I urge everyone to go and read the mailing list post that sets out what the device should be, what it shouldn't be, and what it absolutely won't be.
https://lists.openwrt.org/pipermail/openwrt-adm/2024-January/002479.html
9 Likes
Better than Banana pi r- 4?
I think we are missing one important fact- that this device will be unbrickable.
This will allow people to play around with OpenWrt and will not worry if the need to throw their router in the recycle bin. This will build next gen of developers, who will not be afraid ( or at least know the consequence of doing extraordinal things) to make builds for more routers.
A high degree of safety against bricking is a software concept. Other devices are safe against bricking and easy to unbrick, too.
Before software details might be discussed the hardware needs to fulfill the expectations.
If the goal is to provide only another single board computer as development platform where you need to care for yourself about cooling, case, power supply, antennas and legal stuff I don’t think the target group will be large.
There are a lot of competing OpenWrt targets that are at least complete hardware solutions, ready to work as a router: with cooling, with case, with power supply, with needed legal certifications, with antennas and everything ready to run out of the box.
OpenWrt One should be a complete ready to run wireless router solution out of the box. Like the GL-MT6000 is now, too. Buy it, open the box and be ready to play with software, don’t need to care for hardware details because it’s a complete hardware router solution and not just a single board computer circuit board.
If the goal should be to have another single board computer circuit board alone as development platform, then please, speak it out. To think about software details is nice but we are discussing a piece of future hardware.
Cool. Now I'm use Banana Pi R1.
I think that requires:
somehow? I would agree that this seems highly desirable, but I actually expect this to happen, the rest I agree would be nice (I will however not care and happily assemble a One myself, as I think the specs are 'good enough' for quite a number of use-cases, and I am getting fonder and fonder of 'good enough' compared to 'perfect' as I get older 
)
2 Likes
Yes, I think so. To get for example the FCC certification to sell the device in the USA I assume it needs to be emissions tested with antennas and with WiFi operating. The certificate should only be valid for the tested antenna configuration.
Any attempt on OSH certification? If the SOC pinout cannot be shared especiallt for the WIFI secret source a SOM approach would be the best choice instead of having the half cooked schematics and no PCB design...
Hi all.
I'm sticking with my Linksys WRT3200ACM, but I'm willing to upgrade to a 10GBe router as far as 10Gb FTTH internet connections are finally available in my city.
The replaceable Nvme ssd is a great idea.
I'd like more ram to be able to directly run more vms or dockers.
I guess as soon as workstation level arm chipsets for PCs (like the stunning Apple Silicon Mx) will be available it won't be a mirage anymore
So I hope this project will turn into a success and an high-end device will follow
BEST LUCK!!!!
1 Like
Need two 2.5Gbps ports.
Maybe offer 2 different models???
1 - with Wifi
2 - without Wifi
For me, I think 2 ports (2.5Gbps) is a must-have. Otherwise, the router cannot forward the full bandwidth of 2.5Gbps, and will always be capped at 1 Gbps, why?
Yes we can always add another 2.5Gbps switch behind the router, but if the router cannot pass 2.5Gbps to the switch, that would be a shame.
I am currently using OpenWrt as my main router, but I don't use it as my Wifi router. I have my my own Wireless Access Points (wifi 6 mesh). If I ever need to upgrade it to Wifi 7 one day, I don't necessarily want to replace the OpenWrt One device. The router should remain as wired-only. Instead of offerring Wifi capability, I would like to see the router having wilder compatibility such as more Ethernet ports, USB 3 ports, USB-C ports, MoCA, etc.
I don't really care about the price tag. If the router is meant to be an example of good OpenWrt router, whether it is $100 or $150, I will buy it. To make the market buyers a fair chance, I think OpenWrt One could be released in 2 models, and I'd like to buy the no-wifi model.
3 Likes
@NPeca75 explained this quite cogently, if you want to serve 1 Gbps lan clients on a secondary switch 'full duplex' you need >= 2x1 Gbps between router and switch. Plus in addition to the wan traffic WLAN 2 LAN traffic will also traverse the LAN port. The port asymmetry might not have been for the above rationale (I think it is simpy based on the SoC's capabilities), but thee above still holds...
Will this device do 500mbps sqm cake?
Sounds interesting
It depends. But you likely mean will it be able to reliably traffic shape at 500/500 Mbps, and that will need to be seen. I guess 500/500 might well be within reach, 1000/1000 is considerably less likely.
2 Likes
I remember that hw offload would implement some kind of hardware QOS implementation. It would take much more cpu usage to use software qos
Dual core A53s at 1.3GHz? Doubtful it'll hit 500 Mbps on SQM cake, but it should get close maybe with fq_codel and some affinity tweaks.
Now if multi-CPU DSA ever happens that's another story
Its gonna route all networking to it's PPE and NPU, so CPU won't be used much
Dual core A53 with 1.3 GHz is what MT7622 has. Enough for 500 Mbps with SQM:
more performance numbers from MT7622:
I would expect similar SQM performance for other A53 dual core with 1.3 GHz.
Somehow I doubt that since I barely got that out of my MediaTek Filogic 830 based GL-MT6000 using SQM Cake. It took dropping it to fq_codel and messing with affinity to hit 800 Mbps. I'd like to see a bufferbloat test showing that speed along with CPU usage and +0/+0. Hopefully it does it just seems questionable.
(Meanwhile my old WRT32X had no problem at 600 Mbps with SQM cake but that's mvebu).
Please look into the linked thread, many reports on cpu usage with SQM are given, even screenshots. No need to doubt.
400 Mbps with cake should be a safe bet for dual core A53. More with fq_codel.
Interesting looking at the thread. Not a bad SoC choice after all for the given price target.
edit: there are definitely inconsistencies in their tests though. I've seen similar with my Filogic based. I have reason to believe MediaTek chips are designed for HFO/SFO + WED and don't work perfectly with SQM, but not sure why. My WRT32X was always perfect +0/+0 with Cake, these aren't. It's a minor concern but there is some issue somewhere. Performance is great with HFO/SFO+WED though so that's how I run mine.