A bunch of pictures


Can you show me a picture of a PC which has a passive heatsink with holes in the case to let out the heat?

The only passive ones I know are simply heatsinks that are fully inside of the device.
I have built a PC like that. Full case with holes in it. Encasing a fully passive system.
I received one such computer (made into a decommissioned plexiglass wine cooler with a wooden lid) as a present. The holes on the lid did pretty much nothing. I made the holes bigger, pretty much no impact. I added holes to the bottom to get a natural draft, nearly no impact. I added a fan to the lid, finally a usable system. Natural draft based on buoyancy works but is not a good match for all cases.
 
Please re-examine the solutions that Askarus and I have put forward.
While I could see Askarus idea work (leaving aside the question wether it would be sufficient to get rid of enough thermal energy or beeing even very efficient), I'm not very "comfortable" with your idea (but I am certainly less of an "expert" in that regard then you are).The first question that came to my mind (back then when you first proposed it, months ago) was (obviously): Why hasn't anyone else done this before ? It is either not cost effective or has some drawbacks that make it an unviable solution. Don't get me wrong, I have no interest in dismissing your idea, I am just wondering about potential problems that may occur if you are creating an almost direct connection between the user and the inner workings of an electrical device. Are there any obstacles bringing such a device to market, like a special certification beeing needed? And there are of course mechanical problems, like how you integrate it into the case, or what would happen if the case has to endure a shock that leads to a plastic deformation of the metal part, while the synthetic material easily absorbs the shock by elastic deformation. How to avoid corrosion due to the acidic fluid humans usually produce ?
 
Last edited by a moderator:
Grench, how easy is it to assemble the Pyra with your solution?

Mine won't have an impact in complexity at all.

The machine will take the bridge and place it where it onto the solder pads.

Then you simply have to put the whole board into the case.

(Global Components working) time is money as well as complexity.

For me the whole folding and putting it between something process sounds too complex.

Maybe you could simplify it a little.

Would make it make a lot more usable.
 
I don't know what I'm talking about -- I failed a course in Thermodynamics about 15 years ago, and that's about it :)

But I think the Pandora has no heat problem, and the Pyra should not consume too much more power than the Pandora or else its battery life will be crap. If it gets too hot, there should be a mechanism that slows down the processors,  to a Pandora level of performance or even lower if needed.

In bursts (e.g. when booting or when starting an application), you want to use the full processing power, but not constantly. It's fine if it can't sustain its max performance for hours and hours.

It's a handheld. Weight and battery life are more important than sustained high performance. I don't want to make it significantly heavier just to get a heatsink/radiator that can sustain top performance for weeks, even though one battery charge will only last an hour at that level of performance.

We will already have a heavier battery and probably a heavier screen. I really want the Pyra to be not too heavy -- it's not just the volume, but also the weight that determines its pocketability (and ergonomy!).
 
I don't know what I'm talking about -- I failed a course in Thermodynamics about 15 years ago, and that's about it :)

But I think the Pandora has no heat problem, and the Pyra should not consume too much more power than the Pandora or else its battery life will be crap. If it gets too hot, there should be a mechanism that slows down the processors,  to a Pandora level of performance or even lower if needed.

In bursts (e.g. when booting or when starting an application), you want to use the full processing power, but not constantly. It's fine if it can't sustain its max performance for hours and hours.

It's a handheld. Weight and battery life are more important than sustained high performance. I don't want to make it significantly heavier just to get a heatsink/radiator that can sustain top performance for weeks, even though one battery charge will only last an hour at that level of performance.

We will already have a heavier battery and probably a heavier screen. I really want the Pyra to be not too heavy -- it's not just the volume, but also the weight that determines its pocketability (and ergonomy!).
I'm thinking of the docking station.

I think for a device like the Pyra it's important that some people might consider to use the device with a docking station.

I would not want to be throttled while on a dock and do some gaming,

Forcing the processor to a lower clockspeed is the worst solution for me.

This has to be done if everything else fails.

BTW: I'm using my pandora with DisplayLink at work all the time.

