Repaste done, BIG gains.

So, I repasted my CPU and both GPUs today with MX-4.

First off, I'm thankful that I bought ArctiClean...the Dell paste crap was a nightmare to get off. Would have been far worse without the ArctiClean. As for their paste...no wonder my CPU was getting so hot, it looked like there were areas of the CPU where there was no visible paste at all. I know it's meant to be thin, but it looked WAY too thin to be making proper contact.

When repasting, I decided to spread the paste manually with a card rather than placing a bead in the middle and spreading it with the heatsink. As it's my first time using thermal paste, I didn't want to use too much or too little paste for the latter method. Yes, it probably has introduced air bubbles, but I don't mind taking the ~1C hit to make sure the paste is spread evenly in the first place. My plan is that this application will last for the remainder of the laptop's lifetime. (Besides, I remember reading an article online - can't find it now sadly - that compared many methods of applying thermal paste, and found that ALL of them had air bubbles, and the temperature differences between most of them, including the manual spreading method, were inside of the margin of error for the temperature tests.)

Other than the paste being a pain to clean, it was also damn near impossible to get the heatsinks off the GPUs in the first place. The thermal pads had pretty much adhered the heatsink to the card...I had to remove each card and gently pry the heatsink off.

Had a big panic when I reassembled it...8 beep error code, wouldn't POST, thought I'd managed to break a GPU in the process. Was kicking myself for even trying this. Just in case, I disassembled it, removed the GPUs, wiped the contacts in case anything had got onto them, and reseated them. Well, whatever the problem is, it was fixed in the second assembly. Works perfectly now.

And wow, the gains are massive. Running at a 3.6GHz 4 core turbo, as I was before pasting...previously, when running OCCT, the fan would kick in almost immediately, and the CPU was touching 100C within a minute. Now...it takes a minute before the temperature even goes high enough for the fan to spin up to its higher speeds, and it seems to be peaking at 85C. Guess I'll try 4.0GHz later...

The GPUs seem to be running cooler, too. I didn't measure their temps before repaste, but after, the primary GPU is peaking at 80C in Furmark. The fan is also running quieter than it did before under 100% load. (No idea what the secondary GPU is running at...nothing can seem to report its temperature these days, perhaps the temperature sensor died?) Idle temperatures seem to be down too, the CPU would previously idle around 50-55C, now it's around 45-50C.

Now I'm glad I did it, even though it took me 3 hours in total to replace the paste and fix the error code...the last hour of which was very stressed, worrying that I'd just broken a 3 grand laptop.

 

That's great EvilCorsaiR. Now enjoy your renewed machine :thumbsup:

 

I hit no higher then 85 degree's on my GPU's (680m's on the R2) since I respasted with TX-4), and 98-99 degree's on the 3920xm @ 4.6Ghz at normal usage/loads. about 75% was max I loaded it too...

but I was hitting 100 degrees on everything running a game before, when clocked at 4.3ghz before repasting, when using stock. Much quieter now in my house, no more sounds of aircrafts taking off.

 

MX-4 is great paste overall not so much suitable for heavy OC but it runs very well I know from my own past experience. However, I have tried almost them all and one I havent tried yet (yet waiting for a delivery) is CoolLabs LiquidUltra and LiquidPro and spare pair of CPU heatsinks. I have not seen better thermal conductivity of any thermal compounds yet so far - LiquidUltra rated @ 38.4W/mK and LiquidPro rated @ 32.6W/mK. However, need special application method and some serious precautions, but I would recommend you to put this name on your future to-do list and try this when you get the itchy feeling that " -I could get better temps and results/quieter machine"

 

Well, the two big reasons I went for MX-4 are that it's non-conductive (first time I've ever applied a thermal compound and didn't want to fry something if I made a mess) and a single application will supposedly last 8 years. I didn't want to have to repaste it again, I wanted the application to last for the remainder of the laptop's life. Although now that I've done it once, I wouldn't be too uncomfortable doing it again...still not something I want to do anytime soon.

The conductivity ratings for LiquidUltra and LiquidPro are tempting, but MX-4 is good enough for me. I don't plan on overclocking the GPUs, and the 3720QM is locked to a maximum of 4.0GHz on 4 cores (still need to try that out, currently at 3.6GHz). I'm hoping it won't push beyond 95C at 4.0GHz. Still, a cheap ES 3920XM/3940XM from eBay pushed to 4.5+GHz is tempting...

