important! p870dm, p770dm, p771dm, p775dm thermal paste replacement

Hello Dear forum members!

recently i have been into customizing new Clevo dm series and here is a very interesting discovery i have found, this covers all gpus: 980, 980m, 970m or any other you might want to install:

if you are like me and not happy with the factory TIM, you might have already taken off the heatsink and applied your favorite TIM to your GPU/GPUs, and if you are like me - then you probably know the ammount of your TIM needed to drop in the center of the die to have it sealed tight.



surpisingly the temp did not drop from the factory TIM setup, and even got worse:
now here is what i saw when i took the heatsink off the GPU:




the reason for that is a very thick thermal pads(2mm) that were designed to have a 0.3-0.4mm TIM layer! :0 too much in my book!

the problem is that if you put 1mm thick pads - the memory chips will not meet the cooling.

next i will follow with step by step installation sequency i have made to achieve good contact. it works for all dm model, as the problem is universal for p870dm,dm1, p775dm,dm1, except maybe p771 and 770 models - these i have not yet seen.

 

ok, here we go:

i took and put away all the orignal thermal pads from the heatsink and the gpu

next i got some 2mm flexible pads and used them for the memory chips in the following way:





squized this one in slightly too from same 2mm pad :

 

All other thermal pads, except ones for mosfets need to be strictly 1mm thick!



for the mosfets on the short side of the card i used 2mm pads and covered them with 1mm on top

 

for these elements i used 1mm thick pads, but had to add thermal paste there, because there was 0.1mm gap:





this small step was very important for me to have all the smallest gaps sealed to it's best, i took this polyethilen(or smth) pad from under some old GPU and stuck it under the 980m in a special place:

 

here is the result, i have used the same drop of thermal paste, actually did not remove it after the first time,

this is what i call good contact!

 

Nice post. Let us know what drops in temperature you get. I'm curious how much a difference the different thermal pads make.

I'm not an expert, but it looks like you might have a little too much paste in your repaste. The amount of paste that has bled over the sides looks a little too much.

 

Great post. Looks fantastic! And everything is making nice contact with the cooler.

But, the proof is in the pudding. Results? Do you have any screen shots of before / after temp differentials? Was it a dramatic change?

 

i actually did that to show that even with excess ammount of TIM, the die does not come in full contact with the heatsink.

i've got around 7C drop on both gpus actually, not with this notebook, as this one i have made only to take pictires and make this small guide.

 

why you need a proof? first of all i did not make it to proove anything, but to help someone who may have stumbled on the same issue. the problem is that if you change the paste and you dont know the issue, you might have the contact as shown on 2 first pictures, and you might not even know about it!

don't you see the difference in the paste ammount/thickness that is need in both cases?

i think for any experienced person the difference is obvious.

as i wrote in the post before, this is not the first DM laptop pad layout i have moded this way.

 

It's fairly common for all in one heatsinks (where the core contact area is a single piece with the VRMs) for the pads to be a bit thicker for the stock application of paste, it's good to give people the heads up.

 

it was not the issue with p770zm, p750zm. there you had the contact out of the box.

here you have GPU chip heatsink part and memory heatsink on the same level - that is a big mess actually.

 

I think I chose the incorrect wording in my post. What I am looking for was how bad the paste job was to begin with. Temp delta of 7 C? While that is pretty large, I was thinking it would've been larger, although we don't have pix of what the pattern of the original spread. Was actually expecting for you to say double digits.

Yes. It is very obvious. So obvious, I think even someone who's never pasted a heat-sink can see the difference.

In sum, nice reporting.

 

I think a lot of people here just like to see that before and after, and are just result driven to try to max performance.

I appreciate you taking the time to post all of that, as it can be valuable to owners of those models. The thermal pads can often be overlooked, so it is a good look at them. Thanks!

 

Yes, much thanks to the OP for this very in depth application guide of thermal pads and thermal paste. Thank you for taking the time to show this as I have yet to do either of these things and can certainly use all the help/guidance I can get.

 

yesterday, after changing the TIM, i have decided to test the machine. before, i have only made this to a single GPU dm models, so to my surprise i have found that Slave GPU has 9-10 C higher temperature. 59-60C on Master vs 67-69 on the Slave!

so the rest of the night i have spent sleepless trying to solve the issue, and when i was about to give up and go to sleep, i have found this! never even looked that was, because i was sure the issue is the heatsink:





a thinn aluminium sheet was covering the Slave GPU fan "breathing in" hole! that's weird, isn't it? i assumed it's some manufacturing mistake, so i cut it out:








the result was amazing! both GPUs went to 60-61C during the Valley test with OC fan profile, and finally i got some good and calm sleep. btw, the OC fan profile is not loud at all at these temps!

