I'd go with whatever option provided the best hardware at the best price but still allowed homebrew software .. and supported a Linux environment to actually run software in. I think GP will do the right thing and make the majority of the community's desired changes-- e.g. 4:3 aspect ratio, D-Pad and perhaps the S3C2460 instead of VRender-3D. But if they didn't, and I could still bring my existing GP2X works easily over to this new machine (e.g. Linux, SDL) I'd be fine with that too.
A big thing is, I need to know other people intend to get one, or what is the point of developing for it? With GP, there's the loyalty of the GP32/GP2X community.
Is this some hypothetical thing, or an actual planned machine? (It sounds TI OMAP2-based..
)
EDIT: Ah, I see now after reading through your post history; you intend to build some sort of ARM11-based console ... well, if you did go that route, and honestly think you could fit that fancy hardware into such a small price margin, I'd love to see it. Probably the best you could do, in my opinion, is an OMAP2-based machine with ARM11 at a decent clockrate (that'd provide integrated FPU/Vector processing capability..) OpenGL-ES compatible 2D/3D acceleration (PowerVR MBX Lite/RS would be fine), an acceptable amount of memory (ideally 96MB+ of Mobile DDR), a high-quality, decent-sized 4:3 display, a D-pad and a reasonably compact design. If it ran Linux and supported SDL I'd port Stargazer to it, and it'd be extremely easy ...
I wonder though, what sort of company, if any, do you belong to? What are your connections, how would you get this to sell? GP32/GP2X have a fan following, which is why we know GP will be able to move a significant quantity of XGP units. But with such expensive SoCs you'd need, and not any real following to build on, I wonder how you'd manage to turn a profit without charging a fortune per unit. If anything a cheaper comprimise on system components may be better to get on your feet, at least.