Community hardware project?

I didn't see anything on the forum but figured it was worth asking. Have there been any attempts at a community hardware project for OpenWRT? Something built with OpenWRT in mind from the start? It seems like every time I go to look for a new device to run this on, the current models aren't yet supported, and the older ones aren't available for purchase, and there's a limited pool of what's available after that to choose from.

I'd say this very much depends on where you live.

The BPi-R2 is well supported, no?

BPI-R2 isn't listed in the OpenWRT table of hardware, but the BPI-R64 is, and the BPI-R3 is only supported under snapshot so far. I wasn't familiar with this model before now, looks interesting.

OpenWrt will not run on anything you can cobble together with your grandkid on a rainy afternoon, using an old gutter soldering iron and a handful of through-hole components. So a 'community hardware project' in the sense that anyone can do it easily, with interchangeable components, within a reasonable amount of time and 'on the cheap' is not going to be possible; and we aren't even talking about FCC/ CE (ETSI)/ MKK certification yet.

Any kind of community project would be in the form of a (very-) small 'c', as it would inevitably have to be a complete device, designed, manufactured and commercially distributed in one hand (yes, you may call it crowdfunding, but really, it's not). But even if you leave the immediate commercial aspects aside (distribution, shipping, customs in different countries, warranty replacements, …), you are up against a big problem of the economies of scale. Your traditional commercial vendors aren't building a few hundreds- or maybe even thousands of devices, they don't even start production below tens-, hundreds- or millions of devices and will get much different conditions (and vendor-/ source access) - and once their device hits the shelves, they're already designing and producing the next 3 successor generations in parallel. Meaning your small-batch design will be more expensive (than even x86_64 + and OpenWrt compatible AP and switch) than your competition, it will also age quickly - without you having much of an answer to this. So the only chance to get a foot into the door at all, would be to massively over-spec the design, allowing you to (maaaybe-) upgrade components in the future, but that raises the price point even more, losing more potential customers.

Just as a though experiment, look back 2-3 years -even totally ignoring lead times for design and manufacturing (1-2 years on top)-, can you imagine any SOC (aside from x86_64) picked back then that could be (reasonably-) updated to wifi6, wifi6e, wifi7(!) and 2.5GBASE-T? If you want a more practical comparison, compare the Linksys WRT3200ACM with the Turris Omnia, each at the time of their market introduction - check original prices and how they stack up against each other today - and tomorrow (keep in mind that cz_nic is a bit bigger and longer in the business than your typical community crowdfunding project); for bonus points, look at the Turris Mox (and how that will stack up in terms of upgrades to wifi6e, wifi7 or 2.5GBASE-T, 5GBASE-T, 10GBASE-T).

Beyond that, look at the (fully commercially designed-, produced- and marketed) BPi-R3 and the BPi-R4, check how much it will cost you to make it a full product (simple but robust case, PSU, pigtails, antennas, including shipping, customs and taxes, etc.) out of it and reconsider how the BPi-R3 would deal with wifi7 in ~a year's time.

2 Likes

Just wanted to add that the turris omnia was a pretty decent design (arm a9s are pretty capable cpus witout a noticeable glass-jaw) and it offers some expanfability, e.g. cz.nic offered wifi6 replacement radios. But it certainly was/is not cheap.
Sidenote especially wifi routers but mostly electronics in general require certification that brings with its own cost, that can be ignored when spread over thoudands of devices but will become noticeable for smaller 'artinesal' product runs, but that is not optional in many jurisdictions.

That's why I took it as example, as they went that route of massively over-spec'ing the device for a device designed in 2016 - with the corresponding price tag.

Retrofitting wifi6 is possible, but wasn't easy (it took quite a few trials to find wifi6 cards that didn't break heat-dissipation and (out of spec-) power delivery options), wifi6e only in a limited way (heat&power, antenna mounting and coexistence, performance for HE320…); wifi7 is probably unlikely (aside from DBDC solutions with limited rx/tx chains). The wired backbone is limited to 1GBASE-T and a 1GBit/s SFP port (this is slowly becoming a bottleneck for enthusiasts).
Compared to the WRT3200ACM, this clearly is a winner (for the non-mwlwifi WLAN alone, even though the situation didn't look quite as bleak back then in 2016 (not great, never great, but not as dire)) - compared to an ipq806x device for a third of the price (and lesser specs), questionable, the cheaper entry allows upgrading in quicker succession while still ending up cheaper.

The Turris Mox (aside from not being supported by OpenWrt at all) doesn't really break these limits (1GBASE-T etc.) either and the modularity is kind of a mixed bag (considering availability, pricing and what might realistically make sense to actually purchase).

1 Like

All true. Personally I am quite happy with my omnia, but:
a) I am still on a 100/40 link
b) have no issue running my link well below its physical limit (before the omnia, my wndr3700v2 only allowed ~49/29 shaper rate on the same link)*
c) value the automatic updates from a source I happen to trust

Yet in 2023 I would not recommend it as a generic router for everybody (it is getting long in the tooth and you get more capable devices for considerably less money), but if automatic updates are desired it still is an option.

About the mox I concurr, really nice experiment with modularity, which IMHO showed it is to early for modularity in home routers... (link speeds and CPU demands are still raising significantly, so a router's CPU can not yet 'last' long enough to make replacing the other peripherals an attractive offer).
Then again, IMHO, we are in the territory of dimishing returns in regards to internet access rates, a gigabit link is going to feel considerably more capable than a 0.1 Gbps link compared to the perceived improvement from going from 1 to 10 Gbps... (I will not rule out that we will see higher rate applications in the future that will require higher access rates for basic functionality, but I am not holding my breath yet). So the mox might already be 'good enough', and I am just a cpu-snob for rejecting its a53 cores as too unbalanced/weak.

*) For my usage 49/29 with sqm was simply more usable than 100/40 without sqm, I am not saying that everybody needs to agree to that policy.

Not really... If you have any experience with computer building, upgrading or repair at all, putting together an OpenWrt router on x86 is so easy that no community is required. You can do it all on your own with Google as your co-pilot, using only commodity components, including older ones. Or use an existing commercial board or device (anyone remember Alix and APU from PC Engines?) as a base and work up from it.

Right now, my primary router is a Sophos SG 115 Rev 1 from 2015 bought off eBay in the fall of 2022 for a whopping $50 (factory specs are Atom E3827, 4 GB of removable RAM with processor supporting up to 8 GB, 320 GB SATA hard drive, four-port Intel i211 NIC). Rev 1 had an interesting quirk: wired-only and wireless routers had identical motherboards, so there was a socket for a Wi-Fi card on the wired-only router (with Rev 2, they stopped doing it; the boards are still the same, but wired-only routers no longer had the seats for Wi-Fi cards, only soldering points for the seats). Anyway, I populated the socket with something or other from Atheros (the seat would accept either half-size of full-size card), replaced the stock hard drive with a 16 GB SATA SSD, and presto! — a Gigabit OpenWrt router, community or no community...