Frankenpad T60/61 cooling question

[bwaldow]

I have built two Frankenpads, one on a Merom board, and one on Penryn board, both with Intel GPUs. The Merom board uses a T500 HSF, and the Penryn board uses a T61 42W2820 HSF.

When I assemble the HSF and the bracket, there is a substantial gap between the bracket, HSF, and the GPU chip, such that I can physically move the HSF pad up and down between them using my fingers. An elastomer thermal pad takes up the gap between the HSF and the GPU chip so there is no play, no movement. That's why I use it.

Without the thermal pad there is nothing to hold the HSF pad onto the GPU. The HSF 'floats' between the bottom of the T60 bracket and the GPU chip on the motherboard.

I have read the suggestion of using thermal paste as on the GPU as well as the CPU, and I don't see how that can be done. I'd like to hear from someone that has actually done this.

At this point I'd expect to have to make some modification to the T60 braket such that it holds the 'floating' HSF pad down on the GPU chip with some spring force to avoid breaking the paste seal and introducing air gaps during transport of the machine.

How can I reliably force the HSF to contact the GPU chip so that thermal paste becomes an option?

Thanks,
Bret

[bmwman91]

You have to CAREFULLY bend the heat pipe so that the GPU / Northbridge portion of the heat sink sits lower. The trick is doing it so that the bracket is co-planar with the dies. Some modifications to the spring bracket are also useful.

See here:
http://forum.thinkpads.com/viewtopic.php?f=29&t=105114

[bwaldow]

When you assembled your parts, did you find a noticable gap between the HSF and the GPU below, and the bracket above? I want to know if your situation is the same as mine - it would be useful to hear positively that we have the same situation.

If the brackt is not pushing down on the HSF with at least as much force as I must apply to move the HSF down in contact now, there is nothing to prevent the HSF moving off the GPU and breaking the paste seal, introducng air between. What modificaiton does this for you?

But, again, if you didn't have a noticable gap between bracket, HSF, and GPU your situation isn't like mine.

You have to CAREFULLY bend the heat pipe so that the GPU / Northbridge portion of the heat sink sits lower. The trick is doing it so that the bracket is co-planar with the dies. Some modifications to the spring bracket are also useful.

Bending the heat pipe so that the HSF contacts the GPU chip merely transfers the gap from between the HSF & GPU to instead be between the HSF & the braket.

The gap is just a large as it was before, just now it's on top of the HSF instead of below it. I'd like to know more about the (fiarly substantial) modification to the bracket required to ensure a positive, no-chance-of-movement contact with the HSF from above.

Thanks,
Bret

[bmwman91]

The spring bracket mods are relatively easy. If using the stock setup, just bend the spring fingers downward. I also jammed some silicone based thermal gap pad material between the head spreader and spring bracket at one point, before switching to the T61p heat sink + T500 fan combo. The big aluminum block on there sits pretty high and the spring plate pushes on it without doing anything to it.

[bwaldow]

You have to CAREFULLY bend the heat pipe so that the GPU / Northbridge portion of the heat sink sits lower. The trick is doing it so that the bracket is co-planar with the dies. Some modifications to the spring bracket are also useful.

See here:
http://forum.thinkpads.com/viewtopic.php?f=29&t=105114

I read your post before making my Frankepads, and my bracket is modified as yours is to clear the corner of the metal block on the HSF. I notice also that you are using thermal pads, rather than thermal paste. I am using a thermal pad on the GPU, and Arctic MX-4 paste on the CPU (Arctic themselves say that's better than AS5, and lasts a lot longer).

My temeprature measurements come from repeated re-compiles of the entire software system in Gentoo GNU/Linux, while simultaneously running 7 sessions of glxgears to push the GPU. Temperatures climb from the first re-compile to the second (over two hours).

The Penryn system peaks at 83C. The Merom system peaks at 88C. These are the maximums for both the CPU cores and what appears to be the Northbridge - it's the hottest temperature shown aside from the CPUs. System details are below in my signature.

As your CPU and GPU use more power I would expect them to be hotter, and they appear to be so when you stress the system.

I'm emailing on the Merom system with Xubuntu 12.04 LTS, as it's low load, and working out Gentoo on the Penryn, so I must shut down and trade disks to get similar measurements on each system. This is what I did to get the figures quoted above - but getting more information would require some time and effort. The Penryn system has the Gentoo software load now, and I can 'ask' it any questions without that change-over.

