An almost naked Pyra

[vcoleiro1]

The portion of Fujitsu's solution that sits on the CPU is 0.6mm thick. It's the condensor which is elsewhere that is 1mm thick.

Not sure that makes a difference, but there's that.

By the way, was there no plan currently to have any cooling solution at all on the Pyra?
If you cant squeeze a 0.6mm ribbon in there , what chance is there to use a heat sink?

 

[slaeshjag]

The idea is to use the mainboard as a heat spreader.

 

[vcoleiro1]

The idea is to use the mainboard as a heat spreader.

So use the mainboard as a heat sink, that sounds like a terrible idea.

 

[fusion_power]

Since the mainboard is designed to have direct contact with the CPU, it'd be reasonable to assume that it's very tight with space.

Of course if you want to use such an heat pipe solution, you may increase the gap between SoC and main PCB. This could be done by raising the connectors on the daughter board for example. I'm not aware of the tolerances but I can imagine that 1mm additional height is not that critical. There is still alot of air in the Pyra case, at least compared to a dense Smartphone.
And of course it would need an extra designed solution, the existing one from the phone would not fit of course. But at least, heat pipes are not that expensive, the construction there seems simple enough to copy the principe without much trouble. (They may have patented their solution but nobody says you have to copy it to 100%).

But first, Ed must test temperatures within an fully assembled Pyra, only this will show what we really have there.

 

[vcoleiro1]

@fusion_power
Exactly, except that it would only need to be raised 0.6mm as thats the thickness of the pad that sits on the SOC.

I guess as you mention, ED will have a better idea of what if anything is needed when he does more testing.

 

[EvilDragon]

Fujitsu aren't the only ones with such solutions.
There are some synthetic carbon (I think) pads which let you distribute the heat or move it anywhere you want.
And special silicone mats that can shield the heat (so that it doesn't warm up critical areas like the battery). You can actually put these mats on a 200°C stove and put your finger on top of it

We've got both here and can test whatever we want. We got enough space between the PCB and the battery compartment to shield, we got enough space between the CPU and the PCB to use some heat distribution if need be.

The main idea is to move some of the heat to the SD Card slots (which have a large metal case and are open, so they can exchange the heat with fresh air) and use large copper areas in the PCB for heat distribution.
The most important thing is to DISTRIBUTE the heat, so that it doesn't stay in one tiny spot.

I've opened some tablets which use the carbon stickers I mention above. They mostly only have a 4x4cm sized carbon sticker on top of the CPU. It doesn't connect anywhere, it just exists to distribute the heat to a wide area, so that the overall temperature is lowered.

The EVM doesn't do anything like that.
It has a very simple heatsink on the SoC that doesn't do much: According to my measurements with a temperature meter, it only lowers the temperature by about 10 - 15°C.

Additionally, the EVM is not very optimized regarding power usage. Without even switching it on, I used up to 300mA when I measured.
And the heat from the EVM doesn't seem to come from the CPU only. When I did my testings, it seemed to me that the PALMAS (the power chip) also adds a lot to the heat, and it sits near the OMAP5.
I was able to measure up to 120°C when running the EVM with full stress test for a while, whereas Nikolaus only measured up to 65°C yet on the Pyra CPU board (also with stress test).

I can imagine the PALMAS heats up a lot on the EVM - it needs so much power that a Pandora AC adaptor isn't powerful enough to run it without crashes... and as all the power distribution goes through the PALMAS... yep, that generates quite a bit of heating.
The EVM was never built to be optimized in power usage, it's built so customers can evaluate all features.

When I was running the EVM with X and standard usage, I measured about 55 - 60°C.
Nikolaus measured about 35 - 40°C for the same on the Pyra PCBs.

So I don't think it'll be that bad, but sure, it'll generate a lot more heat than the Pandora when you fully stress it.

 

[Plume]

We really want to use OpenGL ES because it makes it easy to have arbitrary scaling, shaders, overlays, etc. But the drivers aren't ready.

For the next best thing I was hoping for at least something that's comparable to the fbdev DSS driver on Pandora, where you have at least two layers and one of them scaleable (but hopefully better than this). I have no idea how to use DRM but I'll try to learn about that.

As far as I know, fbdev is the old linux interface and DRM is the future (TM).

 

[DAP]

Be careful about using the SD card slots as heat sinks. I have had perfectly working SD cards get very warm all by themselves.

 

[Zink]

You remember correctly, however, the dip switches have been replaced with electronical switches, so you can even change between the eMMC and the MicroSD on-the-fly.

I was a little bit worried when I read this. I thought the additional MMC slot was meant for situations when the main flash breaks or wears out and is unusable any more (no software can be run). How one could switch to the MMC then, if you need to do that from software?

"Sword of Mana" and "Mario & Luigi" can't be finished in gpSP.
I had some very frustrating moments there later in the game thanks to emulation problems.

I had the same problem with Zelda: The Minish Cap - gpSP hangs in one spot of the game. What I did was to switch to the GBA for that single spot. The savegames are compatible. It didin't work smoothly, but was playable until I could save again, and then continue with gpSP.

 

[Klumpen]

I had the same problem with Zelda: The Minish Cap - gpSP hangs in one spot of the game. What I did was to switch to the GBA for that single spot. The savegames are compatible. It didin't work smoothly, but was playable until I could save again, and then continue with gpSP.

