c range and up. I also did a couple of weeks folding too running at 90c. For a DTR I definately abuse mine regularly. Normal silver and ceramics I've used had to be replaced every 3-6 months from showing performance degradation and each pull you could see the failures of the TIM. I have yet to do a CPU change etc. and see where ICD failed in any way............

 

TANWare,

Thanks for taking the time to post the pics, as always it gives us more background and a feel for the process particularly with the higher stressed systems.

Contact and pressure look good from the paste impressions any interest in doing a C/P test? I could use a good C/P test for comparison as you have gotten great results yours would be an ideal comparison example.

Thanks again for taking the time, much appreciated

Andrew

On the abrasion front

Most thermal compounds are liquid sandpaper or lapping compounds by definition.
Light gray compounds are usually/mostly made up of aluminum oxide. AS5, AS Ceramique, MX-4, Shin-Etsu, Prolimatech PK-1 all contain aluminium oxides.

Aluminum oxide is what they make sandpaper out of.....MOHS Hardness scale 9

Diamond on the MOHS hardness scale: 10

Copper MOHS: 3

To be abrasive you just have to be harder than the material to be cut.

Most thermal compounds have particle sizes in the range of 600 - 800 same as your 600 - 800 sandpaper

Abrasive Grit Sizes

It's all about the cutting speed

If you are ambitious, you can calculate cutting speeds and material removal between diamond -- the hardest, (cubic boron nitride is actually harder than diamond) and the second hardest -- aluminum oxide, you will find an insignificant difference.

For the insignificant distance traveled, equivalent pressures, You will not see any difference between the two under normal use.

Abrasives have to move to work.

Take tooth brushing as an example

While if you use normal manual tooth brushing techniques this is generally not an issue however if you are overly rigorous you can cut through the enamel quite easily and to note my dentist observed that with my use of a sonic toothbrush at 31,000 brush strokes per minute I had cut through the tooth enamel at the base of some of my teeth and recommended a less abrasive toothpaste.

Silica's hardness is about 7 (quartz) Copper 3 so ordinary toothpaste will polish a copper penny( or IHS) quite easily if you make a determined effort to put it all in perspective.

A less obvious example here but will make sense when you think about it. Your own copper shims pictures show evidence of a light oxidation, perhaps 1/10 of the thickness of of any laser etched serial number on an IHS and has managed to stay intact through a casual cleaning process.

Normal application and removal is not a issue but if you wanted to you could remove the oxide by polishing with just about any thermal compound except a zinc oxide one and even tooth paste if you were so inclined.

 

It is always appreciated when full/complete data is provided, thanks.

 

Thanks - if you add in idle temps and ambient room temp would be a big help

 

I assume 50 person limit is already passed but if not then...
On my Honor in exchange on free tube I swear to post my temp results and update it every 3 monthes. I type from kindle at this moment so sorry for not the exact same form.
Please choose me. It' been just half a year since 1-st post which is not so long...right?

 

It will be a few weeks but surely I can run a C/P. Always good to add to knowledge bases. I was a machinist for 10 years so I have no issue with the abrasive quality of ICD. Because of my experience though I do not recommend my installation method to others, I know by feel when to stop before the grind!

Edit; I should note once, quite a while ago, I applied direct pressure on AS5 thinking it was not abrasive enough to matter. It squeezed out too much TIM so I pulled the HS. To my surprise radiating from the center you could see score marks just from the TIM moving itself out to the sides on the chip and especially the soft copper HS...........

 

Current temps:

 

Ambient room temp 30C, idles at 40-45C

 

There is another ICD bashing thread going on here, in the guise of what is the best TIM. I just had to point out that while ICD can be a PITA to remove this is actually a very good thing. If it is so hard for us to remove it is even more so for the system and its normal processes to remove/degrade over time. So this is actually a good thing for a long term installation.