It certainly is a lot quieter now, though. It would be quieter still at idle if the secondary GPU didn't keep getting stuck at 3D clocks...it keeps the secondary GPU fan spun up to audible levels.

 

This is worrying. The CPU thermal performance has very rapidly and severely degraded. Now it hits 100C almost immediately after starting OCCT (it was peaking at 85C after 10 minutes of OCCT, immediately after the repaste). Same room, same setup, same BIOS settings (still at 3.6GHz).

I could try cleaning out the heatsink, but I did so very thoroughly while repasting, and there's no way dirt could build up that fast to degrade thermals that quickly.

The GPUs are still around the same thermal performance as they were immediately after repasting - or, at least, if they've degraded then it hasn't been enough for me to notice a significant difference in thermals.

Any ideas on what could have gone wrong here? Will I need to repaste again to fix it, and is there anything that I should do differently this time?

 

thats what you get from a mediocore TIM, you would need to go something like LiquidUltra to sustain low temps at no matter how crazy your OC may be

Im on my ~20h run now with this new which-craft TIM, have never crossed 72c while heavy multitasking in desktop work environments, multimedia, and internet all at the same time while @ 4.4GHz simple ThrottleStop OC no VID increase, so how do you think you like this?

You can follow my application to be extra safe while using this TIM, link here: http://forum.notebookreview.com/alienware-18-m18x/741745-repasting-m18xr2-coolaboratory-liquid-ultra-any-tips-before-i-start-28.html#post9561417

no thermal pastes will last as long or be as reliable unless its a ICD or LiquidUltra, thats what I'd count on from now on from my personal experience. Most TIM's will fail shortly within hours if not days after some abusive OC, they just tend to loose their formula properties rather quick rendering useless.

Remember - with great power comes great responsibility, so chose wisely what you treat your hardcore hardware with

 

MX-4 isn't a mediocre TIM though? Every test and review that I looked at for it was very good, and while very heavy overclocking and heavy use might degrade it...I've only been running at 3.6GHz, and aside from a few minutes of OCCT and a few hours of moderate gaming (nowhere near 100% load), I haven't been hammering it. That shouldn't degrade even the worst of TIMs, and also wouldn't explain why my GPUs haven't degraded much in thermal performance, if at all.

I was just wondering if there was some alternate explanation, such as maybe under-tightening or over-tightening the screws on the CPU heatsink, or too thin/thick an application, or...something. Defective paste did come to mind, but again, I'm not having problems with the GPUs (not yet at least).

Furthermore, I absolutely require that any TIM I use is non-conductive, and - presumably - LiquidUltra is conductive, given that it's metal. I will not risk damaging my system by getting conductive paste on any parts of it, no matter how careful I am or how low the risk is. I would have been willing to try it if it was non-conductive. (And I still can't believe that MX-4 is a low quality paste that would degrade after a week of moderate use, particularly given that it's meant to last 8 years...)

 

Any idea on the Prolimatech's? they have been gaining some traction lately

 

MX-4 has never lasted me long on CPU's no matter what CPU's, for sme reasons on GPU's its ok, but again I had a bad habbit of repasting every month or so on everything so that much I can tell you on MX4. I would only use MX4 on tablets, netbooks, and some i3-i5 cpus maybe nothing more heavyweight. MX4 had ~5-10c higher temp difference vs PhobyaHegrease Extreme on my past experience and extra ~5c drop achieved after applied ICD TIM. So here you go (testbed application was i7-940xm CPU and 7970m) I also ended up using ICD on both prior I got my machine in to new hands. At least you know there is the best known non conductive - ICD thermal compound, give this a try

 

I would have went for IC Diamond in the first place, I was just worried about it scratching the CPU and heatsink (as it's known to do). Should I give it a go?

 

scratching is not so much caused by the TIM as by the user who may not be careful and apply in improper manner or attempt to dry-wipe it when replacing it with fresh one. EVEN IF YOU SCRATCH you die, its so microscopic it will only be visible in person/camera but not that you will feel any performance degradation if any at all! Many people used it many people still do and its one of the top best TIMs before LiquidUltra, so you can give it a go, follow ICD manual from their page and you will not end up scratching anything. Worst mistake you can do besides dry-wiping is - when applying this TIM try not to drag the syringe across the die as this cause little scratch, like I have done, blame the time of the night tiredness and just did what I did, but in no way it caused any harm to my CPU even then

 

It's just...I can't believe that MX-4 would have broken down that fast. It was under a week before temperatures had risen back to their pre-paste state. I'm fairly sure that it just wouldn't break down that fast on a proper application?