So, my systems run cooler than yours now (expected as I'm not pushing so many watts), I have made the bracket modifications you show in your post (thanks for that, by the way), and I use a thermal pad on the GPU as you do.

But I don't see a way to remove the thermal pad and use thermal paste instead - the HSF is not held positively onto the GPU by the bracket.

Once I've cleared this up, and am certain I won't be changing the physical cooling system, I will be putting Thinkfan in and undervolting - I too want a very quiet, reliable system.

Thanks,
Bret

[bmwman91]

Aah ok now I understand the confusion here. I did not use thermal gap pads for the CPU or GPU. That is a phase change TIM (thermal interface material). It looks like a gap pad because the plastic backing was still on there protecting it in the pictures when I was getting imprints on it while aligning the heat spreader. Once I got it all positioned well, I peeled off the pink plastic covering. Basically, it starts out as a thin sheet of thermally conductive material, and then when it is heated to its phase change temperature it essentially softens and behaves just like thermal grease with a nice thin bond line.

The material is made by Honeywell, and I am not sure if it is commercially available. Since I work in the electronics industry, I have access to all sorts of fun stuff in our labs.

I did initially have some gap pad material on top of the heat spreader between it and the spring clip to try to increase the contact force, but that was later removed.

[bwaldow]

Aah ok now I understand the confusion here. I did not use thermal gap pads for the CPU or GPU. That is a phase change TIM (thermal interface material). It looks like a gap pad because the plastic backing was still on there protecting it in the pictures when I was getting imprints on it while aligning the heat spreader. Once I got it all positioned well, I peeled off the pink plastic covering. Basically, it starts out as a thin sheet of thermally conductive material, and then when it is heated to its phase change temperature it essentially softens and behaves just like thermal grease with a nice thin bond line.

And can you comment on your system having no gap between the HSF & the GPU or Northbridge?

On mine, the gap is substantial - several millimetres. The HSF is essentially floating in space, and the bracket does not push it down onto the GPU. Thus, even if I bent the HSF so that the pad meets the GPU surface in it's 'relaxed' state, the HSF is still free to come off under physical load - such as a shock when I put my bag down on the train.

I will only be comfortable using thermal grease when something is positively holding the HSF pad onto the GPU, so that it can't detach when there is a 'bounce'. I presume that would allow air in between and could spoil the thermal bond.

That's what I want to find out about - how do others modify their system to eliminate the possibility of the HSF pad coming off the GPU chip. In the case of the CPU, it's positlvely held by 4 screws - which also indicates the designing engineers thought the system could be subjected to sufficient shock to require such a positive fastening. Otherwise they would not have spent the money.

Has anyone else found this gap between the bracket, the HSF, and the GPU, and how did you deal with it? For now, my system is ok using a thermal pad, which is both a few millimetres thick, tacky, and elastic. It takes up the extra space and I feel I can rely on it.

Tuus assures me that I should be using thermal grease. Fine. But HOW given the substantial gap I find on assembly?

Alternatvely, if someone can comment that without undervolting (yet) my peak temperatures of 83C & 88C are good then I'll know not to worry. Does anyone get significantly better temperatures on their GPU/Northbridge than that without undervolting? If so, it's worth me pursuing using thermal paste. If that's as good as it gets, I don't need to concern myself about it anymore - the pads are good enough.

Thanks,
Bret

[bmwman91]

I just looked at your signature and it looks like you are using one with the integrated GPU, NOT the discrete nvidia one. If that is the case, then I think that that is your problem. From what I have come to understand, the integrated GPU sits a LOT lower than the discrete ones, My Frankenpad has the nvidia NVS140M GPU which is probably why there is no gap in my setup.

Thermal gap pads do have much worse performance then greases since they are so much thicker. However, the Intel integrated GPU also dissipates a lot less power than the discrete ones so you might actually be fine. Do you know how accurate the GPU and NB temperature reading are? What are you measuring them with? There are a lot of software tools that read component temperatures, and they are all different. As far as I am aware, the ONLY accurate one is tpfancontrol since it is Thinkpad-specific.

[rumbero]

I have built two Frankenpads, one on a Merom board, and one on Penryn board, both with Intel GPUs.

To my understanding there is no such thing as an Intel GPU. The Intel graphics capabilities are a built in feature of the Intel chip set, and thus there is no need to bother about cooling any such imagined GPU. I certainly didn't take any such specific measures for my own private Intel graphics Frankenpads, and i never ever had any heat related issues.