 

@tanzmeister can you give me the link where you buy that thermal pad you are using for your gpu?

 

i did not buy them, i got them with a batch of GPUs, i will ask the name for the brand from my supplier. but only mid next week, so if you remind me by PM on wednesday - would be great!

 

nice i will remind you but can you explain again how size your thermal pad using please? my english not well enough so just tell me how thickness your thermal pad using for replace the factory thermal pad (mm)?

 

it is also 2mm pads for the memory, but they are very flexible. in reality you nees something like 1.3-1.5mm, so i have stretched them and pushed them down a bit, after the heatsink installation they tend to raise up and fill the space needed.

for other thermal pads, please refer to the pictures and quotes between them. most of others are 1mm. except for the outer-perimeter mosfet ones and the row of capacitors needs to be look at for the thickness/size of the pads

 

Wow.... I should try this too. Thanks mate!

 

The foil talk reminds me of the @Mr. Fox video - http://forum.notebookreview.com/threads/clevo-p870dm-g.789710/#post-10225400

 

interesting that mr. Fox did not have any foil at all there, only the plastic part.

 

Dude can i ask you 1 thing is what different 5W/mK and 6W/mK on thermal pad?

 

i can't tell you that, i got it from my GPU supplier. any of these will do perfectly, it's just VRAM.

 

It's a rating based on how easily the thermal pads will allow thermal energy to pass through them, however unless the manufacturer specifies the thickness and pressure this occurs at it's mostly a meaningless number to compare between brands.

 

Thanks for posting this up. I am about to repaste so this has come just at the right time.
I have found these pads

http://www.amazon.co.uk/ARCTIC-Thermal-Pad-1-5-Conductivity/dp/B00UYTTMNI

Do you think these would be better than the 2mm then and just double up for the others. (apart from the 1mm ones of course.)

Cheers

Jim

Sent from my SM-N910F using Tapatalk

 

I am confused why didn't you just return it to Sager and tell them to fix it? Now with all the mods you had to make did you void your warranty?

 

should be better than 2mm ones for sure! maybe you even can stretch them if you will need less thickness.

 

i dont think repasting voids warranty, and if you read up, op has the 2nd laptop already with same problem, maybe Sager should look into this too. i already reported the issue to the Clevo factory too.

 

Will any foam do on the back of the GPU. I don't have any spare ones lying about?

Sent from my SM-N910F using Tapatalk

 

Repasting shouldn't void the warranty, as long as you don't damage the computer in the process, I believe.

 

look at you padding, in my case without that foam on the back in a special place, the last raw of pads did not meet the elements: there was something like 0.02-0.04mm gap. so i had to push the gpu up that 0.04mm to the sink by that small foam thing to seal it up properly.

the foam is like 3mm thick and very soft. have tons of these actually, but dont see any reason to send it out, as i am sure you can find a substitute for around your house. just make sure it does'nt bend the PCB, you will need only a slight push.

 

Awesome info in the opening post. +Rep added.

 

We have recently carried out our own internal testing based on the OP and other information in this thread.

We used three different XMG U726 / P870DM chassis with SLI 980M installed. I can't publish our complete findings as unfortunately I don't have enough free time to put them into a presentable format for a forum post, however I have provided a summary below for all to read

- we noticed a similar issue to the OP in that there seems to be less pressure on the corner of the GPU die which is closest to the thermal pads

- the issue on the three chassis (6 GPUs) was much much less pronounced than shown in the OP. Following application of thermal paste, thermal cycles and then removal of the heatsink, on a couple there was a slightly thicker layer of paste in the corner closest to the pads - but on others there was an unmeasurable to no difference.

- we definitely couldn't replicate the extreme difference from corner to corner that @tanzmeister experienced as show in the photos in the OP

Our conclusion is that several factors are at play here.

Firstly it is clear that the thermal plate does not lie completely flat on the die, however in our experience and testing this has not been close to the extent that others have experienced. We can only conclude from this that the problem is definitley not present on all models, in fact it would seem that it is only present on a minority of heatsinks.

Secondly, the chassis and GPUs we tested have the same standard thermal pads as in the OP, therefore the thickness of the thermal pads could be one factor but it can't be the only factor.

The issue is clearly being exacerbated on some units, whist on others there is a very very small discrepancy or none. The other variables which could be at play are the flatness and angle of the GPU die itself and the flatness of the heatsink plate. As always we are in daily communication with Clevo on R&D topics; our report, with added information from this thread (including photos, temperature reports and so on) have all been discussed with them. It would be interesting to see if the more extreme cases are with new heatsinks, ones which have been thorugh heavy and numerous thermal cycles and so on.

