One thing I’ve wondered for a while is if it’s possible to run the VB at 3.3V rather than 5V, since everything is CMOS, which is usually pretty voltage tolerant (and the commercial V810 chip will run at 3.3V or 5V). There’s a bunch of things that it’d simplify, mostly interfacing with newer hardware which is typically compatible w/ 3.3V, but not 5V w/o level shifting (or series resistors at a minimum).
So… yes, it’s kinda possible. I played a game for a few minutes and it seemed stable, though typically you can’t run at frequencies as high as at higher voltages, but I’ve run a VB quite a bit higher than 20MHz @5V, so it seems possible that it’ll run at 20MHz @3.3V. The power OK pin needs to be connected to 3.3V to bring it out of reset.
The problem is the displays. The image is drawn fine, but there’s a random hazy garbage overlaid on the screen. You could probably add a regulator for 5V to the display and 3.3V to the rest of the system (since 3.3V levels should be fine to drive the 5V display since the displays are input only). That’s assuming the problem is the displays and not the Video RAM or VPU. It may also work better running it a little higher (at ~3.6V).
I’m pretty sure it’s a switching regulator, as the voltage drops, the input current goes up, so as the batteries get weak the VB starts drawing more current, putting an even heavier load on the dying batteries. The VB can run at less than 6.8V… you really need to measure under load.
Though yes… if everything is running at 3.3V, it should draw less current (especially if the switcher is modified to output 3.3V rather than running 5V through a 3.3V linear regulator).