Active or Passive Cooling for P2?

[Grench]

I seem to remember seeing some kind of perf/power graph for modern SOCs, showing that when running at a small fraction of their total computing potential, they consume less electrical power than the. SOC used in the pandora.

This suggests that a modern SOC could be underclocked/throttled enough that no kind of heatsink would be necessary.

You could well argue that doing that would be a bad idea, but the same goes for fitting it with a large heatsink / fan combo.

Mmm....  I still like my idea of having a heat spreader sandwiched between the SoC and the battery to a metal/air radiator/grill on the front edge (inset enough to save hands).  Completely passive, gives the heat somewhere to go (out), and gives the unit a metal core that can act as a structural member to screw the case halves to (no more stripped screws).

Even if the SoC doesn't need the spreader/radiator for 'standard' desktop tasks, it would be nice to have it there in the event of running something that tops out the unit - without having the CPU have to throttle the task (as much).

Plus, you know that no matter what SoC gets used, someone on here is going to overclock the snot out of it.

 

[Mr_Loon]

...

Anyway, I hope you are not counting a heatspreader as a heatsink?

By a heatsink I mean something that is basically attached to the factory manufactured CPU after it has left the factory.

A heatspreader is often something that you simply cannot simply "take away" (or hey yeah, sure you can, but overheating really becomes a secondary issue if you do that ).

...

I am counting a heatspreader as a heatsink. Because that is what it is.

You are welcome to your personal definition of a heatsink but I'm using the accepted definition which from my understanding is : an object that absorbs and dissipates heat without the use of moving parts.

 

[TrashyMG]

Well a Heat spreader and Heat Sink are different things. A heat spreader is generally used to exchange heat from a source of heat to one that has a more favorable surface area. A heatspreader on a CPU for example move heat away from the die of the semiconductor mostly a flip chip attached to the substrate to a metal plate that will allow you to attach a heat sink to.. the metal lids on the processors below are all heat-spreaders.. they don't efficiently cool the processor alone and still require a heat sink.

 

[MrConfusion]

You are welcome to your personal definition of a heatsink but I'm using the accepted definition which from my understanding is : an object that absorbs and dissipates heat without the use of moving parts.

Well, my definition is simple enough:

Spreader:

Prevent a given area of the core from overheating locally because lots of heat is emitted from a point smaller than the tip of a pen.

Distribute the heat more evenly across the entire core.

You don't need to conduct the heat anywhere for the chip to operate properly unless the manufacturer so says.

But most importantly: If you buy a chip with one, you cannot remove it, because it isn't just "glued" to the chip or anything, rather it is a part of the chip.

So you really don't get to decide whether there's a spreader or not: totally depends on the chip you choose...

I'm OK with heatspreaders since if a SOC with a spreader is chosen, it will not affect anything, it just means the chip looks a bit different than plain plastic. Typically they don't even add that much to the height of the component, certainly if there's a USB port, audio port, EMI shields... well, anything like that on the same side of the motherboard it won't make any difference to the required height viewed from the "side" of the motherboard.

Sink: Move the heat (literally sink it) from the entire core area to another area of the device where the heat doesn't matter. Whether via plain conduction or a heatpipe makes little difference from my point of view (hmmm.... though you could of course argue a heatpipe is in a way a "moving part"... actually... naah... nevermind...).

 

[Mr_Loon]

Well a Heat spreader and Heat Sink are different things. A heat spreader is generally used to exchange heat from a source of heat to one that has a more favorable surface area. A heatspreader on a CPU for example move heat away from the die of the semiconductor mostly a flip chip attached to the substrate to a metal plate that will allow you to attach a heat sink to.. the metal lids on the processors below are all heat-spreaders.. they don't efficiently cool the processor alone and still require a heat sink.

I get where you are coming from, but a heat spreader is a heatsink, it might not be the type of heatsink that is seen used with desktop PC's but it is still a heatsink. Have a look at the picture of the heat spreader used with the ram on the wikipedia article on heat spreaders linked in your post. This is the type of thing I'd call a small heatsink.

No need to get too bogged down with technicalities, for the purpose of this thread please consider a small heatsink to be one that would not necessarily add to the overall thickness of a P2 and a large heatsink one that would have impact in terms of device thickness.

