Repaste Job

Hello everyone,

Yesterday I was reading some threads about thermal compounds and I found myself opening my laptop to repaste it. I Have a P150SM-A bought from ProStar in August 14' and recently I upgraded from a 870M to a 970M.

The GPU temps are very good IMO (up till now never exceeded 70 C). Iddle temps floating around 40 C.

The CPU in the other hand was not having good temps, sitting around 50's in iddle and above 80's while playing Dota 2 or benchmarking (3dmark 13, 11...). Usually max temps reaching 86-88 C after a couple matches of Dota.

I have in my arsenal a AS-5 and one tube of MX-4. But after what I saw in the threads, I ordered a GC Extreme, which is coming by Wednesday. Of course I couldn't wait and I tried to repaste with the TIMs I had. After removing the heatsink from the CPU (first time in 10 months after purchase) I saw the factory paste job and I was intrigued.



What is this insulating tape close to the die?



This is the factory paste job. ICD.



I Cleaned the surface with Coffee Filters + Isopropyl Alcohol 99%, although could't remove all the residues from those little resistors I assume?


This was my first try with AS-5. The results were bad, reaching 90+ under 3DMark 13. After some re-tries I came to the pea method with MX-4 and now Im having temps around 80 C on benchmarks and something around 85 C during Dota 2 sessions.

I'll definitely repaste Wednesday when the GC Extreme arrives, so I want some advice on the pasting technique, as I can see as the most important part of the process. Also, if someone can clarify what are those little resistors close to the CPU die I would be pleased!

Thanks in advance,
Gabe.

 

With almost all TIM's you want the interface to be as thin and full as possible. With a large rectangle like that and a thin TIM then an elongated thin stripe of about 50-60% of the width set in the middle should do.

 

There is also a break-in period for the paste which can vary from a few hours to a couple of days. After that period you should have a noticeable drop in degrees with respect to the temperature of your CPU.

Also, as suggested, try to keep the interface as thin as possible. The first picture looks like a mess to me, unfortunately, many technicians overuse the paste, as if the more paste you apply the better effect you will have.

 

Also you want to consider the FAN and vents are clean and the fan is operational. This will also determine your system effective cooling needs failing to notice these two other crucial parts will reduce your system cooling effectiveness. I see where just cleaning FAN and Vents along with new Thermal paste combination will act as if it was brand new from factory thus extending your system lifespan usage.

 

That paper has no real meaning. Mine doesn't have it and I know it isn't necessary.

Also, you cleaned the ICD badly; I can see scratches. You should soak ICD first before removing it, and do not try to force it off with wiping. Otherwise your die will get scratched.

 

Meh, scratches, smatches. Everyone is so concerned about that. It doesn't matter. While it isn't a bad idea to soak ICD before removing, those scratches have little to no bearing on temps. It's primarily cosmetic. I know in an ideal world it's best to have a super smooth finish, but I haven't found it to matter much on laptops unless you're extreme overclocking and need to eek every deg C out of it that you can. From that image, that is so much more clean than a lot of the stuff I've seen and done myself.

I would soak those resistors in alcohol though and soak a toothbrush in alcohol too and scrub that out of there. I wouldn't trust Arctic Silver either if it got on that row of resistors.

Also, from my testing I've done lately, I've found ICD has generally been running 2-5C cooler at load than GC Extreme on both CPU and GPU. I still haven't found a better non conductive, non capacitive TIM.

 

It is supposed to be the best. I must admit I've seen Arctic Ceramique 2 come close to the same temps on my CPU, but I've long since put CLU on my CPU and I don't feel the need to change that anytime soon. The temp difference is just too large.

 

Yeah CLU is quite amazing, just makes me nervous.

 

I would try to spread it out with an applicator as well to insure the thin and even that everyone has been recommending

 

Actually, no, you don't want to do this with Arctic Silver 5. Use a pea or line application and let the pressure from the heat sink spread it out.

