For example, the i.MX 6UltraLite could be interesting because it doesn't, as far as I can tell, contain a GPU:

And the GPU in the main i.MX6 line is better and better supported by the ‎Libre drivers all the time (that's the chip in my Novena, so I have some hands-on experience with the GPU).

2016-07-31 14:51 David Boddie:

Not wanting to revisit the discussion about the A20 but, instead, think about other processors that might perform better in the comparison table on this page: http://rhombus-tech.net/crowdsupply/

I have no idea if it's even reasonable to discuss these as possible options, so don't all jump on me at once. :-)

(I have very little saying in the matter, so this is just my opinion).

As much nicer as other ARM 32-bits chipsets might turn out to be, I don't think that they bring substantial changes to the table. Of course, if there are thousands of cards to be sold and they're easy to design... sure, why not? But personally I wouldn't find them very compelling.

If the current campaign and everything else go well, I think that the next step for CPU-cards ("next" not meaning "immediate" -- just "next step in evolution", which might mean years of wall clock time) should be to attempt 64-bit options, either ARM or MIPS (or quasi-MIPS like Loongson).

To go forward into the "freedom trail" while keeping to "tried and true", maybe SuperH sh4 (32-bits) is worth a shot now that the patents are expiring and while the toolchain and things like the Debian port are still almost 100% functional. Maybe the people at 0pf.org will get designs with the 64-bit version at some point in the next few months/years. But support for the 64-bits SH-5 has been recently removed from GCC, so it doesn't look very good in that regard.

I'd also vouch for RISC-V (64-bits) or OpenRISC (32-bits), but Luke explained the problem with OpenRISC in the thread on slashdot, and RISC-V designs are not available yet.

I'm hopeful about RISC-V, though :-)

Cheers.