RISC-V Router
router.start9.com82 points by janandonly 7 hours ago
82 points by janandonly 7 hours ago
> Built on a RISC-V processor with an open-source boot stack and operating system, it is the most open router on the market […]
No it's not [cont'd]
> with a fully open-source boot stack (OpenSBI, U-Boot), open-source Linux kernel, and published board schematics.
You can all get all that for both OpenWrt One and Turris. Possibly more, they go beyond schematics on HW design. And that CPU is no more "open" than the libre end of ARM chips elsewhere.
https://project.turris.cz/en/hardware-documentation.html - that's the bar. CERN OHL (or equiv) with not only schematics but gerbers.
And, y'know, I rather get OpenWrt unforked from the OpenWrt people. Even the Turris people are burdened by OpenWrt "re-maintenance".
> StartWRT: Start9's fork of OpenWrt, including a modern GUI, that reimagines the router experience from first principles.
I wish them the best of luck with their hardware venture, but a custom fork of OpenWRT is not what I'd want for a router from a small startup.
I can't even begin to count how many startups have done crowdfunding projects for new hardware and tried to get too custom with the software stack before the company went under.
Others already covered the high price for the specs, but we really need to see some benchmarks for things that matter: Routing throughput, VPN throughput, and other real numbers. Faster ports aren't helpful if the CPU can't process packets fast enough.
I also wonder why they wouldn't work with upstream in improving the existing GUI (or upstreaming their improvements), instead of putting the burden of a fork upon themselves.
Working with upstream is most convenient for their users, for them, and for the ecosystem as a whole.
A basic Google search leads me to this article [0].
> On March 27, 2026, Start9 CEO Matt Hill hosted a private unveiling of StartOS 0.4.0, the next major version of the operating system that powers the Start9 Server One. During that same session, Hill also gave viewers a first look at StartWrt, the router’s dedicated operating system. StartWrt is Start9’s fork of OpenWrt with a modern GUI that reimagines the router experience from first principles. The interface is sleek, modern, and a clear departure from the technical admin panels that define most open source router software today.
> Where OpenWrt’s default LuCI interface is functional but technical, StartWrt presented a clean, modern interface designed for users who have never configured a VLAN or written a firewall rule.
When you consider the circumstances a fork is the only thing here that makes sense. You can't just open a pull request to OpenWRT where you are like "Here is our purpose built simplified GUI we designed for our router, please merge."
[0] https://www.solosatoshi.com/start9-announces-fully-open-sour...
> When you consider the circumstances a fork is the only thing here that makes sense.
No, because a fork and an overlay are not the same thing. Getting your custom frontend has nothing to do with sharing the maintenance burden on all the grit behind it.
The gui of openwrt is not great. It might be better if you already have lots of experience with linux networking and openwrt specific command line configuration. If not it seems like a mess, very vague and overlapping controls without much explanation. DDWRT and Tomato are much better although openwrt might be more powerful without resorting to straight firewall and routing rules through text.
Under the hood, the StartWRT UI is just another OpenWRT package, and it plays nicely with luci.
And at that point why not OPNSense? OpenWRT for me is what I would run on crappy BestBuy routers that can’t run a proper router OS. OPNSense is 100% amazing.
250k for openwrt based risc v router? Maybe need do more work such as using vyos + fdio/vpp
Love this in theory, but can't do it with only 2 ports. I need backup WAN.
Is Start9 a well known company? The page by itself seems indistinguishable from a scam, but maybe they have a reputation that justifies their asking for $250,000?
It is not well know but I heard good thing about what they do.
It is very similar to Umbrel [0].
- [0] https://umbrel.com/
BananaPi already sells boards with same CPU for around $100 with maybe $15-20 extra for case
https://docs.banana-pi.org/en/BPI-F3/BananaPi_BPI-F3
Is it doing anything different ? I assume at least made in US so it can be sold as router and not dev board ?
Are the banana pi boards able to run a mainline kernel or close to it? I have a memory of getting real close to buying one of those, and then reading a comment on HN about having to run their Frankenstein setup
Depends on the board.
Btw, I don't see anything about mainline in TFA, did I miss that?
FWIW there is also "OpenWrt mainline" and "Linux mainline"; OpenWrt carries a whole bunch of things on top of Linus' tree but I'd still call that "mainline".
Given the similarities in port layout (just missing a HDMI and USB3 header), and that the case is nearly identical, I would guess that this router probably is a custom run of the exact same BananaPi board without those headers. Both also use MiniPCIe in 2026, which is a bit of an odd decision.
The page linked above contains links to their bootloader and Linux kernel tree (6.1 apparently), so chances are rather low.