AS5 has a long cure time, on the order of 200 hours according to the packaging. If you're using AS5 then apply it and use the computer normally: turn it on, do whatever, turn it off when finished, repeat. Use the computer normally for at least a few days before running stress tests.

FWIW, IC Diamond is about as easy to apply as AS5 but has a much shorter cure time, on the order of hours rather than weeks. ICD works better under the same conditions, too so it's win-win over AS5.

 

Arctic Silver recommends the line method for processors with a heat spreader (desktop) and surface spread method for mobile processors that do not have the heat spreader. OP did not apply AS5 correctly. But yes, there are better pastes than AS5.

 

Thanks for the input, guys. I just applied the GC Extreme and I had some better temps now (83-85 max temp under XTU benchmark). I'll let it cure for a couple hours and test again. @HTWingNut there's no problem in using a toothbrush to scrub the residues from the row of resistors? Also, I think in the first try with AS-5 some of it went to the row of resistors after the heatsink was mounted (I noticed that when I was repasting with GC-E today).


I found a little hard to squeeze the TIM out of the tube, tho. This is the first try (and current) of GC-E. I think the drop was too thick =/ Maybe I'll have to clean the resistors so I'll most likely do another try.

 

I buy bulk pack soft bristle toothbrushes from the dollar store. Just soak it in some alcohol first, pour a little on the CPU PCB and let it sit for a bit and scrub lightly, then soak up with paper towel or whatever.

As far as the bead of TIM, as long as it covers the entire die, you're OK. Only way to know for sure though is to remove the heatsink and check. The TIM bead in the first post image looks like a good example.

 

I'd forgotten that Arctic does suggest the spreader method for mobile processors. Probably because of lower clamping force compared to a desktop cooler.

Yep. Pasting well takes some practice. Don't be afraid to apply, take some notes or pictures, run it for a bit, then remove the heat sink and see what it looks like. Not enough? Try again with a little more. Too much? Try again with a little less. ICD isn't expensive, at most about $7 for a small tube from Amazon that's good for at least 3 or 4 applications. That's plenty to practice with.

Note 1: the difference between ICD 7 carat and 24 carat is the quantity of TIM in the tube. It's the same material regardless of the tube size.

Note 2: ICD is more viscous than AS5 but if you warm it first (put the tube in a cup of warm water for a few minutes) then it flows more easily.

 

So is ICD or GCE better?

 

They're both good compounds. Any performance differences are within reasonable margins of error. ICD is a little easier to apply; GCE is a little easier to remove.

You could also use mayonnaise. Just don't use chocolate. Really:
http://www.hardwaresecrets.com/article/Thermal-Compound-Roundup-January-2012/1468

 

ICD is pretty much the best non-liquid-metal compound (unless you ask Razer, who claim their thermal compound is the best, lawl) if you're chasing every few degrees. But it all depends on the application. Maybe GCE was applied better and hence the temp drop.

 

Every TIM vendor claims that theirs is the best. In practice the difference between best and worst, not counting the liquid metal compounds, is less than 5C. It's usually not worth replacing one good compound with another good compound unless you really need those degrees for high-end overclocking.

Application can matter a great deal. @J.Dre (IIRC) had significant heat problems with his P750 due to poor application of IC Diamond from the vendor. It took him a few tries using different application patterns to get the CPU temperatures under control. He eventually did using an X-shaped pattern.

 

I'm not a vendor, and I'm considering various testing from non-commercial users among various instances when I make my statements.

Next, the "difference" is much more than 5c. It depends on the temps. If your max temp with the worst compound in some liquid cooling loop is 50c, then the best might be like 47c. Because most of the heat is already being distributed. If the highest heat was say... 90c, then the difference can stretch to even 10c. And THEN there's also good and bad applications too. If you're hitting the 90s, you can easily see 5+ degrees between accepted "GOOD" thermal pastes, far less "average" ones or bad ones.

