Would a soft core which is 10 times smaller than the original chip being emulated always lose on power? What if it's clocked down so that the emulated chip doesn't run much faster than the original?
In automotive, is power so ferociously tracked? Most of the energy in a car is to get it moving, and even running the cabin fan requires more wattage than a reasonable control system.
> Would a soft core which is 10 times smaller than the original chip being emulated always lose on power?
If that were even possible (which it probably isn't--the fundamental "building block" of an FPGA is easily 10x the size of a similar fully custom gate), the soft core would still likely lose. Power in CMOS is proportional to C Vdd^2 f. Vdd hasn't moved in a while--so you've lost the biggest hammer in terms of power consumption. And C goes up when you change nodes--not down. And we're holding f constant according to your assumptions.
Furthermore, FPGAs are general purpose so they need things like clock spines that are optimized to work at their max frequency rather than at whatever frequency you chose. You burn a remarkable amount of power and area so that the chips can operate at say 400MHz even if you never run them faster than 8Mhz.
> In automotive, is power so ferociously tracked?
Not as such--automotive is generally more interested in whether your chip can survive power spikes. However, automotive is super sensitive to cost which is all about chip area.
> Most of the energy in a car is to get it moving
While true of ICE cars, that is less true than you might think with electric cars and regenerative braking. I can go for a very long time on 20mph roads between charges but the moment I get on a 65mph freeway, my battery gets sucked dry. I think atmospheric drag is something like to the third power of velocity--it's a really huge increase when you go even slightly faster.
However, as you point out, climate control is the real killer of energy--running AC or heating is power hungry.
In addition, the self-driving car folks discovered that running all that vision and radar and then processing it sucked down a remarkable amount of juice. Computes and sensors aren't exactly free.
In automotive, is power so ferociously tracked? Most of the energy in a car is to get it moving, and even running the cabin fan requires more wattage than a reasonable control system.