That’s a driver for the NPU, not the GPU, mostly useful for people who want to run local AI models, image classifiers, etc. But thanks for sharing, I was not aware of the work on this. Glad to see another part of the chip being opened up.
I think we are letting perfection stand in the way of good enough here. In an ideal world everything would be open and libre but the only way to get that is on obscure and or outdated hardware. A GPU blob is a lot less restrictive to most users than the FSF guidelines artificial, self righteous, moral code. Being able to do what I want on a 99% open computer is more useful than a 100% open computer that I can’t do what I want on due solely to self imposed restrictions.
3 Likes
I agree, the goal should be to get to 100% free & open by any/all possible means, but in the meantime I’ll take a usable machine that goes as far as possible. IMHO only an FPGA-based machine will get us to full hardware and software freedom, but the FPGA ecosystem isn’t quite mature enough.
I’m interested in projects like Precursor that provide special-purpose FPGA hardware as a “bridge” to get us through the intervening years before a fast, general-purpose FPGA machine is viable. If I can have a fully trusted device for communications and key storage, with a less trusted device for general computing, that’s still an improvement over where we stand today.
2 Likes
Wasn’t there something from a moderator about maintaining a respectful tone? @bnys
2 Likes
You’re right I didn’t need that, but it was directed at the FSF not another user here. I agree with their end goal, but their policies are too uncompromising to be practical for most users.
Fair enough, so the GPU is not liberated yet nor is the boot blob.
I wonder if pocket mnt reform could have rk3399 has an option without wifi soldered in. I ask because, I have a rockpro64 with 4GB of ram and no wifi chip in it.
I would love to be able to use that instead of the current chip for my pocket mnt reform, or similar. I want to use a no blob setup.
Preferably with a nonsystemd distro or OpenBSD. Only thing is, I don’t know if you can use novelwriter in OpenBSD. That is my main app for pocket mnt reform
FSF considers ivy bridge, 2012 or newer to be non-freedom. Even if you dont need blobs to run on the software end.
They think intel me being removed is the only way it can be free enough. for those devices.
Truthfully, this is why they are not respected as much as they could be. They need their policies to be updated sometimes. The hardware blobs being disabled should be good enough. aka, intel me disabled is the same as intel me removed. This is why FSF fails sometimes. They are too proud to realize they have overlooked many things. SMH…
There are a few 3399 SoMs out there but as far as I can tell none will fit into existing adapter boards. I am surprised there are none that use the CM4 form factor.
That’s a shame, rk3399 seems to be usable without blobs from what I hear.
When choosing processors for the Reform series I’m not focused on 100% blob free SoCs/SoMs because the number of people who intensively care about this—more than about having usable performance—is not large enough to sustain us.
Of course I prefer SoCs and chips that have little or no closed-source firmware, and that have good mainlining efforts behind them. One thing I don’t accept is shipping closed-source software/drivers that run on the main CPU in user- or kernelspace.
Even if the ARM Mali GPU firmware is unfortunate, all in all, the RK3588 is a good compromise so far, as it unlocks a lot of performance (and fun) while being relatively unencumbered by closed-source artifacts. BTW, if you prefer, you can also use the RK3588 purely with software rendering without using the GPU.
I have no interest in bringing outdated Rockchip or i.MX models back. As Reform is OSHW, other people are free to do that of course.
12 Likes
Is RK3399 included in being too outdated for your tastes?
Although, its possible someone will fully liberate RK3588 someday in the next few years. Just gotta wait right?
No. Unless you have specific knowledge that someone is working on liberating an SoC, it’s most likely that nobody is working on it.
How can you be sure that someone isn’t secretly working on such a project until they think its worth revealing to others? Just asking
This question is irrational.
There are 8 billion people on this planet and not all of them think the way you do or me either. You can assume nothing.
That being said, if 100% FOSS means its completely open hardware, then that won’t happen for an extremely long time, not foreseeable future like think way beyond that and add twenty years at least.
If no blobs are needed, then nvm as this has happened before.
I don’t understand why you’re telling me this.
I’m reminded of something: cheap, powerful, completely open source - you can choose any one option.
I’m actually quite happy sometimes with not powerful and relatively expensive, if it helps guarantee a little bit of privacy, and I love MNT for even attempting to do what they do.
Perhaps it is pointless, but I felt the need to say this because you assume no one is working on liberating rk3588. That could be wrong is all I am saying.
The true answer is that its unknown at least now anyhow.
I have not assumed that.
Looks like I got some good news on that front, from a panfrost dev named Alyssa:
This is good news for everyone hence why I share it here.
I asked her and she sent me that link.
1 Like