Every test I found online showed that MX-4 was very close to IC Diamond in temps, and I'd still be worried about the abrasion of IC Diamond if it was a bad batch with larger particles (I've seen forum threads about that online). If the CPU had a heatspreader I might be less hesitant, but with an exposed die...

MX-4 is meant to be one of the best pastes on the market, from all of the reviews I'm read, and it's supposed to be durable as well. That's why I'm thinking that I might just have applied it wrong. Perhaps I'll try putting on a rice grain-sized piece on to the CPU die and letting the heatsink pressure spread it, instead of spreading it by hand before applying...

You might wear it out quickly, but keep in mind that you also clock your CPUs far higher than I have mine.

 

well yano what, why not give it another shot? Since you've some serious ammount of mx4 left do few more series of repaste and log your usage and see how it will perform, maybe you've had just bad application this first time. Try and let us all know. If it still fails like before, means time to order some ICD or other compound like GC GELID EXTREME or also known as Phobya Hegrease Extreme.

BR

 

Yeah, I'll give it another shot when I have a couple of hours to spare for a repaste. This time I'll try a small line in the middle and I'll let the heatsink spread it. Any idea how thick and long such a line should be for our mobile CPU dies?

Is there any way to tell exactly how tight I should have the heatsink screwed on? I obviously don't want to overtighten it, but I also don't want it to be too loose to make proper thermal contact.

 

I used ICD on my R3 and my temps never peak beyond 72*C for the CPU and 70*C for the GPU. Stuff is crazy thick though, gotta be really really patient with it when applying.

 

line as thick as it comes out of the tube, dont stretch it and dont bulldoze it to itself trying to make it thicker. The distance doesnt matter too much as long as its not longer than 2-3mm from each side of the die (length wise).

Make sure you catch all four screws in PCB before you tighten them all the way down to the bottom. Only tighten finger tight (ONCE PER EACH SCREW) in exact sequence numbered on the heatsink and you will be fine. Tightening more than once may cause air gaps to develop etc. Finger tight by means imagine it stopped and if you turn further imagine its a wax candle and it will break, so avoid breaking wax candle. These screws are spring-loaded - they do pretty good job pushing heatsink down appropriately

 

I have nothing but good experiences with the Gelid Cooling Xtreme thermal paste. That's the best stuff to buy in bulk. Really, it's awesome stuff. Has performance that draws close to IC Diamond but for about half the price and twice the amount.

 

Theres nothing wrong with MX-4 - i've been using it for years and it doesnt degrade providing you apply it properly. Do not spread it just put a drop on and then use the pressure of the heatsink to spread.
You might have a faulty batch.

The paste makes very little difference if it's good paste and applied correctly anyway - 80-Way Thermal Interface Test

 

I know, I read a lot of tests like that - basically from what I could gather, all of the high end pastes are very similar and the application method is more important.

I'd also read another test that shows very little difference between the various application methods themselves, provided each of them is done right. Maybe I didn't spread it thick enough, or something like that. I'll try a drop this time, and if it degrades in a week again, I'll contact the seller and report it as a faulty item - it definitely shouldn't be doing this.

 

Repaste done, currently stress testing. Took it over two minutes of 100% load before the fan even kicked in to the higher settings, at which point it was in the high 80s. Under sustained 100% load it seems to be settling in the low-to-mid 80s on all cores (except Core 0, which oddly always seems to be about 10C lower than the rest?) This is at 3.6GHz on all 4 cores, no voltage increase.

Used the line method for this application. I suspect that I've used far more paste than I needed to, but I wanted to make sure that I had complete coverage and I don't mind sacrificing a couple of degrees to make sure it won't degrade and there aren't many gaps. (This isn't like a desktop CPU, of course, where it's okay if there are gaps at the edge of the heatspreader that aren't in contact...we need to make sure the entire exposed CPU die is covered.

I'll report back on whether it seems to degrade as fast this time. I was glad that MX-4 is so easy to clean, though. All I had to do was wipe it off, no need for the ArctiClean 1 to be able to wipe most of it off.

I'm still wondering how the hell people manage to get temperatures in the 70s with a higher overclock. I've seen those sorts of temperatures from some people.