M6700 Heatsink swap/repaste

There is that stressful moment when you don't know what your temps will be during reassembly and before booting... Always makes me sweat.

Personally, I prefer using a little too much paste than not enough and getting the amount just right can be tricky.

 

My heatsink just arrived. The box was rather small, just enough to hold it, so not huge like yours. Pre-applied paste looks good. I'm going to try to stick it in tonight, I'll include a few pics and temps and we'll see what happens!

 

This is a useful guide - I wish I had it when I repasted my heatsink. The process wasn't all that painful. I made a couple mistakes that I will not elaborate upon , but nothing that resulted in damage and now that it is done I am much more confident in working on it. I think the most important thing is to sort the the screws by step (this way you will not forget any screws during reassembly) and count the screws that you remove to make sure you have them all (the top palmrest actually had two fewer screws than were indicated in the manual).

I decided to repaste my heatsink because I was hitting the thermal limit (105C) on two of the cores while running a stress test. Also, core0 was running about 10% cooler that the hottest running core (core0 did not seem to exceed 90C). I was partially expecting the paste to be half off the CPU (causing dissimilar core temps), but the the paste looked evenly applied and making good contact. I cleaned the CPU and heatsink and applied IC Diamond thermal compound. My results after repasting are somewhat of a mixed bag. I was a little disappointed to see that core0 is still running about 10% cooler than core2/3 (core1 runs a little cooler than the hottest cores). However, the "hot cores" are no longer hitting the thermal limit. They seem to max out around 92C (still a little high for my tastes considering the ambient temp is 17C) while the cool core maxes out around 83C. I am still a little concerned that the cores have such dissimilar temps (when evenly loaded). I would be a happy camper if all of the cores behaved as well as core0...

 

It is normal to expect some differences in the core temperatures, it is extremely hard to get even contact on the whole of the heatsink surface. The way you screw it in place has an impact, that is why I screw each screw a little in the order indicated and repeat the process multiple times, to get the contact to be as even as possible. The quality of the heatsink surface also matters so for example, you could lap the heatsink surface to try to get better contact, but that is an extremely delicate and time consuming operation, not to mention the potential to screw up.

It is also worth having a temperature graph, I notices that the temps tend to spike to higher temps for half a second once in a while, so the max temp value can sometimes be misleading. That is one thing Intel XTU or HWiNFO64 are good at providing. Throttlestop allows you to log your temps so you can make your own graphs too should you want to.

If anyone wonders about also repasting the GPU, once the CPU heatsink is removed, you can remove the GPU heatsink at that time so it isn't much more complicated than repasting the CPU. The GPUs in the M6700 usually run pretty cool though so I don't see a reason to repaste.

 

Grumpy,

If it's any consolation, mine has similar characteristics for heat spread, Core0 is coolest and Core3 warmest. My first thoughts are that it has to do with airflow, just a guess.
I was quite surprised that putting the power plan to ultra-performance helped. I was thinking that XTU would overide that but apparently not.
It had improved by 8F from that change alone.

 

I did exactly that

I considered this, but I did not have all the materials required at hand and the potential for screwing something up seemed greater than the potential benefit.

While I am not graphing/logging the temps I am also not relying on the "Max Temp" values. While under the stress test I look at the temps that the cores seem to "settle in to".

I didn't not bother with a GPU repaste. My GPU runs so cool - It does not even seem to break a sweat when overclocked.

That is somewhat reassuring

I too am using the ultra-performance power plan. It was the only way that I could get the fans to stay on while docked, which results in cooler temps.

 

Alright, I've seen people screw each completely in the order instead of doing what we did, same for logging the temps. I agree with you about the lapping part, it's an option, but not something I would personally do.

As for the core temps, I have a difference of about 4 degrees between the hottest and coolest cores on my M6700 when stress testing, even my desktop has non uniform core temperatures so it is in fact pretty normal. 10 Celcius is a lot though, but you can't expect uniform temperature across all cores. You could always try a new heatsink though.

 

I don't expect cores to be identical, but the 10C difference does have me somewhat concerned although I am not sure that there is really anything I can do about it, especially since repasting gave similar results (albeit lower overall max temps). It seems that it must be either a "defect" with the heatsink or possibly even with the CPU. Well, for now, it is not hitting the thermal limit which is good.

 

Yeah, personally, I consider that as long as you're a good 10 to 15 Celsius under the thermal limit, you're safe enough.

Aaron, if possible, could you count the number of screws you have to remove if you have the fingerprint reader and smartcard reader. I'm curious as to whether there are some extra screws.

 

I just finished the procedure. I'll write more later once I get some testing done but I'll comment on a couple of things here quick.

I didn't have any extra screws for the fingerprint/smartcard reader. Just a cable that connects on the underside of the laptop, which is accessible when you remove the secondary drive bracket. The Dell manual shows right where it is in the section on removing the palm rest. Everything else went just like you said in your original post. I counted the screws very carefully. (Thanks a lot for it, btw, it helped out.)

Also, on the issue of uneven core temps. Both before and after the upgrade I am having core 0 at about 5 degrees hotter than the rest. (This is idle temps. With the old heatsink, it evened out somewhat when under load --- I haven't tested under load with the new heatsink yet, though.) I'm not super worried about it, there have been plenty of posts on the issue here in the past and it is normal.

More later once I have some numbers.