Ram Heatsinks

[delicious-pi]

With the possibility of overclocking/overvolting, would ram heatsinks be a viable option?

There seems to be two types, Copper and Aluminium. Copper is heavier, but has better conductive properties and Aluminium, which is lighter and more suited to our pi's.

What are your thoughts?

[mkopack]

I suspect that's going to have mixed results. Remember that on the RPi (and this is true for the Beagle+Pandaboards as well) the RAM is on a module stacked on top of the CPU. That means you have no direct access to the CPU to remove the heat built up in it. Placing one of these on top of the stack will get heat away from the RAM, but there's still going to be a relatively large amount trapped under the RAM from the CPU.

Couldn't hurt to try though, right?

Anyhow, either should work fine, albeit the copper will work a bit better, but I'm sure it also costs quite a bit more $$. So totally up to you. I've considered doing something similar with my PandaBoardES...

[Tharic-Nar]

If you intend to use a heatsink, go aluminium, no point going with copper. Heat transference between the chip and heatsink isn't the limiting factor, but between the air and heatsink, so surface contact area with the air is more important than the material of the heatsink (to a point).

[Gert van Loo]

I know it sounds strange to a lot of people, but temperature is not the limiting factor for this chip. It works or it does not. Cooling the chip the might gain you a fraction of speed, no more. Search for the 'over clocking' thread where I explain all that.

[jamesh]

Dom's been running at 1GHz without any heatsink - no problem at all. Apart from voiding the warranty!

[meltwater]

Digging this thread up because from the latest post:
http://www.raspberrypi.org/archives/2008

It looks like heatsinks will be worth using (the turbo mode is controlled via temperature correct?).

Now to source some heatsinks...

[jamesh]

If you get to 85 then turbo mode is turned off - but, you would really need to be hammering the device, and in quite hot conditions, to get to that. Just read a post form someone hammering the GPU who only got to 70....

[Lob0426]

Point one: most people are running without heatsinks.

Point two: heat "dissipates" towards cooler areas. Heatsinks on the SoC are less effective due to the small air gap between the processor and RAM modules. But any heat moved away will benefit the SoC. With a way to check the SoC temperature, it can now be tested out as to how effective they are or aren't. If you are running yours in room temperature air you probably do not need them anyway. If you have a USB chip that is always hot, then use a heatsink on it.

Point three: heatsinks will not hurt anything. And they are effective on the LAN9512 chipset.

[rilski]

For a peace of mind, I'd rather have heatsinks installed on the raspberrypi.

Just for the peace of mind alone, since it'll help it cool down.

Copper/Aluminum either way works.

[hojnikb]

I only have a tiny heatsink on my lan chip, just becouse there is a bug in the 1V8 rail. It does get warm but still well within specifications. And also Pi looks awsome with this copperish looking heatsink

[malakai]

They do look cool I opt out if done improperly or with a poor heat sink they would probably compound the problem by not allowing heat out or at least that's what my head tells me.

[zboz]

A few hours with the dremel and a butter knife later.

[pygmy_giant]

No half measures there - might obstruct an add on board though....

[kriston]

I only have a tiny heatsink on my lan chip, just becouse there is a bug in the 1V8 rail. It does get warm but still well within specifications. And also Pi looks awsome with this copperish looking heatsink

But don't most Raspberry Pi boards that were shipped do not have this bug?

[Lob0426]

At this time all of the boards, before the version 2 PCB board, have the "bug". It does not affect all of the boards, but the error in the circuits is there. It was corected in the new PCB.

[Synthetic_Darkness]

For the original question, go with Aluminum. It's better suited.

But I must agree with everyone else. I've been running the overclock option on high. The Pi is in a closed box with no air circulation. Plus I live in South Africa and had a 29c degree day yesterday. And hammering the thing for 2 hours, got it to 61.5c

[jbhome]

On this thread of heatsinks, I have a production question.
As someone pointed out, the CPU is below the ram chip.
Would it be possible to place heatsink compound, or a thermal pad between the two to provide a better thermal path for radiated heat from the CPU.

Jim in the Kootenays

[zboz]

You will not be able to cram in anything that isn't in a really miniature syringe into that space. Thermal paste isn't even such a good thermal conductor that this would even be worth the effort. Silicon based pastes have a k of somewhere around 0,5-0,8 which isn't anything to cheer for. As a comparison, aluminium has a k of around 220. The effect of even a minimal flow of air between the RAM and CPU is kind of hard to measure.

I use my heatsink for providing weight to the pi in cheap plastic cases, which thick cables and heavy connectors otherwise throw around on and off the table. On top of the pi PCB was pretty much the only place to fit some extra metal material. Cooling is really not an issue.

[jbhome]

While the paste has a poor K, what is the K for not/limited contact?
I'm suggesting past between to at least aid the conduction of heat from the CPU to the memory and then via the aluminium/copper heatsink to air.
As you pointed out, you won't get a syringe or much of anything useful between the two chips once they are assembled, but during the assembly process, applying material would up the effective surface contact/transfer area.
Some quick research shows thermal conductivity
air 0.026
heat sink compound/silicon grease 0.20 to 0.70 at least 10 times as effective.

Keeping things cool is always good.
Getting the highest reasonable clock rate while not degrading the part also good.
Facilitating both these things at a low cost, priceless.

Jim

[zboz]

I'm seeing stall medium between the parts as the problem, not just the direct thermal transfer in air. Air will always be moving, although very little, when temperature differences are present in a system. Still, I'm not saying it's a bad idea, just that, at least for me, the controlled enough environment for credible measurements of the heat spread to be taken are quite hard to achieve unless the difference is really drastic, which I doubt.

Sure, why not? It still sounds like one will be left with a back problem and a mess of silicon paste but I'd be happy to be proven wrong!

[Burngate]

From what's been said elsewhere, there are a lot of metal connections between the SoC and the Ram, and most of the heat travels through those rather than across the air-gap. So filling the gap with some compound is unlikely to improve matters much

[RaTTuS]

...
A few hours with the dremel and a butter knife later.

That's actually pretty nice looking,
fixed via the mounting holes ?

would not fix on mine though ...

[zboz]

That's actually just a mockup; it's not mounted at all. It will be mounted with screws from the top so one can use spacers under it to still be able to mount the board using the holes. It hits the audio jack so I can't even get it straight on and a few components on the board are higher than on the CAD drawings I found somewhere. Edges are razor sharp on it too and it needs a fly cut finish for better looks.

[RaTTuS]

aha , ;-

[nhslzt]

to cool all 3 chips...
http://www.etsy.com/shop/nhslzt