this is something i think regularly about doing.
what stops us from having an open source motherboard for a modern-ish platform, like am3/am4/am5? i know firmware can be a pain, but if the chinese manufacturers can do it somehow we could too.
the motherboard is usually the component that’s the most fragile in a machine without a gpu, thus what makes most sense to be open and repairable. plus not having to rely on the goodwill of manufacturers to actually sell their shit to us.
Good news! On the firmware side there is some serius work being done to support it!
https://www.basicinputoutput.com/2025/01/amd-opensil.html?m=1 is AMDs work to opensource hardware intialzation
https://opensourcefirmware.foundation/projects/ Has a good list of mostly BIOS/UEFI replacements.
Idk about memory, Usb, sata or PCIe controllers though. (something else?)
i was aware of coreboot and opensil, but i figure they are not ready for something like this yet?
they would solve the hardest part. nowadays everything lives under pci express and a lot is orchestrated by the cpu/pch, which is the responsibility of… firmware to initialize. deciphering and reverse engineering this process would not be trivial at all.
i know some smaller controllers have firmware baked in, and some are on an outside flash chip that may or may not be able to be read and copied. some may have good documentation available and even reference implementations you might see repeated on different boards. some others might be easy to obtain.
but yeah, i don’t think it would be easy (or possible at all) to have firmware be all open. i would bet there are clauses in some of their licensing officially forbidding us of all of this.
RISC-V
I have very mixed feelings about RISC-V since the non-profit organization moved their headquarters to Switzerland after taking a bunch of US tax dollars and grants to develop it at UC Berkeley.
You clearly haven’t been to the us in a very long time
With how things are going in the US right now. I’m not mad about people moving certain groups out of the country. I’m sad, but not mad.
Note: I know the reason they moved to Switzerland is unrelated.
i like what risc-v is supposed to do but it is still some ways out though.
The firmware is probably going to be the hardest part. I have heard that people who write micro code are rare and far between.
That chips for modern boards need to connect to multiple buses. I’m not saying it’s impossible, but it might make a board significant larger.
Yes, a hardware person I met through a family member said it’s hard and had they known how complex they would have chosen another dicipline
If that was an option I’d still be rocking my i7 920 with tri channel ram. That sucker could take anything I threw at it until the mobo started dying in 2021.
The only new mobos I’ve found with an lga1366 are dual socket server boards.
Had my i7-920 for 11 years. Started with 6 GB RAM and updated to 12 GB later. Went though 4 GPUs (just upgrading, none died). Changed from HDD to The machine was still going when I replaced it with a Ryzen a few years back.
This looks really cool, but I have immediate concerns about the lack of ZIF socket for something designed to take 30 year old chips.
The ZIF socket we know is patented. You would need to reinvent it to use it.
I would be surprised if there was anything on the board that didn’t have a patent covering it.
ATX itself is a joint standard specifically for creating specifications vendors need to meet, so that relaxes a lot of things.
You need to pay to use HDMI from what I understand, but not DisplayPort.
That being said, I would think that integrating components would mean you are buying parts that have been patented and not that you are leasing the patents themselves. That will increase the cost of the boards, as would be the modularity for users to swap parts themselves.
I do know that soldered connections are much more stable than pins to board. So many times I have fixed a computer by reseating memory or a card. Corrosion seems to build up between connections that is much less common than properly placed and soldered chips. I personally prefer my memory and CPU modular.
If we do use modular components, we should look into first making a board that ONLY has CPU and memory, placing the remainder of components in PCIe.
I don’t think populating the board with a ZIF instead of a pin hole socket would have changed anything.
Yeah, I talked my way into realizing the component is probably the patented part and it’s being resold.
The chip shortage makes people take desperate measures.
Could I make RAM modules using breadboards? 🤔
