Heh, yeah. Native BC isn't possible, but they appear to have figured out some type of BC layer/workaround. No clue how it works.
Thanks ^^
I suppose it really depends how the numbers were sourced (assuming they're real). If they came from some type of spec sheet they're gonna be without the OC.
If they were measured using a watt meter on a SDEV-like device (more likely IMO since 28.6 is very specific), those would automatically pull more power than a production device because they contain an entire FPGA + ARM combo daughter-board and OC would depend on the software being tested (edit: and if NG currently implements oe::SetCpuOverclockEnabled).
Q. Sorry but have you found some evidence for this in the firmware?
A. No.
Q. Not trying to be rude at all, but...source?
A. "dude trust me"
(I don't really want to/can't say anything. I'm also not 100% certain it is what I think it is, some of this shit is like working with riddles.)