Improving P170EM 7970M GPU cooling performance (foiled)

i've done that as it's described in slickdudes guide...
not working. maybe when the enduro fix is available i'm in for a swipe and clean install.

 

Hello,

Just thought I'd post my experience with this mod. Tests are made with Kombustor for 10mn, DX11 GPU burn-in all max.

Before: 92°C and still rising.
Repaste: 89°C max and cycles between 86°C and 89°C due to max fan.
After taping: 82°C, maximum if I let Kombustor run is 84°C.

Bridged the gap between the heatsink and fan, and also sealed the side of the fan... works like a charm thanks very much for the tip.

Edit: ah and my laptop cooler seems to have an impact now, probably due to good air circulation! max temperature with it is 82°C.

 

what paste you using?

 

Coollaboratory Liquid Pro but I simply replaced it. I figured I had not put enough the previous time so I stopped listening to people and put a generous amount on both the GPU die and the heatsink, with the small "gap" between both in mind

I tested in-game temps and they dropped from 88°C max to 75°C max on the same game

 

Sorry about the thread resurrection, but I just wanted to share my experience with the foil mod, the results of which were definitely mixed.

I applied the mod twice, monitoring the temps while running Kombuster before and after applying the foil. I ran two tests with Kombuster: Furry and Tessy test (GL4), and 3D API (DX11) Burn-in, both for 4 minutes with the fans running at maximum speed.

First time applying the mod

Before: Furry and Tessy - 81 C
3D API Burn-in - 94

After: Furry and Tessy - 80
3D API Burn-in - Reached 94 degrees after only 2 minutes, so I stopped the test knowing the temps would only rise and I wasn't comfortable subjecting my 7970m to higher temps

Second time applying the mod

Before:Furry and Tessy - 81
3D API Burn-in - 94
30 minutes of Sleeping Dogs - Max 90, mostly 86-89

After: Furry and Tessy - 78
3D API Burn-in - 91
30 minutes of Sleeping Dogs - Max 88, mostly 81-85

The first application was definitely a bust. I decided to try again after wondering if other variables could have affected the outcome (application method, ambient temps, etc.) As you can see, every test improved after the second application. I'm still a little worried about the heat breaking down the adhesive and damaging something though. To others who have had this mod for a while, how is it holding up? Do you have to reapply it every now and then?

I can't help but feel skeptical about the mod, especially with the varying results, but I'll leave it in for a while and continue to monitor temperature. It does seem to have lowered temps this time though. Thanks to OP for making this thread and everyone who has posted with helpful info and pics so far.

Here are some crappy cell phone pics of the second time I applied the foil

 

It sounds like the setup is very sensitive to minor changes in airflow pathing.

 

It doesn't look like you plugged the hole at the far end of the radiator of the GPU fan (at the interface between Fan and GPU RAM radiator):
http://forum.notebookreview.com/sager-clevo/674825-improving-p170em-7970m-gpu-cooling-performance-foiled-3.html#post8691667

You lose pressure there. Perhaps enough to make a good difference. Otherwise, is sounds like the thermal paste is the problem.

 

You mean the side towards the edge of the laptop, right? Foiling the other side (towards the cpu) won't make a difference, as the fan is already pressed up against the rad and there aren't any gaps. You're right though, I hadn't pressed the foil up against the side of the radiator when I took the picture. I was still trying to figure out how to do it without it looking bad lol. And I've repasted around 3 times, I kept redoing it in slightly different amounts in order to get the best results. My idle temps are in the low 30s and peak temps are normally around 80 in demanding games, but I used Sleeping Dogs for the test because my laptop runs hotter on that game than any other for some reason.

Edit: I made a few adjustments around the side of the fan and radiator, and temps dropped 3-5 degrees across all the tests. I'm really surprised and impressed with how effective this mod has been.

 

I've got a bit to elaborate on this whole tape mod thing.

Spoiler

I've retaped my P170EM a few times just to see what differences in tape application made to temps and I'll just summarize it here. Basically, the best idea is to use one large piece of tape to cover the main top part of the heatsinks instead of multiple small pieces. This helped with my temps. I also ran my nail down the tape pushing it into the small indents in the heatsink so that there wouldn't be as much air trapped between the tape and the heatsink. I'm not sure if it made a difference or not but it helped me achieve best possible temps as is.


I also put a small piece of tape between the two GPU heatsink rails just because of the possibility of air escaping there. The way I did it was to cut a "T" shaped piece with the long part facing downwards and push it into the crevice(make it slightly larger) with a small screwdriver.


