Imperf3kt wrote: ↑
Sun Mar 18, 2018 12:07 pm
You may note that that document is dated 2007.
Here's one from 2012
You can't fix a specification by using another totally irrelevant. USB PD (power delivery) is a new and great spec that deals with higher amperage ratings and even defines higher voltages since 5V is a joke with high currents since both cable and contact resistance is too high and the voltage available at the other end of the cable will drop too much (again: the problem is 'voltage-drop under load')
You do not try to push more than a few A through circuitry at 5V since it's a bad idea (again: Ohm's law). The USB PD specs define a maximum of 2A at 5V, for any higher amperage rating the voltage has to be increased too since you do not use 5V with high currents anyway.
See here for some details and the 5 profiles: http://www.electronicdesign.com/interco ... r-delivery
So if a device wants to be compliant to USB PD it needs to implement at least Profile 1 and new receptacles and cables are needed. The maximum current rating of an PD compliant Micro USB receptacle/plug is 3A but this is only allowed with 12V or 20V. So while such a PD compliant Micro USB connection using new cables and connectors is good for 3A no one right in his mind does try to implement this at 5V since... voltage drop too high.
Anyway, this whole spec stuff doesn't really matter since in reality based on my and other tests you can't exceed 9W consumption with the Pi and this is well within the Micro USB specs (5V / 1.8A). Still the problem is usually under-voltage (the lower the voltage the more cable/contact resistance becomes a problem --> voltage drops) but users are usually not aware. Those with a display seeing the yellow lightning bold might get the idea they ran into under-voltage but are told to upgrade to higher amperage ratings. Replacing their PSU usually then fixes the problem by accident (better cable between PSU and board).
Headless users usually don't take notice but if I understood @jamesh right, Pi Foundation wants to fix this: viewtopic.php?f=63&t=208057&p=1287779#p1287779
If you want to test for the limits by CPU utilization you would need to use something heavy. Stress, sysbench is all pretty lightweight stuff. You need to get cpuburn-a53 for example (and with a Pi 3 then run for sure into under-voltage being frequency capped at 600 MHz): viewtopic.php?f=63&t=208057&p=1287349#p1287276
Edit: The Pi is not PD compliant of course and I don't think PD compliance is of any use since only increasing the costs. With USB-C (cables and receptacles made for up to 3A by design while Micro USB receptacles are rated for 1.8A max and 'average' USB cables only made for 500mA --> voltage drop) there exist two true 5V modes with 1.5A and 3A that only require some primitive negotiation. That's the way to go in the future and then this future Pi might also be used to reliably power external disks and such stuff... at the moment with that input circuitry that's gambling