 

[TrashyMG]

I would agree with you assessment that what was shown on the memory are more a kin to heat sinks as there are fins that would cause some sinking, but there is a technical difference between the term heat spreader and heat sink.. I suppose it's a curse of working in the semiconductor industry were these technicalities bug me.

 

[MrConfusion]

I get where you are coming from, but a heat spreader is a heatsink, it might not be the type of heatsink that is seen used with desktop PC's but it is still a heatsink. Have a look at the picture of the heat spreader used with the ram on the wikipedia article on heat spreaders linked in your post. This is the type of thing I'd call a small heatsink.

Oh, you mean this picture :rolleyes: :

--------------------------------------------

Description

An edited photo of 2gb of DDR2 RAM with integrated heat sinks.

Source

self-made

Date

2-21-2008

Author

Victorrocha

Permission


(Reusing this file)

May be used for any reason with credit given to the author

----------------------------------------------------

It seems whoever took the picture knew exactly what he was shooting whereas whoever later added it to the heatspreader page might have been slightly confused. Gotta love Wikipedia...

But whatever... for me the main point is:

No "extra" weight on the PCB, all legs of whatever's stuck into it as visible as easily as possible (not covered by a block of aluminium if avoidable) and no size increase of disassembly trouble from glued on cooling accessories.

For me it looks like I could have fitted your definition of a small heatsink too, but whatever, I've already voted...

Edit: Added these ----------------------------

 

[Batou456]

One possible way to increase the heat dissipation capacity without causing major disruption to the Pandora's form factor, would be selective use of relatively flat Aluminum panels associated with the cut out for the key mat and backing of the screen.  There are multiple processes to make these limited panels that are not excessively expensive.  The rest of the case would remain injection molded polymers, although it's worth putting some consideration into having an Aluminum panel be the battery door, possibly with it sliding in on rails from the side of the case, to help with the dissipation of the battery's heat.

An appropriate metal hinge, and potentially other means of effectively conducting heat from the key mat panel to the screen backing and potentially battery cover panel would spread out the heat load and use the form factor to effectively have a radiator area of potentially 3 times that associated with a smart phone that can take advantage of natural convection. With good heat piping it _may_ be possible to use a 10.1" tablet SoC as intended for a 10.1" Tablet SoC, but this shouldn't be assumed out the door.  Some kind of heat spreading system that hooks into the dissipation panels would also of course be necessary to get it from the source to where it can be gotten rid of.

Care should however be taken when sizing the load associated with this as while these panels will enhance heat removal, it also means there's a need to be mindful about temperatures of something that will by its nature be resting in people's hands.

So in short if the Z3770 or something else in the 2-4 Watt range absolutely positively must be your SoC, instead of a Smart Phone SoC, then you'd want something along the lines of an Aluminum bar that the SoC would be thermally greased to and would carry the heat to the sides, where it would thermally interface with the side panels.  The side, battery, and screen backing covers would all ideally be textured to maximize effective area, and measures would be in place to insure heat was effectively carried and spread between the dissipation panels.  Coatings for the Aluminum might be appropriate, with a light vapor based diamond coating be ideal, although not necessarily cost efficient. 

It however should be noted that Open Pandora isn't really in a position to try to win an arms race with nVidia's Shield.

 

[Hồng Thất Công]

..

It however should be noted that Open Pandora isn't really in a position to try to win an arms race with nVidia's Shield.

If we have the opportunity I don't see why not.  But yeah it maybe very hard

 

[Klumpen]

Nvidia tries to fill the niche now too, but ED was first so I think I will stick to him and his devs.

I'm sure, that the next Shield will have even more of Pandoras advantaged of today and try to top them,

they may be studying community stuff for some time now.

 

[Hồng Thất Công]

Nvidia tries to fill the niche now too, but ED was first so I think I will stick to him and his devs.

I'm sure, that the next Shield will have even more of Pandoras advantaged of today and try to top them,

they may be studying community stuff for some time now.

 of course they have people spying on this forum since many crazy ideas have been born here

 

[Neelix]

I seem to remember seeing some kind of perf/power graph for modern SOCs, showing that when running at a small fraction of their total computing potential, they consume less electrical power than the. SOC used in the pandora.