It's mostly for playing music but it gets pretty hot already (10% cpu load. Rest is charger and USB)

I would not underestimate that.
 
It's probably mostly the charge which makes the unit hot.

So that + cpu load drives the unit high. Had a bunch of temperature warnings when I tested units with notaz bench.
 
Last edited by a moderator:
Yes, of course the processor can be clocked down when the heat is too much (that's the same as in all smartphones), but we're basically discussing about worst case scenarios.


When I did my tests with high CPU load, the temperature went up to 80C within a short time but needed 10 minutes under full load to reach 100C, and 15 minutes to reach 110C.


Running KDE, Gimp, LibreOffice, FireFox, PCSX and Picodrive at the same time maxed out at about 80C (even when running it for hours), and that was on a yet unoptimized X fullHD system where X alone used 60% CPU.


Notaz sent me a better stress test (which also uses NEON), but I've not yet had the time to test it.


However, it is very very unlikely to have something run that does max out the SoC in all aspects. There is NOTHING on the Pandora that does that.


Worst thing you will probably do is run a compiler for a longer time, but then you will probably not hold the Pyra in your hand.


The normal temperature of the SoC is something about 50C.


The max temperature of Pandoras SoC is about 70C, and that's with full load.


The OMAP5 has a lower temperature for a couple of times the power the Pandora has.


Still, with full load over an extended period of time, it'll really get hot - but I fail to see the scenario for that to happen.


Normal usage will lead to short peaks, but those won't do anything.
 
I compile packages during one straight week, should be less with the Pyra ^^.
 
So, just Play whit 3 PSX Emus at once, whyle you dont charge, and it shouldnd be a problem..

Sad, that this seems now a big problem on pyra, while the pandora ditnt had this problem.. (no wonder: 2 ghz is much more than 700 mhz)..

I know from my Smartphone the Problem: Sometimes its "frezze" when its on the charge cable because it gets too hot.., and this is only USB...

On long hot Summer Nights (Jimi Hendrix), it should be Save to play Gameboy and Atari ST on Pyra..
 
So if there is a shim based solution, it could be thermal-glued to the SoC before assembly of the moduleboard. Then all the corners are spaced against the mainboard pcb with offsets at each corner.

Rectangular ones can be bought bulk, but its not that hard to stamp out a custom one that fits a bigger area.

Its kinda hard to follow the suggestions, because people are talking about solutions they arent linking to.
 
I don't know what I'm talking about -- I failed a course in Thermodynamics about 15 years ago, and that's about it :)

But I think the Pandora has no heat problem, and the Pyra should not consume too much more power than the Pandora or else its battery life will be crap. If it gets too hot, there should be a mechanism that slows down the processors,  to a Pandora level of performance or even lower if needed.

In bursts (e.g. when booting or when starting an application), you want to use the full processing power, but not constantly. It's fine if it can't sustain its max performance for hours and hours.

It's a handheld. Weight and battery life are more important than sustained high performance. I don't want to make it significantly heavier just to get a heatsink/radiator that can sustain top performance for weeks, even though one battery charge will only last an hour at that level of performance.

We will already have a heavier battery and probably a heavier screen. I really want the Pyra to be not too heavy -- it's not just the volume, but also the weight that determines its pocketability (and ergonomy!).
I'm thinking of the docking station.

I think for a device like the Pyra it's important that some people might consider to use the device with a docking station.

I would not want to be throttled while on a dock and do some gaming,

Forcing the processor to a lower clockspeed is the worst solution for me.

This has to be done if everything else fails.

BTW: I'm using my pandora with DisplayLink at work all the time.

It's mostly for playing music but it gets pretty hot already (10% cpu load. Rest is charger and USB)

I would not underestimate that.
Could perhaps the charger and/or dock provide some kind of heatsink in some way? Maybe the charging µUSB port could be thermally connected to the heatsink of the Pyra, and the charger and dock side of the connection could be designed in such a way to provide more heatsink and radiation surface?