The majority of those wet TIMs boil out, or dry out and crack. Either causing less contact and thermal transfer. So unless you plan to every 3-6 months replace the TIM then those wet TIMs are not the way to go.

I find it amazing people complain about exactly what they want! A TIM that is efficient and will hold up a long time. Then through a hissy fit about it being hard to remove, DUH........................

 

Thanks for the update - great presentation of the numbers, certainly makes my job easier

 

Much appreciated

 

New temperature readings.
You can check my older posts here.

Tools used:
HWiNFO64 v3.93
Prime95 v26.6
MSI Kombustor v2.2.2

Ambient temps should be a little higer than last time i measured but since i force my fans at 100% it should not affect it too much.

Idle temps:


CPU stress test temps:


GPU stress test temps:

 

had to repaste and remove chims my cpu will upload new results as soon as i have time and let the paste settle and then run some benches will use a better format this time
reason of repaste was geting very hot light load temps like watching a streeming movie would give me 78c on cpu with a ambient of 20c
my gpu's are still running very cool at same moment like hiting 37 to max 39c ca 1c higher then idle.
right now wrighting this on hottest cpu core 49c pendaling up to 52c and back again

sorry for not posting more frequent

 

We have been reviewing IC Diamond reliability results - From the 15 reported results so 30% Measured approx +15 C or more and 10 tests or 70% indicated little or no change.

The spread is interesting in that some show no degradation from 6 months to 3 years and like technos show a rapid decline in as little as 6-9 months and others like InfectedSonic and MexicanSnake have no change after 2 or3 years.

So why the difference in results?




Note the following bell curve of desktop CPU's, the population tested was 652 end users with before and after applications of IC Diamond.

We Followed through with some Contact and Pressure (CP) tests on the same group with a sample size of a little less than 10% around 55 tests.

Keep in mind that it is not directly comparable as these were desktop systems vs laptops here, on a laptop the delta temps will be double due to the higher heat flux which is why you see greater improvements.

Thermal compound is about as simple as it gets but 3 factors are key for performance

1.) Proper application amount

2.) Good Pressure

3.) Good contact

Note Arrow 1 - In this region left to right it was found with the C/P testing that those who gained in temperature or had a minimal improvement had some "blend" of poor or marginal C/P. Possibly a good mount vs a bad mount or 2 bad mounts not enough compound, ambient errors etc..

Note Arrow II - As you move right the blend of C/P increased along with performance and I would generalize here and say that between 3 and 5 or 6C improvement improvement correlated with expectations that being somebody @ 60psi vs another @ 80psi might see another deg. C improvement above the mean on a diminishing return basis as well as those whose baseline compound was a poor performing stock compound or ran at higher OC increasing the delta temps.

Note Arrow III - The higher returned values increasingly as you move right are more likely to be the reverse of Arrow I, things like a bad initial mount vs a good IC Diamond mount or ambient errors, failed base line compound etc. for an overstatement of performance which you would not see in a more controlled test

Probably can see where I am going with this as I said thermal compound is simple and straight forward, we emphasize application amount enough that there is less likelihood of a problem there it really boils down to whether you have a good mechanical joint or a marginal (bad) one while the sample size is small, only 15 we generally we like to see 30+ 15 is large enough to indicate the approx. trend.

Thermal Compound is all about mechanics, electronic oriented people often overlook the fact.

I would also note that TanWare's post on the previous page on an 11 month install with copper shims showed great contact, good pressure and the compound still liquid. I assume the mod is done to improve performance which is counter intuitive as it added a layer of resistance but works due the bigger benefit of good contact and pressure?



What's interesting to me is that the publised survey I posted/linked to 6 months ago had Laptop failures at 30% in the first 2(3 yrs. ?) years, 10% were attributable to issues like dead pixels or impact damage with 20% being component failure. Note Arrow I, In our sample study of C/P and those that had poor or marginal results associated with limited C/P approximated 20%, maybe the Pareto's Principle or the 80-20 Rule at work here?

