If you are sure that it is a hardware issue, the manufacturer won't help, it is out of warranty and you have some skill, then you could open it. I would take a camera to record each step, because putting it back together is gonna be hard. You may also want to find a tear-down video / howto before you start.

If there is a large space between the heatsink and the CPU, maybe a shim would be a good idea. Aside from that (hopefully) rare situation, shims will make thinsg worse. But with the 'race to the bottom', 'cheap as the manufacturer can get away with' laptops, it might be far more common than in better laptops or desktops.

IMO the cause of overheating in this laptop is the 'switchable graphics' setup..the Radeon HD 6470M is probably just drawing power and creating heat pointlessly.

Underclocking the CPU, gah. By aprox 50% as well, thats _really_ going to hurt performance. It might help overheating when compiling, but that is ignoring the real problem...and its quite likely the laptop will still overheat.

Just FYI, that was 'bumpgate', and not just limited to mobile GPUs, it also affected desktop GPUs. They never totally fixed the problems with G7X, G8X and G9X.

Also, NVS 135 is G84M, and basicly the same chip as used in the 8400M G/GS/GT.

Does it mean I need to compile the 3.13 kernel myself? I don't have experience doing that. Isn't it hard to do? :(

Yes, that's what I thought too. I think I'll try to install the 2nd newest propietary driver first. Btw what do the commands do?

OP go from general to specific. Either narrow down the cause (best) or eliminate all possible causes (weak, but workable)

I am unfamiliar with this runaway problem with switchable graphics so all I can say is either meter this to be certain this is happening or if there is a simple software fix, try it and see if the temps drop. Software problems should be addressed first especially if it is not in a "from the factory" condition ie - you installed a new system. Fix this first because it is easier and faster to see results.

It can also be worthwhile just to remove the battery if you expect a long term on AC power, or as OP mentioned, during a particularly intense task on AC.


If, OTOH, it is hardware related that isn't as hard to fix as many think. Regarding the above, I emphatically disagree. There can be NO SPACE between a heat sink and the heat source because air is an insulator. I have never seen nor can I even imagine how a proper shim to take up an air gap could ever possibly "make things worse". There are even videos on YouTube showing how to do this. It is that common, and as I said it isn't beyond the scope of even moderate skills.

In addition to making certain there is good heat transfer between the CPU and it's heatsink (and also the GPU and it's, if it has one) one should look to see if there is any obvious spot to improve airflow. Just don't drill any holes, especially in metal, without the means to insure complete and utter clean up afterwards. A Nibbler on plastic works wonders and a thin sheet of packing foam can act as a filter and cosmetic cover if you decide to add what amounts to a new air intake.

If you have a good reason not to modify your hardware, even fear, then just don't, but for most people with any decent tools and skills to use them, the improvement can be absolutely dramatic. I have seen 20 C reductions at idle, and far greater resistance to change under load. Just more stable.

Other than for testing I wholeheartedly agree with the above. Castrating your CPU should not be a viable solution when other means exist.

Interesting to note that I have a D630, two D620's and a couple C series. Each one has gotten too hot and I've start to disassemble and replace the paste and clean the fan when I get them, all bought used. My belief is that Windows never reaches the full potential. The D620/630 have the gap between the fan and the fins on the heat pipe that fills with hair, dust, etc.

check this out http://www.linuxquestions.org/questi...8/#post4830372

Actually, it is pretty easy. AlienBob has a nice HowTo for that (of course you have to adapt the kernel version): http://alien.slackbook.org/dokuwiki/...kernelbuilding

Umm...'air gap' is a bit of an overstatement. Though it can happen, as some heatsinks are slightly curved, cheap ones in particular. In those cases while there will be contact between the CPU and theheatsink, its not full contact, and there can be 'air gaps' at the sides of the CPU/heatsink interface.

Its also possible that there will not be enough pressure on the heatsink, and in those cases adding a shim can help.

Adding a shim in most (all?) other cases just adds more layers and more thermal paste, which reduces the efficiency of the cooling system.

Its not really runaway. The issue is that with switcahble graphics setups, the 'unused' video adapter is generally not 'switched off' with linux, which just sucks power and creates heat.

Its fixable in most Intel CPU/AMD GPU or AMD CPU/AMD GPU situations with vgaswitcharoo or by using the fglrx closed driver. (Intel CPU/nVidia GPU and AMD CPU/nVidia GPU setups are now almost always 'optimus' and work better with bumblebee).

Just some comments for clarity

I do know enough about switcheroo, bumblebee(d) et al that I know it isn't really a runaway. I'm afraid that choice of words just reflects a near obsession with thermals. To me 65 C idle is just insanely hot and completely unusable in that condition, requiring an immediate and if needed, drastic fix.

I am fully aware and extremely experienced with heatsinks and thermal removal systems of many varieties including passive, water, oil, and phase change as well as electrical (ie- Peltier). "Air Gap" is not an overstatement for this reason as I am not referring to the minute differences in surface flatness/smoothness. That requires polishing, which I commonly do on PCs but is of little value on a laptop due to most having CPUs soldered in place, making the risk not worth the gain.

The occurrence of an actual and substantial air gap on many laptops is due to factory assembly methods. Parts that fit together precisely are generally anathema to assembly line speed, so manufacturers often resort to extremely thick wafers of "thermal pads". This is at best, "fooling oneself" since even the best thermal paste does not have a coefficient even approaching metal to metal contact. It's just better than air. Thermal paste should be applied in as thin a layer as possible. Ideally much of the mating surfaces should NOT be separated by paste so that it only fills the low spots preventing air pockets.

