[Arm-netbook] ARM Cortex-A15

[Xavi Drudis Ferran]

El Fri, Aug 05, 2016 at 03:43:31PM +0100, Luke Kenneth Casson Leighton deia:
> ---
> crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
> 
> 
> On Fri, Aug 5, 2016 at 3:35 PM, Wolfgang Romey (hier)
> <hier at wolfgangromey.de> wrote:
> > Hallo Luke,
> >
> > surely you know the ARM Cortex-A15.
> 
>  yes.  it's the much-higher-power-performance variant of the A15.  as
> such about the only possible SoCs we could put into EOMA68 if they use
> A15 is if they were single-core.
>

I'm not sure I understand Luke's text. Are you saying that OMAP5432
(the SOC in the Pyra, dual A15 cores + dual small M4 cores + etc)
dissipates or consumes too much power for EOMA-68 ?


I have no idea how much power it consumes, I've only found: 

   The TI OMAP5432EVM uses a 12 Volt, laptop-style power brick. When the
   board is plugged into power and left off it used 0.7 Watts. At idle
   without X Window running, it used 3.4 Watts. Running one instance of
   "openssl speed" needed 4.4 Watts and two simultaneous instances moved
   up to 5.2 Watts. Playing the h264 1080p Big Buck Bunny took 4.7 Watts
   when using the special hardware video decoding. Note that these
   numbers are using the 60 Watt laptop style power brick which itself
   will likely be running at a fairly low efficiency at the low end of
   its power spectrum. So, for example, the OMAP5432EVM might only be
   seeing less than half the 4.4 Watts consumed at the wall.

https://www.linux.com/news/omap5432-review-texas-instruments-dual-core-arm-a15

I don't know enough electronics to understand it but I think it means 
they measured the power that went into the power source, not the power
that went out, and they think the power source efficiency might be less than
50% (but they don't seem to know). This tells little of the actual power 
used by the SOC or SBC, just an upper limit. That's with 2GB RAM and 
4GB flash. 

The passive heat dissipator in the photo (other photos here
http://www.ti.com/tool/omap5432-evm#0 ) does not seem
huge, but a dissipator obviously won't fit in a EOMA68 card. 

When I search for images of "Allwinner A20 board" I don't see
dissipators over the A20, so it looks like the OMAP5432 will 
get hotter. If you add flash and RAM (you might want to put more than
the current EOMA-68 cpu card) ... looks like too hot. But who
knows, it doesn't look like terribly hot...

If this is current and I interpret it right...

http://elinux.org/Embedded_Open_Modular_Architecture/EOMA-68#Physical_Dimensions

Connectors or antenas can protude outside of a EOMA-68 card. 
Maybe some type of kind of heat sink could protude also (if you can 
protect it against burning someone's hand or something), or maybe 
a card could be longer than 90 mm with the more heating components 
falling outside the device for easier ventilation ? Not sure that'd
be compliant, though.

>  the pyra uses the OMAP5 which is fine... i just won't use it because
> of PowerVR.  every single company that has ever gotten involved with a
> PowerVR GPU has had absolute hell from its users.  that includes
> Intel, with their early Atom series about a decade ago.
>

That's just the 3D acceleration. It could go unused just like Mali
goes in the RYF A20 card. 2D aceleration is Vivante GC320 (the
analogue to G2D) and might eventually have free drivers with etnaviv
(needs porting/adaptation). But there may be more components without free
drivers (video acceleration?) .