Since this has a foreign-made processor and WiFi module, would this be blocked by the Trump FCC's foreign-made router ban?
Turris Omnia NG is also "open source" and has 2x 10 Gbps SFP+ and 4x 2.5 Gbps ethernet ports. StartWRT and Turris OS are both forks of OpenWRT, which is kind of annoying. The Turris project has been around a long time and has an active community.
Quick glance of their page only mentions Turris OS being built on open source. If I can't blow away Turris OS and install whatever, then it's not open and uninteresting.
You can install whatever you want on Turris devices. And you get schematics and gerber files.
Vanilla OpenWRT runs Omnia (Marvell Armada). OpenWRT support for IPQ9574 is WIP. Omnia NG is IPQ9574.
Turris has its own OpenWRT warapper, but you can just wipe it and install the stock OpenWRT.
Single WAN, Single LAN, is not actually what I would (or do) use for "home-based self-hosting". That hosted stuff gets its own network.
that is what vlans are for. but having only gigabit ports is limiting here.
I helped a bit to develop this UI myself. Support for vlans was baked into it from day 1. The idea being good admin/guest/iot/hosted/etc separation without extra access points.
RISC-V is quite wimpy this far, so it’s not even clear if it can saturate a gigabit with features turned on. The one benefit is that it doesn’t have Intel IME/AMT, AMD PSP or ARM TrustZone backdoors built-in, but I would be extremely surprised if the Chinese SpaceMiT CPU didn’t have Chinese backdoors of its own.
> it’s not even clear if it can saturate a gigabit
If that's the case then it's not the CPU's fault. I can't open the linked site but assuming it's really the same as a BPI-F3 i.e. a SpacemiT K1 chip, that can do 2.8 GB/sec on large RAM to RAM memcpy using a CPU core i.e. 44 Gbps total, 22 Gbps each read and write. Plus I assume it's got DMA so no need to involve the CPU anyway.
Here is a test I ran in April 2025 on a Sipeed LicheePi 3A same chip).
https://hoult.org/K1_memcpy.txt
> RISC-V is quite wimpy this far
The new K3 chip from the same manufacturer does 8.7 GB/s RAM to RAM memcpy using a dual issue in-order A100 ("AI") core, just over 3x faster.
Sure this pales in comparison to recent Apple / Intel / AMD but it's a lot faster than home networking.
Although your benchmark is interesting, I don't think it's very relevant here. In my experience, you'll saturate the CPU through packet decoding, routing, and firewalling long before memory becomes a bottleneck.
That's why all network SoCs have hardware to accelerate such thing, otherwise in software alone they can barely handle simple routing at a few hundred mbps.
That chip doesn't seem to have that: https://cdn-resource.spacemit.com/file/chip/K1/K1_datasheet_...
1 Gb/s is only ~100,000 packets/s at standard MTU. You literally get 10 us/packet which is a eternity. Normal fast-path router operation only really needs to consider the header of <100 bytes/packet, so you are getting ~100 ns of compute per byte of considered data and on even a 1 Ghz processor you are getting over 100 instructions per byte of considered data. Failure to achieve a measly 1 Gb/s really says more about those software implementations than it says anything about the impossibility or difficulty of the problem.
> The one benefit is that it doesn’t have Intel IME/AMT, AMD PSP or ARM TrustZone backdoors built-in, but I would be extremely surprised if the Chinese SpaceMiT CPU didn’t have Chinese backdoors of its own.
That seems worth paying for. How could china hurt me more than my own government?
Yep. It's crazy how effective the US Gov has made it seem like China are the bad guys, when it was US/Israel all along.
VLANs would appear to defeat the ease of use aspect here. Plus that means you need managed switches, and know how to use them.
> Router
> Ethernet: 1 WAN Gb, 1 LAN Gb
> $250000
Awesome.
Cost is 300$ not 25k (for the end user) it looks like
But the fundraising goal is.
Yes, but you presented it like the product cost would be $250,000. You also failed to include the main selling point... It's a RISC-V chip.
> Ethernet: 1 WAN Gb, 1 LAN Gb
Really? In 2026? Pass.
It needs to be _at_ _least_ two SFP+.
Note that most people, worldwide, only have <1gbps internet access if at all.
Sigh. 1gbps is widely available, even in relatively poor countries.
And if you're making a _new_ device that should last for 5-10 years, it's just stupid to use technology that is getting obsoleted even now.
>Sigh. 1gbps is widely available, even in relatively poor countries.
No, it isn't. Not even by far.
>And if you're making a _new_ device that should last for 5-10 years, it's just stupid to use technology that is getting obsoleted even now.
Anything higher than 1gbps would ramp up the cost today.