Example: look at this table below, and look at how the difference in pastes increases as the temp potential (clockspeed + voltage) gets higher (the CLU application was not good; it performs better than in that test)

 

Was commenting on the quip about Razer. ALL vendors claim theirs is the best. It's marketing.

With temperatures normalized for ambient and consistent application, yes, the difference between best and worst really is on the order of 5C. Here are the air cooled, low pressure (like you'd typically get with notebooks) results from the Tom's Hardware roundup from 2013:
http://www.tomshardware.com/reviews/thermal-paste-performance-benchmark,3616-19.html
The best other than the Coolaboratory metallic stuff is GC Extreme at 36.7C, the worst is Akasa tape at 42C, a difference of 5.3C. Okay, I was off by three tenths of a degree Celsius.

I'm sure that if you looked hard enough you'd be able to find compounds that are worse than Akasa tape. Chocolate comes to mind.
(Edit: fixed quoting)

 

Marketing is one thing. False marketing aka lying is another.
Something I still think MSI should be sued for for their GT72 and GT80 titan marketing.

 

One thing to keep in mind is that is desktop. I know it is a good indicator, but you also have to look at clamping force of the heatsink. That pressure can be pretty high for a desktop cooler, but it's relatively low with a laptop because they're thinner, more delicate, and have an exposed bare die. Many compounds have a minimum clamp force to be effective. I know as a fact ICD does not. Liquid Ultra does not either. I haven't tried NT-H1 though, may give that a run. I do recall running ceramique on my desktop years ago and it seemed like temps were good at first but then it quickly started getting worse over time.

 

I was simply using it to show temperature variance getting wider the hotter things are, and for no other reason. Many websites etc that test paste will do it with an exceptional cooling system, see ~3 degrees difference, say it doesn't matter and call it a day.

As for Ceramiqué 2, I can personally attest to consistent temperatures over at least 5 months, in addition to pretty decent temperatures on notebook CPUs.

 

Tom's Hardware use Intel/AMD stock coolers with the push pin mountings for their low clamping force tests. They still get variances within a range of about 5C in their bench tests.

 

I'm a big fan of Coollaboratory Liquid Pro/Ultra but it does come with caveats.

Am impressed with Noctua NT-H1 - according to that graph it performs as well as CLP/U in all but two CPU frequencies. I might have to give it a try when I next come to change the TIM.

 

He didn't use CLU properly; it was supposed to give an even better performance gradient. I've personally seen 15 degrees of a drop from AC2/ICD to CLU

 

Ah, I see. I jumped straight from AS-5 to using CLP and have stuck with it since due to consistently good temperatures. Thank you for your reply!

 

Again, though, that is in a relatively low-temperature set-up. Every "real" paste kept temps at sub-40 C, and even toothpaste stayed at 50 C, for crying out loud. The delta between the core temp and room temp isn't big enough for differences in pastes to really become apparent. The difference between good and bad pastes gets more exaggerated as the thermal load increases. If you're dealing with a laptop where even the best normal pastes struggle to keep temperatures at 80 C, then the delta between pastes increases.

 

@D2 Ultima what are the chances of you doing an application guide for CLU?

 

None, because @Mr. Fox already did one:
http://imgur.com/a/JUQvK/#0

 

I've been using Thermalright's CFIII (CF3). I researched a few years ago and still have it lying around. Would it still be good to use now or do these things degrade over time if sealed properly? Might try one of those recommended in this thread like CLU. Cheers guys.

 

Awesome! Thanks buddy.

 

If a notebook struggles to run at 80C then that's either a deficient cooling system, a defective cooling system, or the ambient temperature is too high.

The ideal ambient temperature for computers is around 60F. The ideal ambient temperature for most humans is around 75F. Tom's Hardware and Anandtech and the rest run their tests at around 70F because that's the happy medium between these two ideals. Yes, the higher the ambient temperature, the greater the load on the cooling system. The paste, however, is not the problem. You are. It's your fault for operating your computer in a too-hot environment. Don't blame the paste for being "inferior" when you're running off spec.