This suggests that a modern SOC could be underclocked/throttled enough that no kind of heatsink would be necessary.

You could well argue that doing that would be a bad idea, but the same goes for fitting it with a large heatsink / fan combo.

This to me seems like a very bad idea.

QFT: 

Plus, you know that no matter what SoC gets used, someone on here is going to overclock the snot out of it.

Not just someone, I would expect a large portion of the user base to do so. Being an open device there would be no way to stop the user from pushing the device to beyond its limits, especially if they are using an alternate OS. I expect if this approach was used there would be either a fairly high return rate, or worse a safety risk from overheating devices.

- Neelix

 

[_wb_]

I would assume that all those recent quad or octa core SoCs could start a fire if you overclock them, do a stress test for a couple of hours, and somehow disable the temperature sensor that keeps it from overheating. Even if it has a 1kg heatsink and a big fan, if you store it in your pocket or a closed backpack and keep producing heat, something or someone might get burned eventually.

That is no reason to put a big ass heatsink in it. It's just an argument to make sure that the SoC has a temperature sensor with some automatic down-throttle and/or power down mechanism.

 

[slaeshjag]

As long as a temp sensor can stop a catastrophic overheating, the unit itself should be safe. I have more than a hundred times forgot to put my laptop to sleep before putting it in my bag for a few hours. It does get very hot however. Which will shorten the lifespan for at least the battery.

 

[Exophase]

I would assume that all those recent quad or octa core SoCs could start a fire if you overclock them, do a stress test for a couple of hours, and somehow disable the temperature sensor that keeps it from overheating. Even if it has a 1kg heatsink and a big fan, if you store it in your pocket or a closed backpack and keep producing heat, something or someone might get burned eventually.

That is no reason to put a big ass heatsink in it. It's just an argument to make sure that the SoC has a temperature sensor with some automatic down-throttle and/or power down mechanism.

The SoC will have thermal sensors embedded and it and a lot of the other components probably have hardware-based failsafes that will shut them down if their temperature is so high that they risk malfunction. Have you ever actually heard of anyone burning themselves from a phone?

 

[xiongxioi]

The SoC will have thermal sensors embedded and it and a lot of the other components probably have hardware-based failsafes that will shut them down if their temperature is so high that they risk malfunction. Have you ever actually heard of anyone burning themselves from a phone?

With respect, that's not strictly true, Exo. Many has been burnt by their phones due to the combustible batteries inside.

Okey, that's a bad attempt at a joke.

I do wonder, if cooling methods are useful for intense computing only.

Under the assumption above, does that suggest a high performance soc would be useless if we can't provide sufficient heat transfer? (since once the soc overheats, it will thermal throttle.)

 

[hedwards]

As long as a temp sensor can stop a catastrophic overheating, the unit itself should be safe. I have more than a hundred times forgot to put my laptop to sleep before putting it in my bag for a few hours. It does get very hot however. Which will shorten the lifespan for at least the battery.

Probably, but you could easily wind up in the situation where you can't use the device if the temperature outside gets above freezing.

Ok, that's not likely to be easy, but for a given temperature, it would be easy.

Personally, I'd kind of like the entire case to be built out of aluminum. Which has it's own issues as aluminum would have safety issues in terms of insulation. And probably be way too expensive anyways.

 

[Batou456]

Die-casting would involve Magnesium-Aluminum alloy, and you couldn't make up the entire case because it screws with wireless signals.  Hence the iPad for instance uses the glass as it's window to get wireless signals out.  An Magnesium-Aluminum alloy die cast doesn't have much to advertise itself, and I expect an approach involving stamped aluminum inserts into the injection molded polymer case would be superior in all aspects.  Being able to select where you want a thermal wall versus a thermal conductor is a good thing.

Machining straight from an aluminum alloy block is of course entirely unreasonable in terms of costs.

 

[Binky]

aluminum would have safety issues in terms of insulation.

Safety of what? The user or the electronics?

 

[Grench]

You could, in theory, use my idea above of a internal heat spreader & front radiator grill that also acts as the central 'backbone' of the chassis and do that in cast aluminum. Then the front fins could be vertical and more efficient.

My concern with doing that out of cast instead of thin sheet aluminum is weight. It's hard to cast aluminum to sheet thicknesses.