It is not uncommon to see thermal wafers used that are 1/16th inches thick (0.0625) and I have witnessed some closer to 1/8th inch (0.125). It is only a very slight improvement to replace this entire thickness with the same thickness of even high quality thermal paste, the primary benefit being simply eliminating the adhesive layers on both sides. Not the best improvement, but some.

Consider this: A person could coat his finger tip with a thick layer of thermal paste and apply a match to it with very little danger of a burn. DO NOT try that with a copper shim with a thin layer on both sides. Does that make it more real for you?.

A less dramatic test that is safer to try is this. Put a blob of thermal paste, or better one of those ridiculous wafers, on a nonflammable surface and hold a match or a lighter to it for say, 5 full seconds. The blob, if the mass approaches anywhere near the amount actually used on many laptop heatsinks, will not even be uncomfortably warm in 5 seconds. A wafer can get uncomfortable but not deliver a burn even though the adhesive will ignite and burn for a few seconds too, and in direct intimate contact with the paste. Again, DO NOT try this with a shim of similar mass with a thin layer of paste.

FTR, I have done both of these tests, but only after I employed a VOM with a thermal transducer probe to measure actual temps. (I may be crazy but I'm not an idiot :P ) Heat in electronics is the enemy and I take it very seriously, and defeat it with extreme prejudice and knowledge gained from many years of work in the field.

You do what you want, but I will never own for long any electronics (other than Vacuum Tube) that will idle at 65 C.

I wouldnt call imperfections in the heatsink an 'air gap'. If the heatsink and/or CPU heatspreader is not flat, you can get what I'd call an air gap.

BTW, its normally called 'lapping', not 'polishing'.

Thermal pads can and do compress. Sure, if there is flatness issues you will get thick bits...and if a manufacturer is using a really badly fitted and/or designed heatsink you can get some quite thick bits of thermal pad.

But that is bad design and bad manufacturing.

I seen lots of people suggest shims over the years, and sometimes I've seen people even say 'adding a shim will help in all situations becasue copper conducts heat better than aluminium'. Thats a bit silly, as you can get pure copper and copper bottomed heatsinks. But even with an aluminium heatsink, its still very uncommon to see temps go down, they mostly go up.

If there is a normal ('good' I spose) heatsink/CPU interface, adding a shim will just make things worse. The more thickness and layers you have, the harder it is for the heat to get out. That is why you can remove the intergrated heatspeader found in pretty much all current CPUs and improve temps. No, I am not suggesting that is a good idea for any normal user)

Its not going to be as hot as just a thin layer of thermal paste. ;)

Either Virginia isnt that warm, or you have aircon, or both. ;)

I ran many systems with 40C+ room temps during summer for years. At 40C+, idle temps tend to the mid to high 50s at minimum. I've had idle temps go as high as 60C+ at times. No, its not a good idea.

IMO its software related, and should be fixable with vgaswitchroo, and/or possibly other methods (e.g. using a 3.13 kernel as suggested by TobiSGD).

I'd be trying to reinstall windows asw a temp test and/or install vgaswitcharoo, a new kernel etc. with linux before modding hardware.

Ummm neither did I. I thought I made it clear that the gap comes from removing an overly thick wafer that was used by the manufacturer to keep assembly costs down. To me, 1/16 inch is far too large a gap to be spanned by thermal compound.

Of course I am aware of the hobbyist terminology. Assume I just didn't want any beginners trying to lick their CPU. :P
<snip>
Those people are ignorant of the Science, and if their temps go up, they are also inept at the mechanics.

Why even mention this? The specific comparison was between a thick layer of compound/paste vs/ 2 thin layers and a metal shim. There is no need to twist or invent things upon which to disagree.


In the Summer, Virginia can have occasional days of 40 C but 32 C is rather common. I presently have aircon but use it conservatively and have lived or worked in locations without it. It doesn't matter. Under no circumstances would I allow idle temps of 65 C. which for reference, is the very hottest that people unconcerned about energy cost and environmental impact will set their Water Heaters. It will scald a person, so obviously we never have to deal with an ambient temp within even 10 C of that condition. There is no severe obstacle, other than obstinance or ignorance, to achieving Delta T's of less than 10 C from ambient at idle. So even at 40 C I expect and commonly get an idle no higher than 50 C. That is VERY different from 65 C.... nearly 30 F different, for those less familiar with Celsius.


See above and exactly! and I already agreed that software fix should come first. My intent was simply to demystify the mechanics of working on laptops. It's a bogeyman easily dispelled. Hopefully that is clearer now.

Its actually the correct terminology.

http://en.wikipedia.org/wiki/Lapping

You can 'polish' a sphere. You cant lap one. ;)

*edit- actually, technically, you can 'lap' a sphere, but spherical lapping is not the same as plate lapping.

You are young, arent you?

20C+ idle deltas are uncommon now thanks to CPU frequency scaling and better power management. But it used to be fairly common.

http://www.xbitlabs.com/articles/cpu...y/p4-temp.html

Thats a 13 C delta with Hyper-Threading, 21 C without.

Given a 40C room thats a 53C/61C idle temp.... and that is with an aftermarket heatsink, which IIRC was better than the stock P4 heatsink...

You're forgetting that thermal paste isnt as good a conductor of heat as copper or aluminium.

http://en.wikipedia.org/wiki/Thermal_grease

'Science' says that more layers of thermal grease = worse performance.

I think you mean 'thermal pad', and I havent seen any evidence that vaenbys laptop has a large 'gap' between the heatsink and CPU.

Why even mention the possible manufacturer cost cutting when you have no evidence of it? Creating a dumbed down secanario doesnt make it true.