I've wondered about this myself, and even posted a similar question here some time ago. I've started to suspect the issue may be with Windows itself rather than nVidia slacking off.
If you read nVidia's doc on the Variable SMP architecture (
https://forums.windowscentral.com/e...s%2Ftegra-whitepaper-0911b.pdf&token=dSBp3ngI) the way it's supposed to work is that as CPU load decreases, the system will park cores until it's only running on a single regular core. If the load drops low enough, that single core gets switched to the battery saver core.
On my Surface RT with both Windows 8 and the 8.1 preview, leaving the resource monitor running, I've never seen less than two cores active. Up to two cores get parked, but never three. So not only is the 5th battery saver core not supported, but apparently neither is running on a single regular core. This brought back a memory. Windows NT had (has?) distinctly different HALs for single and multiprocessor systems. There are support articles like this one (
Use Device Manager to Switch from Uniprocessor to Multiprocessor Support) that talked about using Device Manager to update the HAL if adding a second processor to a single processor system. There are also some discussions like this one where potential performance issues are mentioned if running a multiprocessor HAL on a uniprocessor system (
ACPI Multiprocessor Hal on Uniprocessor Machine). Now these are somewhat old, and don't directly apply to this situation with the Tegra, but given that Windows RT currently never seems to use less than 2 cores, I wonder if there's some architectural issue in Windows NT that prevents it from being able to dynamically switch between single and multicore operation. It wouldn't surprise me since prior to the Tegra, it was probably never a design goal.