Distributed Power Supply vs. Wall Power Supply

[sshukes]

I am doing some research for using Raspberry Pi3 in grid computing environments. I have a distributed power supply, 5v 20A 100W going to a spitter so that i can power multiple Pi's from the same source. Currently I have 2 pi's and a 2 5vFans wired up. I am using aircrack-ng as a benchmark.

If I use the distributed power supply my speed is 400pmk/s using the single wall plug in it is 750 pmk/s. Everything else is the same.
Has anyone else noticed something similar to this? Thanks

[W. H. Heydt]

Postmarks per second? Pots Meet Kettles per second?

[stderr]

I am doing some research for using Raspberry Pi3 in grid computing environments. I have a distributed power supply, 5v 20A 100W going to a spitter so that i can power multiple Pi's from the same source.

It immediately becomes clear why the pi should support, say, anything from 3.3v to 24v power in. 20 amps is ridiculous but utterly believable at 5 volts.

[Heater]

sshukes,

My experience withthe Pi 3, and that of others here, is that it will throttle it's clock back when it the 5v
supply droops. The upshot is that I do indeed see it running benchmarks at different speeds with different power
supplies. So far I have yet to aquire a wall ward that is "stiff" enough.

You can see if this is happening if the red LED flickers at all, Or use

$ vcgencmd get_throttled

So what about your 20 amp supply?

Are you sure the 5 volts it puts out is actually 5 volts and still 5 volts by the time it makes it to the end of your
distribution wiringing and splitter?

I suggest you measure it.

Be sure to use sufficiently "fat" wire capable of delivering the current without much voltage drop.

With that kind of current available I'd want to see some fuses in use for safety. But then you have another voltage drop problem....

[Heater]

stderr,

It immediately becomes clear why the pi should support, say, anything from 3.3v to 24v power in.

Nice idea. It would add to size and expense. No thanks.

20 amps is ridiculous but utterly believable at 5 volts.

There is no way a Pi is using 20 amps. A hundred watt's would set the thing on fire pretty smartish.

[hippy]

Currently I have 2 pi's and a 2 5vFans wired up. I am using as a benchmark.

If I use the distributed power supply my speed is 400pmk/s using the single wall plug in it is 750 pmk/s. Everything else is the same.

aircrack-ng seems to be a wireless thing and it is possible that different power supplies could have differing effects on the hardware and/or transmissions. If there is more or less ripple or EMI from one supply or the other that could create a difference.

A single supply would give a common 0V in the supply for both Pi's and circuits, where two isolated wallwart supplies may leave both with their own 'floating' 0V. That could have different effects including how things present themselves as ground planes.

I would suggest comparing how a Pi using both walwart and distributed supply benchmarks with the other Pi only on its own wallwart supply, then reversed. Perhaps benchmark with three Pi's, one always on wallwart, doing that three times with different Pi's on the wallwart to see if the difference is reproducible and consistent.