Nothing more to report for the time being, I'll update if there is something of interest. At the risk of missunderstanding, it is in no way my intention to dissprove other ifnromation posted in this thread or to deny that there are any issues. It's clear that @tanzmeister spent a lot of time collating the information and photos - we have meerly taken this info and expanded on it by using our own rescources. It's clearly an issue in some units and so requires a resolution, which Clevo are now better placed to understand and deliver on

On the plastic cover mentioned in the OP, this is similar to other chassis. It's a method long used by Clevo to help draw air over other components and through the chassis i.e. cooling the MB areas rather than just pulling air directly from the outside and then through the CPU / GPU HS. It does of course mean that the GPU may run warmer than when the plastic cover is removed, but this in turn compromises the cooling of the MB and other components. Ultimately it's up to you to decide if you want to make this trade, we don't have any data to show whether it has a detrimental effect on the system or if it increases the failure rate of a MB or other components if this plastic is removed. I'm just making the point that the plastic is there for thermal reasons, though you are free to not agree with this reason!

 

personally, based on my experience. there is enough air still to vent the MB, but instead, higher temperature on the CPU and GPU die will result in higher air temperatures inside the case. this is simple physics, so it's absurd.

 

It's not absurd, you just have to measure both methods to know which one is the best.

 

To me this is common sense. If you drive up the CPU and GPU temps by restricted air flow and trap the heat due to less ventilation, and doing so increases overall chassis temps you are robbing Peter to pay Paul.

Removing all of the blockage to ventilation makes more of a difference if you use a cooler pad as I do. If you use it sitting flat on a table or desk then it probably doesn't make a huge difference one way or the other.

 

I tried out a Cooler Master SF-19 pad and it only dropped temps about 1C. Is the U3 more effective?

 

Probably. We don't have any specific feedback, but the general opinion is it works better than single, large fan coolers. If you try it, let us know how it works!

 

With better case airflow the radiated heat inside has less impact as it is near the point of extraction. The machines are designed to operate in all climates after all.

 

Those who experience high PCH temperature should fix their machine on warranty or use a mod U3 with decent 12v fans powered with external 12v power adapter. U3 with this mod is the only choice if you really need a cooling pad!!

 

it does matter even without cooling pad, that thing was blocking the entire 2nd fan uptake area, meaning that the fan had to work really hard to suck the air out from another holes.

and yes, its common sense! you have too cool the hotest parts first in order to control the avarage tempertures better.

to say it is "designed in that way" or "manufaturer knows better" makes me wanna laugh. and that is obvious in the OP.

 

radiation has hardly anything to do with the air flow. guess which of these 2 saturates objects with heat faster? hot air circulation or direct heat radiation.

plain physics here. and efficency is the key.

if something needs extra cooling, there is not need to "fire a cannon at sparrows".

 

but isnt PCH located on the other side of the motherboard in p870dm?

 

Yes, but U3 with proper 12v fans and the underside without these black covers cools the whole motherbord quite good.

 

Agreed, I just like to tinker and I like the satisfaction of fixing things. In my case I feel like I've exhausted all reasonable courses of action to rectify the PCH temps.


I'm going to start a dialog with Sager regarding a replacement. I don't want to start overclocking on a bum board.

 

True, if you unblock the vents then the air is moving slower and will have time to heat up, this is why the vents are channeled so air can be drawn in, over the parts and ejected in a timely manner. This does provide the optimum internal temperature.

 

We can hope Clevo invested enough in engineering the thermals for that to be true, but it seems clear to me they did not give it their best effort. The foil plate on the bottom cover restricts (in fact, it almost blocks) the secondary GPU fan. Other aspects of the design suggests they could have spent a little more effort to get it better than what they did. I'm not saying it is bad, because it is not bad. I never would have purchased one if it had a lousy cooling system.

However, the design doesn't reflect as much thought and effort as one might have hoped for. The most glaring example is the fact that the secondary GPU fan is smaller and so is that radiator. For the single 200W 980 this is not a problem because it receives all of the cooling hardware that a dual GPU system would and it runs very cool. When you split this apart into an SLI system, then it does start to matter. Removing the foil plate that restricts air flow for the secondary GPU fan seems to improve GPU cooling for everyone that does it, so I do not understand the rationale for encouraging P870DM-G owners to leave it as is. If a person wants to leave it as is, that is totally fine. But, I don't see how they are going to actually realize a benefit from doing so.

 

It depends, if you are in a hot environment doing a lot of I/O intesive applications you may find yourself needing that airflow to make sure everything runs ok. That's why I say the machine has to work in all environments. If you want to tweak it for your environment and workload then that's up to you.