Thermal Paste Test

I think its about time you did a review on Thermal Paste so we know which one provides the best heat transfer

 

Thanks for the suggestion, unfortunately:

Thermally testing TIM is actually really, really difficult. To make the test fair, you have to use the same CPU and the same HSF. The problem is, they're only pristine - i.e., 100% clean - *once*, which is when you first get them out of the box. Having talked to engineers at HSF and TIM manufacturers (Akasa and Zalman) while researching the article, we were told that although TIM clean gets most of a layer of TIM off, it doesn't remove it all - this is because TIM is really designed to be applied once, not constantly applied and removed.

As a result, as the test goes on, you're going to get a build up of TIM on the CPU and the HSF and this will therefore compromise the results. Then you've got to contend with the fact some TIM needs time to "cure" - and this can be days and/or weeks.

It's tricky, and we'd rather not present test results just for the purposes of having some numbers in the article - if results are being used, we want to be sure they're 100% accurate.

 

I know mx2 and mx3 seems to be a popular choice on these forums. (Is there an mx4?) Me? Iam happy with Arctic Silver 5.

 

I guess the only workaround would be to use more than one chip and reverse the application order, that is the first used on one chip is the last used on the other and average out results.

But then you have to contend with the inherent differences in the chips themselves as well as the "cure" factor...

Tricky.

 

But then the middle chips will always be the middle chips?

It will be very hard to test, best just sticking to recommendations.

 

As a newcomer to TIM, are some types clearly better than others?

And are they better on all CPUs - or do some types of TIM work best with a particular CPU?

 

It a shame that you cant but it would be intresting to see if the makers of tim clams are true

 

Why don't we have an open/public poll, All of bit-tech is to vote in this poll, the TIM with the highest votes is obviously the best.
I would set it up myself, but as i only know Arctic Silver 5, it would be a little bias.

Sam

 

problem with polls is that everyone can only choose stuff they have direct experience of, or guess, neither of which are particularly representative of anything other than the WOMM certification (works on my machine)

I've used 4 or 5 different tims, the accuracy of the application makes as much difference as the material, sometimes even more so of you get it badly wrong ( as I have on one occasion).

I wonder, however, the mags bean counters are looking for ways to make money, how about bit tech/cpc staff do their own testing, choose one, and sell it as a rebrand or an endorsed product?

 

Tiz true. If you paintball it on it is baaad.

Mind you in the beginning I tried all sorts of different methods, different amounts and so on. What I learned was so long as you don't use too much the way you put it on makes very little difference.

 

Yeah, it'd be difficult to keep consistency between experiments unless you did something like reapply five times for each paste, with a picture each time of the paste before putting the heat sink on - but that'd quickly drive a man insane.

Truth be told, I've only really seen around 3C difference between normal stuff. Proper cheap budget crap does suck however; on the other hand some of these diamond based pastes are pretty damn good if they're to be believed. I tried to get some diamond stuff myself a while ago but the company who sells it still seems to be setting up.

 

what about an artificial test? we are measuring the thermal transfer abilities of the TIM, it doesnt necessarily have to transfer from a CPU to a sink.

achieving 100% accuracy would still be challenging, but not as much so as with multiple CPUs.

im picturing a row of TECs, hot side up (temps verified to be even with an IR thermometer), a paste applied to each, then something like a square of aluminum (all cut from the same sheet) placed on top. plug in the TECs and watch what happens. i think an IR camera would probably give the most accurate results since im sure the plates/TIM wont heat 100% evenly.

 

Good idea but then you'd also have to apply pressure as that will affect how they perform and with an artificial setup you then bring a whole new bunch of variables into the calculation. I think we're discovering how tricky it may be.

It is funny though, we use them mainly for escaping to a virtual world and yet demand real world testing.

 

Interestingly, TIM-testing used to be bread-and-butter for hardware sites, but is deemed "too difficult" now.

 

Diamond powder.

 

Wise up a bit guys, you shouldn't actually ask a company whether it's worth testing their product - they'll either emplore you to run every kind of test they know they'll pass on it because they're confident or tell you it's untestable and hence you shouldn't even try, if they're worried.

Now, I agree you're not going to get 100% accurate numbers, but frankly Bit-tech staff, you're a modding/hardware/gaming review site, you hardly need to worry about looking poor in a peer review. Making a reasonably accurate review and letting people know the methodology would be far more helpful then saying it's not possible.

Next up, the "build up" of TIM on the CPU should really be pretty minute so long as you're washing the stuff off with some form of alcohol cleaner (which dissolves the TIM easily). Ok, there will be atomic residue on the thing, but everything that hits the air is exposed to passing atoms which sometimes stick to it anyway. If we only care about temperature differences greater than of say +/- 0.5, then this argument doesn't seem valid at all.

As for the stuff about TIM needing to "cure" - sounds like a bunch of hogwash to me, I'd ask them to explain that one to you next time you speak to whoever claims it. If any movement is to be done due to pressure it should be done immediately upon being fitted with the HSF, and the temperatures we're dealing with in CPU world are really far far too low to be affecting most materials.

Finally, if you're really not happy with the idea that it's accurate to re-use the same HSF and CPU (which personally I'd encourage) then do a larger study. Take say 4 TIM's, for each TIM take 5-10 brand new cheap as hell CPUs (all the same type) that you beg from Intel or AMD or some retailer - they can be complete pieces of crap from the 90's for all we care, so long as they gets hot. Now at this stage you could either get 20 motherboards and PSU's etc, or just do the testing round by round over the course of about 2-3 days. You could announce for those who believe in such things that you aren't leaving time for "curing", and just do say 1Hr of P95 per CPU + TIM + HSF.

Average out the results and you're done. Obviously the larger the sample size the better.

Anyway, there's plenty of ways of doing this and just asking the chemists over at companies who make these things whether they're testable and then actually believing their answers seems a poor way to go.

 

Hm. Would the minute build up be a problem? I mean most of us apply and reapply while we tinker about so perhaps performance on chips that aren't brand spanking new would actually be more useful.

What specofdust said is right, we could have results that haven't been 'cured'. Mind you I'm sure, then again, different pastes cure differently, some improving their performance more than others and perhaps later outperforming those that appeared better after the initial non-cured testing.

If you know what I mean...

 

The reason why i brought this up is that all we have to go on what TIM manufacturers claim there products are able to do. For all we know is that they sell us the same product under a new name.

 

It could be old toothpaste for all I know.

 

I've always been a fan of Arctic Silver - would rank it number one along with MX2/3. I've also used the coollabratory liquid metal stuff a couple of times which is really weird but definately works well. 99% of the time though, it's a well applied spot of AS5 that I use