If the cooling system is defective then that's a repair job for the vendor.

If the cooling system is deficient then you're stuck with it. "Superior" paste won't make it any less deficient, and there are more effective ways of compensating for inadequate cooling.

 

You guys are talking about load/gaming temperatures? Because my unit is floating around 87C during gaming sessions (BF4, Project CARS, and other demanding games).

 

Yeah, but we can't always plan around just the ideal conditions. Non-ideal conditions can and do happen. The lab I work at is one of the older ones on the university campus, and the ventilation system isn't very good. When it is really hot outside, the temps in the lab are frequently over 75F. I don't have any control over that, but I still have to use my notebook to work with SolidWorks assemblies in those conditions. I don't particularly care if I'm "running off spec", I have work that needs to be done, and I don't always get to choose the conditions I do it in. So if two pastes perform similarly under ideal conditions, but one performs better under non-ideal conditions, then that is absolutely relevant when you talk about the quality of the paste. If they perform within 3C of each other under the best conditions, but Paste B has 10C or 15C higher temps when the ambient temp is elevated, then how is Paste B not inferior to Paste A?

 

My servers are in a climate controlled server room, constant 50F/50%RH ambient. Paste B performs as well as Paste A, costs half as much, is easier to apply and lasts three to five times as long. That's how Paste B is superior to Paste A. Therefor your extreme conditions are entirely irrelevant to discussions about the quality of the paste.

Well, no. That entire paragraph is as much a logical fallacy as yours. Cherry-picking an atypical environment to prove a point is bad research. Maybe Paste A works better than Paste B under elevated ambient temperatures. This does not meant that Paste A is superior to Paste B; it means that Paste A is superior to Paste B under those conditions. If you are not absolutely clear that you are using extreme-ish conditions as a base line the the recommendations you make are at best misleading.

Also, my server room isn't that cold, but I do use Arctic Silver 5 because I got a bulk pack of tubes for free.

Also, disable CPU turbo if the firmware exposes the option. It does wonders for keeping core temps down.

 

I think all he was saying is that a superior paste will show a greater temperature difference from a normal/inferior paste when under load/hot.

Edit: Errr, nevermind

 

Sorry, yes, I should have added the caveat of "under those conditions". But that almost goes without saying - rarely will you find a Product A that is superior to Product B in literally every conceivable aspect. It might have enough advantages that it is better in, say, 90% of cases, but given the fact that Product B somehow still exists, it is probably better in at least certain edge cases. You're almost always better off if you analyze your use case and do your best to identify which product most closely matches your needs.

I was mostly replying to your statement of "not blaming the paste for being inferior", but I over-generalized and should have made it clear that it is inferior in that aspect. Even if Paste B ended up being the overall superior option (lower cost, ease of application, longevity, etc.), it absolutely still is inferior than Paste A in that one aspect. In this case, the performance under non-ideal conditions can be pretty important, especially for laptops where they're not necessarily confined to an area where you have control of the conditions.

 

You say this, but Dark Souls 2 says you're wrong right here:

Spoiler

(I couldn't resist)

 

Well guys now I tried IC Diamond to compare it against the GC-E. The results improved, now I'm getting around 83 C during XTU benchmark and no more than 75 C during Fire Strike. I think overall the ICD is slightly better than GC-E. Now I'm looking for CLU lol

 

Well guys, round 43 of TIM application. This time the "wonder boy" CLU.





The performance is not quite the extreme one as some of the users reported here. Idk if my application was bad, but I had a 8C drop in CineBench and XTU benchmark against ICD. Yeah I know its something but I read something like 15C drops or more... Any hint on that ( @Mr. Fox , @Papusan, @D2 Ultima ; masters of the Liquid Ultra)? Thanks.

 

The

The 15C drop will come from a bad pasting job. The difference for me between AS5 and liquid ultra was around 8C drop with the liquid ultra.

 

Yeah thats what I thought. But I'm loving the CLU. Very stable and the delta-T between cores is really low.