[TuuS]

rumbero, the T61 with intel graphics has Intel X3100 graphics which is a gpu it's just not a separate chip, it's integrated into the boards chipset. From the factory it's cooled by a different heatsink and uses Thermal paste instead of the grey pad that cools the chipset on nVidia boards which don't contain a gpu. When I use a discrete fan on an intel board I always remove the pad and replace it with thermal paste.

[Binh]

Without the thermal pad there is nothing to hold the HSF pad onto the GPU. The HSF 'floats' between the bottom of the T60 bracket and the GPU chip on the motherboard.

Since your frankenpad is based on integrated Intel graphic, you can not use spring bracket for dedicated nVidia GPU (I guess bmwman91 is talking about his nVidia version, 8890-A96). Instead, you need to use spring bracket for integrated GPU version (it actually consist of two separate brackets). Look at the hardware manual for T61 standard screen, section on removing fan assembly (page 84/92) for the picture of spring bracket for Intel graphics.
There is also a small problem: the T60/p structure frame does not have one of the two holes for fixing lower right part of the bracket, but without this hole the force on GPU/northbridge chip is still enough.

[Pete B]

As others have already pointed out, I think that you're referring to Intel Graphics as a GPU is causing some confusion here, or perhaps I'm not following you. But as I understand it both of your systems do not have a dedicated GPU, rather they have an Intel Santa Rosa Chipset (965 family) with built in graphics so that would be the GM version. This is not a T61 but I believe that it is at least the same chipset and shows what we are talking about with the 965GM chipset:
http://www.laptops-schematics.com/wp-co ... REV-1A.jpg

Is this what you are talking about? My preference is to refer to Intel graphics as Intel integrated graphics rather than a GPU.

That is usually referred to, or generically referred to as the Northbridge chip that you're trying to cool. It interfaces to the processor FSB and therefore handles the high bandwidth interfaces such as memory and graphics, slower interfaces are handled by the Southbridge.

I'm not an expert on these systems but I believe that since you're trying to cool a Northbridge chip and not a high powered graphics GPU you don't need more than a thermal pad to conduct the heat. On the other hand it would be good to compare how the original heat sink for that exact motherboard interfaces to the Northbridge since I believe that the heat sink you are using is designed for an NVidea GPU and a lower power Northbridge without internal graphics.

I'm planning to build a Frankenpad so this thread is of interest to me.

[bwaldow]

As others have already pointed out, I think that you're referring to Intel Graphics as a GPU is causing some confusion here, or perhaps I'm not following you. But as I understand it both of your systems do not have a dedicated GPU, rather they have an Intel Santa Rosa Chipset (965 family) with built in graphics so that would be the GM version. This is not a T61 but I believe that it is at least the same chipset and shows what we are talking about with the 965GM chipset:
http://www.laptops-schematics.com/wp-co ... REV-1A.jpg

Is this what you are talking about? My preference is to refer to Intel graphics as Intel integrated graphics rather than a GPU.

I cannot tell from a schematic which chip is being cooled on the motherboard. I have been referring to the object being cooled by my HSF as "GPU/Northbridge" in some messages, although I haven't been certain myself.

I'm not an expert on these systems but I believe that since you're trying to cool a Northbridge chip and not a high powered graphics GPU you don't need more than a thermal pad to conduct the heat. On the other hand it would be good to compare how the original heat sink for that exact motherboard interfaces to the Northbridge since I believe that the heat sink you are using is designed for an NVidea GPU and a lower power Northbridge without internal graphics.

The pad on the HSF that is closest to the palmrest contacts the chip on my motherboards. Given also some characteristics of what load heats it most, I can agree it is the Northbridge. Loading the 3D graphics has little effect on temperature, saturating the CPUs has an obvious effect on temperature, producing the highest measurement, and it appears to match the order of the list shown in the ThinkWiki page on temperature sensors.

Tuus affirms the factory uses thermal paste for this chip. The peak temperature I measure (previous posts) is acceptable, but if using paste brings the temperature down even more, that's acceptable too. I'm looking into getting some of the T61 clips to try to work out a way to arrange this - but that opportunity will likely be several weeks away.

[bmwman91]

If you have one open, it would probably be super useful to take a couple of pictures of the HSF you are using, and of the motherboard with the HSF removed. Maybe a shot of the gap that you are seeing, too. I bet that doing that would clear everything up fast.