I would like to get some C/P samples from the high temp guys to see what they look like if any would like to run a test......

Record those Ambient Temps!!!!

Will be back for an edit

 

This might seem counter intuitive at first, but given that the thermal resistance (in K/W) is also a function of the overall surface area, i'd say you have two competing effects going on there. The conduction resistance in cartesian coordinates is thickness/(k*A) if that interests anyone.

1. The thermal paste layer adds extra resistance where both pieces touch each other perfectly due to the lower thermal conductivity of the paste.
2. The paste drastically reduces the thermal resistance where the shims weren't touching each other due to the really poor thermal conductivity of air which is k=0.03 W/mK at 76.85C (350K) and even lower at lower temperatures. That's real bad thermal conductivity.

Add to that that with good pressure, the layer of paste will be very thin, and effect #2 is greater than #1.

One more question, i haven't looked through the thread, but were those systems clean of dust when the testing was done. That is one huge factor to take into account. I always remove as much dust from my system before taking temp readings and i assume everyone probably did the same, but it never hurts to ask.

 

Right, contact is everything. While the shims add slight thermal resistance three primary things were accomplished.

1.) This shim(s) thicknesses were calculated out so that between the CPU and NB the heat sink stayed level and flat so that, unlike the NB with the original pad, would not try and torque the heat sink from laying flat on the CPU.

2.) The larger copper area of the CPU section ensures all copper contact between the two cores. This would not have been an issue with a C2D but is an issue with the C2Q design.

3.) Raising the default height of the heat sink makes for more pressure to both the CPU and NB as it is tightened down. This of course increases thermal transfer.

Edit; now they do also ever so slightly add to thermal mass but the shims are so thin I do not think it is of any significance...............

 

On desktop systems we had the same issue to a lesser degree as most OC's/hobbyists tend tend to be a little obsessive about their temps and systems in general. I assume you have some but not enough to significantly affect results and most are noted as observed.

Generally you have to play the averages due to the limitations of public forum testing and getting everybody on the same page is like herding cats.

Some will make the argument that the result might be skewed as the system might not have been cleaned before the baseline test, it follows though that if some % of the sample group is not rigorous about dust in the first initial test you will have a similar % on the other side of the equation on the updated result so they cancel each other out on a large enough sample.

Take ambient temps for example. Few reporting results have what you would call accurate equipment, thermostat wall thermometers probably for the most part, some will be high some will be low and some will be dead on, average it out on a decent sample size and you will be close to a final number.

Take the William Tell example where he makes a 100 tries at hitting the apple, some hit high, some hit low, some to the left and some to the right. If you average out the locations of where the arrows hit will give you a pretty good idea of where the centre of the target lies, larger the sample and accuracy of the inferred centre location increases.

With approximatively documented 850 forum tests I had one group of undervolters of about 30 in the sample population with improvements of only +two degrees improvements due to the lower power. When averaged in even though an outlier only affected the oveall average a couple of hundredth's of a degree. The sample size will be an issue as the challenge is to get people to update their results.

I only recently became sensitized to the dust issue in laptops as I bought a couple of laptops in the fall and when I noted the system slowing up a couple of week ago I installed speed fan to see what was up and found I was running 94C with a 65 ambient with my usual 30 browser windows open and a video running. I also have a secondary laptop cooler that reduces temp about 8C.

I hit the air ports with my shop vac and temps dropped back to 70C under the same load conditions. So in 6 months temps increased 24C hot enough to prematurely kill a thermal compound. So as you rightly point out people should be mindful of dust factor and I will ask from now on.

Great data presentation btw

lol, Apple should have used IC Diamond 120C! Might hit 130 C with summer temps on the horizon.