I did something similar for Sword of Mana: Switched to a slower emulator just for this part and switched again afterwards.
Unfortunately I didn't find an emulator that was able to run Mario & Luigi without crashes and didn't get to a secure save point in time after my last quicksave so would have lost hours of gameplay anyway.

 

[Exophase]

I was able to measure up to 120°C when running the EVM with full stress test for a while, whereas Nikolaus only measured up to 65°C yet on the Pyra CPU board (also with stress test).

Sounds a lot better. 65C might be sustainable depending on where that measurement is and what the temperature is like by the time it gets to your skin (or other sensitive components)

Maybe 2x1.5GHz isn't completely unrealistic (if you're willing to sacrifice battery life)

I'm still not totally sure about the representative of the stress test though, would be good to see some other tests too. But of course only once more important problems are tended to!

 

[EvilDragon]

Well, as mentioned, Drastic runs fully playable with hires rendering on my OMAP4 phone (at least Okamiden and some other games I tried), so it would be weird if the vastly faster OMAP5 couldn't do it.

 

[matzesu]

Allready playable emus whe allready know from the Pandora but whit hires rendering or whit more batterie time sounds quite awesome..

Im curently starts to play minish cap on pandora gpsp, so if there is an bug in the emu, i think i should wait until the pyra and mgba came out..

 

[PokeParadox]

Allready playable emus whe allready know from the Pandora but whit hires rendering or whit more batterie time sounds quite awesome..

Im curently starts to play minish cap on pandora gpsp, so if there is an bug in the emu, i think i should wait until the pyra and mgba came out..

Nah... just save often. The native saves are compatible.

 

[WizardStan]

How one could switch to the MMC then, if you need to do that from software?

Easily enough: during boot check the uSD slot first for a bootable card, then fall back on the eMMC.

 

[fusion_power]

Fujitsu aren't the only ones with such solutions.
There are some synthetic carbon (I think) pads which let you distribute the heat or move it anywhere you want.
And special silicone mats that can shield the heat (so that it doesn't warm up critical areas like the battery). You can actually put these mats on a 200°C stove and put your finger on top of it

Carbon is heat conductive? Interesting, didn't know.
Well, good to hear that there are tolerances inside the Pyra that can give options for various heat spreading solutions if really needed.

 

[Grench]

Carbon is heat conductive? Interesting, didn't know.
Well, good to hear that there are tolerances inside the Pyra that can give options for various heat spreading solutions if really needed.

Sort this by descending TC: https://en.wikipedia.org/wiki/List_of_thermal_conductivities

Most of the top thermal conductors are different forms of carbon. Unfortunately, pure forms can also be quite expensive. Non-pure forms (mixtures of different carbon based chains) can still be effective though.

I see this as a two part problem. Dispersal and removal.

First is to disperse heat from an intense point source (SoC). ED's solutions above do this by spreading the heat throughout the case. This helps take spike loads away from their origin. Spreading heat this way is immediately helpful, but over time the heat in the unit itself would continue to climb if there is no place for it to exit.

Second is to remove heat from the unit as a whole. It may prove to be difficult to get the heat out of the case once it is away from the SoC. This continual heat build up is sometimes called heat soak.

Standard Polycarbonate is an effective electrical and thermal insulator (~2W/mK). There are alternative formulations (more expensive) of Polycarbonate that are many times more efficient at heat transfer (up to 40W/mK). There are trade offs though as some of the other properties of PC get changed a bit in the process. It is an open question as to whether or not the PC used in the Pyra case examples will be efficient enough at removing heat for the system to reach steady state with the SoC under it's shutdown limit.

There are two paths to getting heat out of the unit. The first is through the case itself. Depending on the formulation of the plastic, it's surface area and whether it then conducts to solid (desk) or liquid (hands) or air (suspended) can all make a difference on how efficient the case plastics are at shedding heat from inside the unit to an external environment. The second is to create a shorter path using a more efficient transport to a more efficient heat exchanger. Aluminum, for example, could be used to conduct heat directly from the SoC to an external heat exchanger on the outside of the case. The drawback to this is that it pretty much defines creating an intentional hot spot - potentially a very hot spot.

My preference, if this needs solved, would be for solving this with case plastics formulation. Making the case plastics transport heat better would allow the heat to exit the system via radiation (small amount) and conductivity (mostly) to the desk it sits on (big heat sink) or the human circulatory system holding it (liquid cooling). If the unit is suspended in non-moving air or stuffed in a sock drawer while on full tilt, I would expect it to get hot and shut.

All of this is 'unknown' though as we don't know yet what formulation is being used for the plastics and if it's really an issue. Lets let ED do some tests and see where it goes.

 

[fusion_power]

Again something learned, the Interweb rulez!

I remember an older sketch of mine, wehre a heatspreader could be placed on the Pyra, if some sort of heatpipe from the CPU is needed.


Not sure if this place could be connected to the SoC but at least the heat spreader would not disturb much there, under the hinge.

 

[ElPoco]

The second is to create a shorter path using a more efficient transport to a more efficient heat exchanger. Aluminum, for example, could be used to conduct heat directly from the SoC to an external heat exchanger on the outside of the case. The drawback to this is that it pretty much defines creating an intentional hot spot - potentially a very hot spot.

Now I'm picturing the thermal clips from Mass Effect, or a steam vent on one side of the Pyra.
Or maybe it could be used to heat water for tea/coffee to fuel your gaming session.

 

[Plume]

Carbon is heat conductive? Interesting, didn't know.

Basically, any electrical conductor is also a good thermal conductor.

Don't ask me why exactly.