Bugs and Calculations


Only if you use the right DTB...
But still many DTBs and all the trouble to load the right one.
There is a single DTB that defines the union of all components and drivers that can maximally be available. I.e. the completely populated Pyra boards. If a component described there is not available, it will try to load the driver, and run the probe() function. Probe will not find the chip and further on, the extra information in the DTB is ignored.
A different DTB is only needed if there are PCB layout changes (e.g. a chip is moved from I2C2 to I2C3 or gpios are swapped/inverted or used differently). This are major device variants. Therefore we have a different one for the first prototypes (because of such wiring differences) and a third one for the OMAP5432EVM.
So there is no need to have different DTB files for each subvariant created by leaving out components of an otherwise identical design. It is like removing a light bulb from the socket which does not need to redraw the house plan of the light switches :)
Different eMMC sizes are automatically detected (like different MMC card sizes). Only different RAM sizes are tricky, but more a problem for different U-Boot/MLO variants.
 
Different eMMC sizes are automatically detected (like different MMC card sizes). Only different RAM sizes are tricky, but more a problem for different U-Boot/MLO variants.
So I suggest a 64 GB eMMC variant that "happens" to have 4 or 8 GB RAM chips, but only 2 GB RAM of which is advertised/officially supported. :)
Would that work?
 
In theory yes, but the MOQ (minimum order quantity) for the eMMC is 1600, which means we need to sell that many units of that variant.

With the RAM chip, we don't need two different chips but we just need to populate less of the same one.
 
In theory yes, but the MOQ (minimum order quantity) for the eMMC is 1600, which means we need to sell that many units of that variant.
With the RAM chip, we don't need two different chips but we just need to populate less of the same one.
So in theory it wouldn't be that big of a deal to offer both 4G and WiFi-only with 2GB or 4GB?

If you got two prototypes with different amounts of RAM but the same setup in general, could you somehow measure the additional power drain or just test how much longer it runs with 2 instead of 4?
 
In theory yes, but the MOQ (minimum order quantity) for the eMMC is 1600, which means we need to sell that many units of that variant.

With the RAM chip, we don't need two different chips but we just need to populate less of the same one.
Although, if the memory setup is like Exophase mentioned in another thread, populating half of the chips would incur a severe performance penalty.
 
If you got two prototypes with different amounts of RAM but the same setup in general, could you somehow measure the additional power drain or just test how much longer it runs with 2 instead of 4?
Would be nice to know the numbers with the 4 GB hardware with only 2 GB enabled (powering off one of the modules)...
 
So in theory it wouldn't be that big of a deal to offer both 4G and WiFi-only with 2GB or 4GB?

Since the 4G chip is on mainboard and the RAM chips are on the CPU board, I think they can be mixed and matched, right?
 
IF after testing the Pyra, it is found that the SoC thermally throttles after a while . Then maybe it might be worth adding a tiny fan which only kicks in at a certain temperature.

I noticed on the GPD WIN discussion, that someone had put forth this tiny fan which is only 10mm x 10mm x 3.5mm high


Here it is here: https://world.taobao.com/item/40085...name=h.qU2T5&cv=AAETfMjW&sm=92d8ca&app=chrome


TB2na.oXVXXXXXzXpXXXXXXXXXX_!!1098980938.jpg
 
@ED.

Get one and try it out. - it's supposedly meant for a mini notebook, so it should suffice for the Pyra. They come in all sizes if you can fit a bigger one.

But yeah, worth getting one to test out. If not to quell the fan discussions that will otherwise go on for ever :)
 
Last edited:
Well, to be able to blow out enough air... I guess it will need to whirr with an awkward sound :)

IFF you're going to try something like that, may I recommend having it mounted high in the non-wired hinge side void (where Pandora has light pipes) and blowing -in- such that the SD card slots become the exit vent?

If the fan pushes -in- 1 mole of 21*C air and the air picks up enough energy to move to 100*C, a considerably larger volume of air will exit than enter.

1 mole of air weighs ~28.8 grams.
1 mole of air at 1 atmosphere and 21*C is ~24 liters (volume per mole in).
1 mole of air at SLC is ~24.5 liters.
1 mole of air at 1 atmosphere and 100*C is ~30.6 liters (volume per mole out).

Air Specific heat = 1.01J/gC

Changing air from 21*C to 100*C for 1g of air = (1*79)/1.01 = 78.2 Joules transferred per gram of air in.

Assuming the system needs to shed 3 Watts per hour, 3 Watt-hours ~ 10800 Joules.

10800/78.2 = 138.1 g of air per hour.
138.1 g ~ 4.79 moles ~ 115 liters per hour ~ 2 liters per minute in.

It wouldn't have to blow all that much air in to be effective. If you're going to consider this, be sure to have someone else run the numbers. What I did above is very back of the envelope and doesn't take into account any of the other methods for the heat to get out.
 
@Grench I assume by 100*C you mean 100 °C? To heat up the air to come out at an average of 100 °C, you'd need the CPU (presuming it to be the warmest part) to be way hotter than 100 °C. I like my CPU to be way cooler than 100 °C. (For those who live in a world with weird physical units: at 100 °C water boils, 21 °C is nice room temperature.)
 
Back
Top