As for the other side of the GPU heatsink it was all trial and error. There's quite a large gap in there so what I attempted to do was to get tape all the way to the bottom and around the bottom 'jaw' like structure as to force the air to go through the heatsink without any possible diversions.


As for the CPU I just used one piece of tape that went all the way from one side to the other. I also made sure to cover those small holes on top of the heatsink.

(sorry about that)

 

^ Please use the {spoiler}{/spoiler} tag with "[" and "]" instead of "{" and "}"

 

I wonder how many of you considered doing what I just did. Instead of replacing the thermal pads for the memory, I did what ViciousUSMXC did with his G73JH. I used ICD7, just put large dots on each memory chip. It's working better than thermal pad here. Just a suggestion if anybody else wants to do that.

 

Does that work, isn't the gap too large to just use TIM?

 

With other TIM it might be, but ICD7 is super thick and it works. It's actually been posted by ICD to say it can be used for memory usage. I know Vicious uses it on his memory too and it's been running fine for years now.

 

How do you know that it actually is running cooler? AFAIK there aren't any temperature modules on the vram chips?

 

Because ICD7 is better than thermal pads.

 

that indeed works great, used to do that on my old asus c90s where i built myself a franken-heatsink out of copper plates i ordered on ebay in order to accomodate a 9600m gt. i put several of these plates on top of each other and filled the gaps up with oczfreeze. worked like a charm

Sent from my Galaxy Nexus using Tapatalk 2

 

Well, he doesn't.

IC Diamond requires curing to provide full performance, and it won't cure unless spread thin&compressed. I would suspect it still should be better though, because commonly used rubbery silicone heatpads are pretty shocking in heat conduction. High-grade headpads are very brittle and almost non-compressible.

BTW if you want better alternative, I suggest using thermal putty (instead of thermal grease) - this is better for filling the big gap between heatsink & chips, also it will better support the heatsink. Without good bottom support, heatsink could shift if subject to certain movement - this could detach it from thermal interface.

 

The heatsink shouldn't shift though as its secured quite well with the 3 bolts. It's an interesting concept, I've also been considering something to improve my 680m's memory overclocking, but probably I would just try to get Fujipoly 11 W pads first. (I realize IC7 is something people have on hand though so it's a cool thing to try anyway, pun intended)

Also for curing, it will cure with time anyway, but I wonder if it won't develop air gaps when it does. Depends on the thickness. at 0.7mm or so, it might be fine.

 

The way I've seen it when disassembled mine, bolts in the heatsink made to press it against heat pads, however, without pads to support it, it could potentially "squeeze" thermal paste a bit each time laptop moved not so gently. This could potentially create small gaps between paste blob and the plate.
I doubt it will cure in time, as it pressure-activated - it does not cure inside syringe after all. Or it will take ages.
That's why I said that you need a thermal putty, not thermal grease. Grease consistency is bit like yogurt, while putty is like play-dough.

 

ICD7 works but I found that the gap between the DRAM and the memory heatsink is as large as 1.0~1.5mm, I said that was because the 1.5mm Phobya Thermal pads I placed were still able to move (They aren't sticky) when I push them with a screwdriver. I had to use 2mm pads to make sure it stays tight.

 

That's pretty bad then, the pads from factory are 1mm. I think some of the Clevo heatsinks are bent or manufactured incorrectly to have such a big gap. I think i saw a slight gap problem on my heatsink as well for the vram chips closest to exhaust/back of laptop.

 

I did a thorough inspection on mine and the gaps seems a bit less than 1mm and those foil tape mods don't make sense to me as there are no gaps between the HSF and HS fins itself? maybe some heat sinks are just bent out of the factory???

 

well "gap" is a big word here. minute distances of one mm or less would cause warm/hot air to escape the heatsink back into the laptop. the foil just forms a closed up "tube" so to say, through which the hot air is funneled directly outside the laptop. this way it cant recirculate back into the machine and heat it up more

all in all, that bit of foil really does make a difference, translating into 5C lower max. temps for both GPU and CPU in my case, for example

 

you would need some good quality aluminium tape then to withstand constant 80c+ temperatures. i guess i could try again as I did not see much difference the first time i tried it, the difference was pretty non-existant in my first test -1c/+1c

 

well the 80C+ temps are just directly on the gpu core, the air being blown out of the machine isnt nearly as hot. id say around 50C but not much more than that.

 

Touche, good sir.

Will try again when I get some more tape, haha.

 

You have to also monitor your fan speed, remember that you have active/adaptive cooling. At the least you will be a step lower on the fan speed which is quite a large difference.