 

thats in F yes ? im not stupid just woundering if its a typo since 6 is just over 3 on the keyboard

never use a vac to clean you laptop it can set of an staticspark and if un lucky your system is Fubar
use a conpresed air container or do like me every 30 days i open my bby and remove the fan's clean the cooling flenshes with first with conpresed air then a quetip with medicly clean 98% alcohol
but remember either you use vac or air you need to put some thing in the fan so it wont spinn up it can fry your bairings
sorry for my spelling

 

Opening a laptop every month is a bit much to be honest. If you dust it out every 2 or so weeks you don't give enough time for anything to build up in there... unless you live where I do, but even still. I've only opened my laptop once to dust it out and it was never really as bad as I've seen some, where the dust itself makes a filter on the heatsink

 

like you can see in my signature i have the np7280 that is whery easy to open only 5 screws and the hole cover plate of cpu and gpu's is un covered
the toshiba i only blow out with the air canester and open maybe every 6month since its a pain in the a to open ca 25 screws and have to dismount every thing
but i must agree im a bit obsessive with my 7280 but after paying 5k for it i dont whant to risk any thing wile clocking it

 

my updated test will take a bit longer am going to do a modd of the cooling array it self to be able to get the same contact pressure i hade with the shims
i will update later but it might take a week or so since right now i feel like im dieing got neumonia and im atibioticks :-(

 

These are my current Temps after running Orthos for 20mins. Room temps 20 degrees C.

 

Ah I got the test kit today, I need to do some thermal tests before I rip apart my HDX again though for the 9 months in test though, I should have those up tomorrow.

 

I've to say this because I always see a huge bias towards IC products. Maybe due to these great advertisings and offers that they make on forums, which is really good and I think everyone appreciates.
I tried IC D24 once and I regret I spend so much money for just so little performance. For the huge price I was expecting it to be in the top (if not the best) of thermal compounds but it was very disappointing to see that temps were 1ºC higher then the previous paste I had which was Arctic silver 5, which is nowadays pretty average.
I've tried later with others with much better results. I was a sys builder so I had to tried dozens.
imvho IC D24 is a good thermal paste but isn't worth the money because there are many other compounds available with the same and better performance at a lower price.

 

So, what would you suggest / recommend then?

 

I just don't like IC diamond because it may blemish the cpu due to the "polishing" effect. it may void cpu warranty. That's a big issue...
you have so many to pick from. there's dozens of reviews around. I prefer something not expensive, non conductive, easy to spread, good performance.

 

We are doing a series of video's on IC Diamond one of them deals with compound application and removal. In the segment on removal we generate an oxide layer on copper that is only 100 Angstroms (a hydrogen atom is about 1 A in diameter, a carbon atom is about 2 A in diameter) or 10 nanometres thick (1% of a micron) Your typical laser etched lettering is approx. 200 to 500 % thicker @ 2 to 5 microns.

In this extreme example we load at 50lbs then remove the compound with Zero damage to the oxide coated layer.

We then demonstrate the delicacy of the coating by taking a soft Kleenex and polishing it to a bright copper shine in a few seconds and then follow with a second example using a bare finger to get the same polishing/ oxide coating removal with a few seconds also.

With proper removal/application IC Diamond is less abrasive than a soft cotton cloth, Kleenex or even a bare finger.....All it takes is a decent solvent and a minute or two of your time to do.

I'll have the Video up in a week or so.

 

Independent end user tests





Independent end user tests on this forum with Approx 2X performance over due to higher heat flux with the smaller die.

 

Blaze-Senpai & others involved in the test

when you take the test write the test number that is on the data sheet on the unused corner of the printed result so I can link the thermal to the C/P Result.

Much appreciated

Andrew

 

Thanks for the detailed explanation. I gave ICD a try last year and it still performs very well under both the idle and load states (it's about 40C as I'm typing this message with a few apps opened in background), so hopefully there was no any damage to the CPU / GPU die surface but we will see next time I decide to re-paste again.

 

April update. Temps on CPU went up a bit (about + 2deg C on avg) on full load to around 70
GPU temps holding stable for the most part under full load (~68 deg C on avg)