IMHO what we need is not more powerful boards but more affordable ones.
At $99 it's too expensive for small projects.
At $99 it's too expensive for small projects.
Do you have any pointers to good boards at this range?
I'd pay a premium for that.
Although for me, it's Haiku that I want to run on here, and for now it's x86 :)
The almost universal lack of Mainline support is what makes ARM so frustrating. If it is not mainline, I don't want to know. Get a Minnowboard MAX and I suspect all the SoC functions will be available with a mainline kernel.
Get any ARM SoC and you are stuck with choosing between a mainline kernel with incomplete functionality OR a custom kernel that might have more functionality (if you are lucky). iMX6 might be the closest SoC to getting kernel sources for full functionality (but still stuck with blobs for Vivante GPU), but how much of that has made it to mainline (genuine question)?
Not sure what you mean by "full functionality" other than GPU driver source, and most of us using these for production purposes don't really care about that at this point.
Does it need to be mainline to make your product functional?
Rather my experience is merely playing around trying to use ARM-based devices as general purpose computing devices in the same way you would use an x86 device, e.g. Allwinner A10 set top box as a low power server and the Samsung ARM chromebook. In that context, I think mainline support is important. Otherwise, as someone who is not a developer, getting even simple things to work requires so much more screwing around than anything x86.
As for "full functionality", look at an SoC like the Allwinner A10/A20 and see how much of the functionality they built into the SoC can actually be used in linux, despite the excellent efforts of the Linux sunxi team.
Edit:
Grepped through the FreeBSD Ports tree for architecture specific ports, and the most common reason not related to using a binary of some sort tends to be that the program uses embedded assembly, usually floating point and/or SIMD extensions, and the developer hasn't coded a fallback of one sort or another. Some of them could probably be ported fairly easily, others may be more difficult projects where it'll be easier/mentally cheaper to just spend the extra money to get an x86 board and be done with it.