[Pete B]

Just wondering Bret if you compared the heat and temps with the Merom vs. Penryn CPUs. The Penryn has a smaller process technology at 45 nm that should be a significant advantage. I suppose you'd have to have used the same HSF assembly to be sure or perhaps there is a drastic difference that makes it obvious?
Edit: I just noticed that you mention 83 and 88 deg but I'm not sure if the HSF differences make much difference.

[bwaldow]

My systems boot GNU/Linux - either Xubuntu or Gentoo. I have some spare disks and I have the restore DVDs for Vista on these machines, so if people can state what/where for testing software I could set up to test on that.

Just wondering Bret if you compared the heat and temps with the Merom vs. Penryn CPUs. The Penryn has a smaller process technology at 45 nm that should be a significant advantage. I suppose you'd have to have used the same HSF assembly to be sure or perhaps there is a drastic difference that makes it obvious?
Edit: I just noticed that you mention 83 and 88 deg but I'm not sure if the HSF differences make much difference.

The other approach I tried is to run one copy of 'burnP6' for each core, definitively saturating each one at 100%. This actually raised the temperatures a bit higher - 'temp1' which appears to be the Northbridge, went to 89C on the Merom and 87C on the Prenryn.

The Merom has the T500 HSF, and the Penryn has the T61 HSF. Both use Arctic MX-4 paste on the CPU, and a thermal pad on the Northbridge chip.

Tuus will be sending me T61 HSF clips in a week or so, and I'll look at what it takes to modify those to work with a T60/61 Frankenpad. If I can lose another 5C before I start undervolting, it's worth the work.

If I hear back about benchmarking on Windows before then... otherwise, I'll stick with the GNU/Linux measuring approach mentioned above in this thread.

Cheers,
Bret

[Pete B]

Nice work by the way. You might want to note ambient temp in case others want to compare since it is the differential that matters obviously. I worry about some of my systems that get into the high 80s at a room temp of 70, since we really don't run the AC that strong and often keep the house at 76 to 78 degrees. I guess these systems throttle back if you use them in the summer heat without AC with ambient up around 90 to 100.

[bwaldow]

You might want to note ambient temp in case others want to compare since it is the differential that matters obviously.

Ambient would have been about 20C at that time.

In the summer here (January - March) the ambient temperature can go above 40C, up to 45C. I haven't attempted running my systems at those temperatures yet, but I've run them in the high 30Cs.

I worry about some of my systems that get into the high 80s at a room temp of 70, since we really don't run the AC that strong and often keep the house at 76 to 78 degrees. I guess these systems throttle back if you use them in the summer heat without AC with ambient up around 90 to 100.

I have to speculate on where one finds an ambient temperature of 76, and what life must be like there...

The Earth is too cold for such temperatures anyplace in the breathable atmosphere, Venus is too hot. I think that one might carefully choose the right location in the gradient between the sunward and "dark" hemispheres of Mercury to attain such ambient temperatures...

Or are you mixing units?

The 89C I am reporting is approximately 192F, and the processors are rated to work at the temperature of boiling water (unit independent measurement, although we must specify ambient atmospheric pressure at one standard Earth sea level atmosphere).

If all your numbers are F scale, then your systems will work at 90F, unless something is very seriously wrong.

Cheers,
Bret

[Pete B]

Yes all the room temps that I mentioned were F, and I do understand that most report chip temps in C.

[Bobroberto]

Just finished installing a shim at GPU and now running around 70C, down from 88 - 93C. Used sterling silver, best heat conducting metal and had some around anyway. The squeeze out on old pad with measured with in a few thousandths of .032" a match for 20 ga. silver. Honed the surfaces flat on old whetstone. SMALL drop of Arctic Silver on GPU, place the shim, then a bit more on shim, slip fan in place and set the screws. Placed laptop upside down for the rest of the day to prevent any drip toward the board. Ready for summer in the Sonoran Desert.

[WalkGood]

Just finished installing a shim at GPU and now running around 70C, down from 88 - 93C. Used sterling silver, best heat conducting metal and had some around anyway. The squeeze out on old pad with measured with in a few thousandths of .032" a match for 20 ga. silver. Honed the surfaces flat on old whetstone. SMALL drop of Arctic Silver on GPU, place the shim, then a bit more on shim, slip fan in place and set the screws. Placed laptop upside down for the rest of the day to prevent any drip toward the board. Ready for summer in the Sonoran Desert.

How did that work out for you now?

[Bobroberto]

Ran cool all summer in rooms with 80-90+F temps