But please always keep in mind that there are settings that might both influence performance and consumption that you aren't able to read out on a running system: the dvfs table where operating points (clock speeds together with appropariate voltage settings) are defined.
LeMaker for example currently uses 5 operating points up to 1.3 GHz (with wrong core voltage settings in my opinion -- either 1.1 GHz is too high or 1.3 GHz is too low):
When I did my guitar review I defined
instead for tests with 1.3 GHz (the same amount of more voltage for an additional 204 MHz than what was defined to jump from 900 MHz to 1104 MHz). As a consequence with my kernel the Guitar might overheat way more easily than with LeMaker's settings when running at 1.3 GHz. But with their settings you might run into stability problems since it makes absolutely no sense to increase core voltage by 150mV between 900 and 1104 MHz and then just an additional 50mV to jump from 1104 to 1308. Their settings are wrong anyway.
Currently on the Roseapple Pi when you rebuild the kernel the following operating points are used:
(defining 240 MHz while using the same 950 mV as with 504 MHz might not help with consumption but only decrease performance with ondemand governor unnecessarily)
Anyway: If one tries to compare both boards without using a) exactly the same dvfs settings defined in the kernel sources, b) performance governor, c) identical definition of scaling_max_freq and d) identical trip_point_0_temp definitions, then results might vary for scary reasons:
- due to different core voltages the consumption might differ
- due to these voltage differences the S500 on one board might overheat earlier and thermal throttling jumps in
Regarding the latter you also have to have a look at
since this defines the treshold when throttling starts (LeMaker wants to decrease this value from 105°C to 95°C with their next OS image). And throttling currently is implemented as immediately adjusting the clockspeed to scaling_min_freq (Roseapple Pi: 204 MHz, LeMaker Guitar: 408 MHz -- you get the idea what happens once throttling starts?) unless a specific treshold is reached and then increases the clockspeed back to scaling_max_freq.
Without identical kernels and something like RPi-Monitor you're simply lost since devices that have to behave identical will perform differently. I would always use the same /boot/uImage when comparing two S500 devices since otherwise test results might differ for no hardware related reason.