In normal use, a Pandora consumes between 0.6 and 2 Watt of energy, and I guess most of that goes to heat (and some light). It has a 15Wh battery, so it lasts between 7 and 25 hours. (Of course you can always consume more energy by attaching USB devices, but that will not produce more heat where it matters.)

I assume that the Pyra will consume between 0.6 Watt and maybe 5 Watt in normal use. Maybe more if you do a parallel compile while playing a GPU-limited game, but that's pushing the definition of "normal use". It will probably have a 22Wh battery. So it will probably last between (roughly) 4 and 30 hours, depending on usage. (Of course less if you attach USB hard disks etc to it.)

(Also, of course that 30 hours battery life is only reachable if we can get the same excellent power saving stuff as on the Pandora.)

OK, so 4 hours is still quite a lot and people may actually want to use those 4 hours without down-throttling for thermal reasons.

Adding an extra copper or aluminium heatsink / heat conductor / radiator element to the Pyra is only acceptable to me if it does not add too much weight. Otherwise I prefer just stricter temperature-based down-throttling. The total thing should remain under 400g in my opinion. The Pandora weighs 330g (battery included). Smartphones are all under 200g, and most are closer to 100g. Every gram counts. A larger screen and a larger battery are not going to help. Sheets of metal don't help either.

Edit: Maybe something like this is useful? http://semiaccurate.com/2013/08/15/panasonic-make-a-10-micron-thick-carbon-phone-heatsink/
 
Last edited by a moderator:
When I did my tests with high CPU load, the temperature went up to 80C within a short time but needed 10 minutes under full load to reach 100C, and 15 minutes to reach 110C.


Running KDE, Gimp, LibreOffice, FireFox, PCSX and Picodrive at the same time maxed out at about 80C (even when running it for hours), and that was on a yet unoptimized X fullHD system where X alone used 60% CPU.


Notaz sent me a better stress test (which also uses NEON), but I've not yet had the time to test it.
With NEON it goes over 120C according to that sensor (without even using the GPU or DSP), but I don't really trust that reading.. Intel CPUs would have crashed or reset long before reaching such temps, but OMAP5 runs fine, so who knows what it's measuring..

However, it is very very unlikely to have something run that does max out the SoC in all aspects. There is NOTHING on the Pandora that does that.
This is a bad thing to assume. For example, Exophase did good job parallelizing DraStic and it makes good use of multiple cores. With slow SDL scaling and slow Xorg updates it can already max out both cores.
 
Last edited by a moderator:
This is a bad thing to assume. For example, Exophase did good job parallelizing DraStic and it makes good use of multiple cores. With slow SDL scaling and slow Xorg updates it can already max out both cores.
Maxing out both GPP cores can happen easily enough. But maxing out both GPP cores and the GPU and the DSP and whatever else is in the SoC, simultaneously, that I think is pretty unlikely.
 
but I don't really trust that reading..
Temp sensors having an offset or strange scaling are not that uncommon, if there's no sensor module that applies known corrections the measured temperature may be off by quite a lot.
 
Is it possible to run the CPU board behind the screen and distribute heat through the glass like tablets do? Connect to the main board through a ribbon cable rather than being connected directly. The screen will offer 2 large surfaces to dissipate heat into, one being glass that is more conductive than plastic. And neither of which you will be holding directly in your hands. Also add a copper plate to distribute the heat more evenly to prevent damage to the LCD.
This is a nice idea, but this would delay PYRA another Two MonthsTM, I think, and would make the PYRA lid thicker and hence the whole device.  Depending on the hinge design, we may have problem with this ribbon cable like we do on the Pandora.  But yeah, I like this idea.  Well, I like all ideas that prevent the SoC from being throttled down because of overheating.  I am a gamer like Askarus, and I would like PYRA to run as fast a it could :)  
 
Have you been able to test Reicast ED, or is that not possible yet?

If it runs well, that'd be a pretty good demonstration of the pyra's power for gamescom.
 
Temp sensors having an offset or strange scaling are not that uncommon, if there's no sensor module that applies known corrections the measured temperature may be off by quite a lot.
I've seen 20C degree shifts with the some of the products I work with.