 

Ahh Good point, I'm such a dope, fan profiles... completely forgot..

 

Try speed tape that is what I used, worked a treat

 

There is a slight problem with this that I overlooked before. Using alum foil tape translates the heat into the GPU memory heatsink meaning it /might/ be heating the gpu ram sodimms. I don't know by how much but it is something to take into consideration. In this regard speedtape might work better.

 

what's speedtape?

 

I think a kind of electrical tape, I would use electrical tape just for the sake of safety.

 

Hello, getting ready to do this on my np9170. Can i use this tape?

2INX10YD ALUMINUM FOIL TAPE - Newegg.com

If not, could someone recommend one that is available on newegg in the us? ordering something there today and want to pick it up if i can. Thanks all!

 

Still trying to decide what tape would work best - aluminum or copper. Any suggestions? Can be from anyplace that ships to the US

 

IIRC there isnt much difference between the two (some nbr user has done a comparison already, but forgot when and where), main thing is that the tape forms a closed-up duct through which the hot air can flow outside without part of it fleeing back into the machine through those gaps.

 

@jaybee

So the aluminum tape i listed a few posts up should be ok? Assuming so, my next question is rather blunt : How can I screw this up? Any possible way for me to short the system (cant see how) etc?

 

wrong thread.

 

yup, should be fine. i myself just got the cheapest aluminum tape i could find on ebay and it works as it should

heres what it looks like in my machine (check the gpu and cpu heatsink fins taped to the respective fans):



cant really do much wrong there ull do fine, no worries. just make sure that it sticks well without and air bubbles trapped or something like that.

 

@jaybee

Did you add what looks to me like extra copper shims to the heatplates on the CPU and GPU?

 

we shall see if this works when I get my 7970M from Kolias - I shall do a mini comparison (p150hm)

 

haha nice, good old kolias was "side-grading" to a 680m, i remember him telling me about it im sure ull be more than happy with that new beastly card

@ajfountains: yup i did. doesnt lower load temps per se but prolonges the time it takes to reach those max. temps at load definitely worth it for those few bucks these copper shims cost.

 

@jaybee

Interesting. Did you have the mod the back cover at all? Any problems with space? I imagine I have a tad more room in my NP9170.

Where did you get the shims? How did you get them to stick (thermal compound?)

 

no modding necessary on the back cover, since those shims are very low profile. they already came with an adhesive tape on the bottom so u pretty much just take them out and stick them on easy-peasy

dont really remember where i got them tbh... found some examples on newegg and ebay tho:

Copper Heatsink - Suitable for Raspberry Pi | eBay

http://www.ebay.de/itm/Copper-DDR-D...et_Cooling&hash=item4ab96144fa#ht_1940wt_1163

Enzotech BCC9 VGA Cooler - Newegg.com

Enzotech BMR-C1 VGA Cooler - Newegg.com

Enzotech MOS-C10 Memory Cooling - Newegg.com

the first two are closest to the ones i have, concerning their dimensions. the other three have longer pins, that would probably improve their cooling capacity but u better check first how much space u have between ur heatsinks and back cover plate.

 

Awesome, thanks. I'll look into getting a few. Ive noticed that core 2 in my 3630 is about 3-5C hotter than the other cores, so i think a repaste is in order. Will try to do everything at once.

 

3-5C is negligible really, so that alone wouldnt warrant a repaste if u ask me temp differences between the cores are normal as long as they stay in the range of below 10C. no matter how perfect ur paste job is, ull probably never get the cores to sport identical temps

 

Good to know. I haven't noticed any performance lag and I know that HWinfo temps arent always exactly accurate. The highest i have seen it hit is 80C on max with the other cores around 75-77C. I think I will still repaste as I have never done it before. Having such an easy access laptop has really inspired me to start tinkering and learning more. With a repaste, all i can screw up is to put too much or too little on, correct?

 

Don't worry about it. One of my cores is a good 8C lower than the others(32C now while the other 3 are 40). Under load it's the same(65 while the others hit 70).

 

thats correct. if ud like some visuals on a paste job, i can recommend the 2960xm upgeade thread linked in my sig. it includes pictures of the cpu swap and also how the paste job was done

Sent from my Galaxy Nexus using Tapatalk 2

 

Something just cam eto mind - Do you need to reapply the foil everytime you want to take out the heatsink to clean it? Or to switch out GPU/CPU as well, correct?

 

yup thats correct.and i wouldnt reuse the foil since the adhesive is only good for one application. once u get the hang of it tho it doesnt take longer than a minute or so to apply

Sent from my Galaxy Nexus using Tapatalk 2