Yes, we did it, we used LS1021A in our prototype without doing thermal analysis and now kicking ourselves every minute of every day. The board is getting hot, very hot, and there is no option to increase the size of the enclosure or install a blower.
Perhaps, a little flat heat sink but that's pretty much it.
Looking through the messages on this forum I couldn't help but notice that quite a few people got in a similar situation and there are no readily available solutions to remedy this problem. I thought it would be nice to summarize all we know about LS1021A power consumption and the ways to reduce it in one place.
So it appears that lowering the CPU clock frequency does not do much for the power consumption. That shatters my long time understanding of the relationship between the clock frequency and power consumption in microcontrollers. Normally, dynamic power consumption is proportional to the frequency, so we can expect that a completely static process power consumption wold be almost negligible (and definitely under 1.6W in our case).
Disabling a core is equivalent to stopping its internal clocks (PW15 low power state). Estimated power saving from moving a core into PW15 state is provided in the QorIQ LS1043A, LS1023A Data Sheet, 3.5 Low power mode saving estimation.
For the LS1043A the saving is not significant, so there will be very small benefit in disabling cores.
So, it looks like LS1021A was designed to run hot... And I see no other way to cool it down but use external (mechanical) devices such as heat sinks, heat pipes, radiators, fans, etc.
Is there any other way to reduce power consumption in LS1021A? And I mean, hypothetically speaking, what would be the absolute minimal power consumption for LS1021A (other than turning it off, of course) and under what conditions?
Is there any other way to reduce power consumption with software other than sleep modes and clock frequency?