Throttling is good and all, but you know we're a bunch of people that will push the hardware to the limits despite risk of melting the thing. So I think some measures need to be made, further testing should be done of course, which is of course what ED mentioned in his prior posts. But it's not time to freak out, I noticed it doesn't take much to dissipate the heat to tolerable levels.. I'm just concerned if it's just a bare module it could be an issue.
 
Last edited by a moderator:
This is a bad thing to assume. For example, Exophase did good job parallelizing DraStic and it makes good use of multiple cores. With slow SDL scaling and slow Xorg updates it can already max out both cores.
Maxing out both GPP cores can happen easily enough. But maxing out both GPP cores and the GPU and the DSP and whatever else is in the SoC, simultaneously, that I think is pretty unlikely.

That may be true,  but I'm thinking probably not really relevant to the discussion at hand...  Maxing out just the CPU core on the Pandora for any extended length of time it tends to get very hot.   And as Notaz pointed out maxing out both cores for extended periods is much less unlikely than one might think.

I'm the type of person to leave a program running for hours on end and come back to it when I have time. 

When I was playing a turn based RPG  in DOSBox  it seems it was running the CPU at full burn the whole time...  even when the program itself was apparently idle.  The first time I came back to it after leaving it be for a while I found that the Pandora was hot - and uncomfortably warm to the touch when I next picked it up.

When I accidentally left Widelands running overnight it not only failed to charge via the USB connection to my other machine, it was warm to the touch. (I woke just in time to see it shut down due to low battery)

- Neelix
 
Is it possible to run the CPU board behind the screen and distribute heat through the glass like tablets do? Connect to the main board through a ribbon cable rather than being connected directly. The screen will offer 2 large surfaces to dissipate heat into, one being glass that is more conductive than plastic. And neither of which you will be holding directly in your hands. Also add a copper plate to distribute the heat more evenly to prevent damage to the LCD.
This is a nice idea, but this would delay PYRA another Two MonthsTM, I think, and would make the PYRA lid thicker and hence the whole device.  Depending on the hinge design, we may have problem with this ribbon cable like we do on the Pandora.  But yeah, I like this idea.  Well, I like all ideas that prevent the SoC from being throttled down because of overheating.  I am a gamer like Askarus, and I would like PYRA to run as fast a it could :)
Added development time aside, this is the time to think of and do such things, not later after it becomes a problem or when an upgrade is slotted to ship. The real engineering concern I think would be how to connect the mainboard to the daughter boards through the hinge with the high data rates etc. And not run into a problem the pandora had with corrupted or broken communication going through the hinge. That is a legitimate concern with that solution. These creative designers reinventing the wheel should be focused on that problem, rather than radiators etc.

If internal temperatures become a problem, engineering any physical solution that would allow for maximum heat dispersion is going add thickness and/or weight regardless. Thing is, SoC's seem to only get hotter as they become more powerful, the thermal footprint over time is on a upward trend in mobile. Even if boarder line acceptable temp limits today, tomorrow will likely not be and will need a redesign of some sort. So if ED wants something that is truely a single design future proof device for drop in replacements, I think he should go this route.

Size wise, If stacked, or side by side (CPU pcb + LCD pcb combined are smaller surface area wise than the main board http://boards.openpandora.org/uploads/post-1-0-22204500-1403032996_thumb.jpeg) it's a good hand off to the alternatives.

The possible solutions I've heard are to add externally exposed radiators of different designs, cpu throttling, and passive air venting. But all of those solutions will still leave something hot in your hands. (external radiators might even cause injury and plastic is more of a insulator than a conductor) Mobile phones and tablets have been pushing heat slowly through the screen for years and I have personally yet to have one failed display because of it, and if you add a conductive heat shield to disperse the heat more evenly across the entire screen, it is less likely to damage to it, again, there's tablets and phones with way faster+hotter processors out there that are doing the exact same thing without problem.

EDIT: additionally, I'd rather have a hot screen than a hot battery, both of which generate additional heat when used.
 
Last edited by a moderator:
Back
Top