Hey Youness, hey Todd,
On 23.12.2017 04:06, Youness Alaoui wrote:
I think there is a plan to move librems to non-x86 architecture eventually (considering that RYF is our long term plan, there is no choice in moving out of x86 eventually),
that would be great.
I think the efforts on the risc-v front are the most promising and I think that's where the true competition to x86 will be, but to be honest, I don't really follow, understand or know much of anything that happens in the hardware space since I'm a software guy at heart (i.e: all I know is that x86, ARM, PPC and Risc-V use different instruction sets).
RISC-V is just a different ISA. Ok, it's free, but as it's BSD licen- sed, silicon vendors can build around it whatever they want. Delivering an owner-controllable platform is not in the scope of an ISA anyway. So RISC-V can't magically change the game by definition.
I hear a lot about PPC (with Talos for example), but I don't think PPC is as open as Risc-v (ISA or something). All I know about PPC really is that it was fun to reverse engineer during my PS3 days :) Anyways, as far as I know, for risc-v, it's not there yet, so we're waiting for that to be ready for the masses before moving to it. I have absolutely no idea if it's "close" or if it's really a long term plan for risc-v to be able to compete with x86 in terms of performance/power usage/features/etc...
It doesn't matter how close somebody else is. If I understand Purism correctly, the idea is not to jump into a market of owner-controllable devices once it exists, but to pioneer that market. The only thing that matters is what you buy *today*. The choice of i.MX for the Librem 5 is a move into the right direction. i.MX6 was the best thing you can get for mobile devices, IMHO (controllable and publicly documented). If you get the i.MX8 for it (and it turns out to be as good documented), all you have to do is to ask for a board with the most powerful version that physically fits a Librem 13 [1]. Then you can offer trustworthy hardware vs. performance and let your customers chose.
There are ofc alternatives to i.MX. Most use a graphics core where free drivers are a problem. Though, a proprietary driver in the OS is far less troublesome than blobs in your firmware (or the ME). And you might find something that is already available and delivers higher performance than the announced i.MX8 versions.
Once you buy a reasonable quantity of an SoC, you can ask if they can make the next generation with RISC-V instead of ARM. Unlikely to get that soon, but way more likely than Intel changing their silicon for you.
Nico
[1] I'm convinced that this is easily doable. At least compared to the effort you already put in liberating the unliberatable. If the i.MX8 turns out to be as controllable and well documented as the i.MX6, you'd be catapulted towards the end of your freedom roadmap.