> 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."
> 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.
> designed for users who have never configured a VLAN or written a firewall rule.
I always get the impression that when things are designed this way, you can't configure a VLAN or write a firewall rule, and so far I've never been proven wrong. :/
Their OpenWRT wiki page for installing on my router was a mess, but I got through it and took extensive notes about where the page was wrong or confusing. Then I asked for access to their wiki and was… ignored. After a week or so I forgot all the info and the notes started to look like gibberish.
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.
> we really need to see some benchmarks for things that matter
Honestly, we don't. We know it won't be competitive with the plethora of high performance ARM network SOCs found in commercial routers. If you use this with advanced features enabled (traffic shaping, packet inspection, etc.) on a fast uplink you will be CPU bound, and the CPU isn't fast. This shouldn't be a surprise to anyone that knows why this platform has any appeal.
You don't buy this expecting to max out your 10 Gbps fiber. There are other, valid reasons, but not that, and I'm glad it exists: one day, there will be RISC-V network SOCs that dominate benchmarks.
More open-source forks of OpenWRT and open-schematic router board designs are exactly what we need. It would further raise the cost of planting backdoors in routers at meaningful scale. We're currently too dependent on the OpenWRT project for router firmware. It's a high-payoff target for XZ Utils [0] type of multiyear infiltration by malicious actors.
The StartWrt port supposely adds some nice features, of which VPN chaining looks especially useful. And a better UI will make it more accessible. There are plenty of people out there who are willing to switch out their routers and chain VPNs to escape gov/ISP/big tech surveillance but don't have the technical means to do so. These are welcome improvements to reduce friction if they manage to pull it off.
The specs are not too bad for the price considering this is a startup project. It has 8 cores with per-core performance similar to Cortex-A55 + 4GB LPDDR4 + 16GB eMMC, which is better than most off-the-shelf routers. I wish they released the WIP schematics and code though, because there seems to be nothing at the moment.
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.
> 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.
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?
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
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.
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".
A backup WAN can be connected over USB, if it's a backup for graceful degradation when the primary high-speed WAN goes down. USB3 gives you a very decent speed, so if your WAN is not very fast (a typical 300-500 mbps home Internet connection), it can just be adequate.
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.
> there is no open-firmware option for modern WiFi from any manufacturer
I wonder if this could be changed, if enough people got together and had a WiFi chip fabbed, or paid a company to open their firmware? I'm guessing the bar is higher than that, because the WiFi trade assoc. probably mandates closed firmware. So you'd have to create a competing (but open) WiFi standard and probably have to lobby the FCC to let us use it.
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).
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.
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?
Yes, you have to decide in your threat model which is worse. There are people who’ve built entire systems on RISC-V FPGA soft cores like Bunnie Huang’s Precursor, but none fast enough to serve as a router.
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.
Working in this field - yeah don't underestimate what's available in "poor" countries. Most do better than poorly governed "rich" countries. And don't use it as an excuse to aim lower. CPE class devices do have a pretty long lifetime.
> 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.
This is pretty much what GL.iNet does. A nice slick interface for normal people, full OpenWRT nerd power a couple of clicks away for HN readers.
> designed for users who have never configured a VLAN or written a firewall rule.
I always get the impression that when things are designed this way, you can't configure a VLAN or write a firewall rule, and so far I've never been proven wrong. :/
Their OpenWRT wiki page for installing on my router was a mess, but I got through it and took extensive notes about where the page was wrong or confusing. Then I asked for access to their wiki and was… ignored. After a week or so I forgot all the info and the notes started to look like gibberish.
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.
> we really need to see some benchmarks for things that matter
Honestly, we don't. We know it won't be competitive with the plethora of high performance ARM network SOCs found in commercial routers. If you use this with advanced features enabled (traffic shaping, packet inspection, etc.) on a fast uplink you will be CPU bound, and the CPU isn't fast. This shouldn't be a surprise to anyone that knows why this platform has any appeal.
You don't buy this expecting to max out your 10 Gbps fiber. There are other, valid reasons, but not that, and I'm glad it exists: one day, there will be RISC-V network SOCs that dominate benchmarks.
As someone not involved in this space I assumed there was specialist hardware. My pi Pico can do fancy things with DMA etc, without the CPU at all.
So why isn't there this kind of stuff in routers?
There is, and it is. Just not in this specific SoC, originally intended for general-purpose computing.
More open-source forks of OpenWRT and open-schematic router board designs are exactly what we need. It would further raise the cost of planting backdoors in routers at meaningful scale. We're currently too dependent on the OpenWRT project for router firmware. It's a high-payoff target for XZ Utils [0] type of multiyear infiltration by malicious actors.
[0] https://en.wikipedia.org/wiki/XZ_Utils_backdoor
The StartWrt port supposely adds some nice features, of which VPN chaining looks especially useful. And a better UI will make it more accessible. There are plenty of people out there who are willing to switch out their routers and chain VPNs to escape gov/ISP/big tech surveillance but don't have the technical means to do so. These are welcome improvements to reduce friction if they manage to pull it off.
The specs are not too bad for the price considering this is a startup project. It has 8 cores with per-core performance similar to Cortex-A55 + 4GB LPDDR4 + 16GB eMMC, which is better than most off-the-shelf routers. I wish they released the WIP schematics and code though, because there seems to be nothing at the moment.
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.
You’re not running a BSD on an embedded device with full driver support any time soon. Linux won this space.
Under the hood, the StartWRT UI is just another OpenWRT package, and it plays nicely with luci.
> 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".
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/
250k for openwrt based risc v router? Maybe need do more work such as using vyos + fdio/vpp
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 ?
The case on there is also the same, the vent holes and layout is identical (sans the other USB and HDMI)
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
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.
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".
The page linked above contains links to their bootloader and Linux kernel tree (6.1 apparently), so chances are rather low.
Love this in theory, but can't do it with only 2 ports. I need backup WAN.
A backup WAN can be connected over USB, if it's a backup for graceful degradation when the primary high-speed WAN goes down. USB3 gives you a very decent speed, so if your WAN is not very fast (a typical 300-500 mbps home Internet connection), it can just be adequate.
I need a dmz. ;)
Looks cool. I'd hoped for usb-c for power at least. Trying to get rid of usb-a.
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.
https://project.turris.cz/en/hardware-documentation.html
Vanilla OpenWRT runs Omnia (Marvell Armada). OpenWRT support for IPQ9574 is WIP. Omnia NG is IPQ9574.
https://openwrt.org/toh/turris/turris_omnia
Turris has its own OpenWRT warapper, but you can just wipe it and install the stock OpenWRT.
> there is no open-firmware option for modern WiFi from any manufacturer
I wonder if this could be changed, if enough people got together and had a WiFi chip fabbed, or paid a company to open their firmware? I'm guessing the bar is higher than that, because the WiFi trade assoc. probably mandates closed firmware. So you'd have to create a competing (but open) WiFi standard and probably have to lobby the FCC to let us use it.
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.
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.
Not all packets are 1500 bytes.
> 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?
Yes, you have to decide in your threat model which is worse. There are people who’ve built entire systems on RISC-V FPGA soft cores like Bunnie Huang’s Precursor, but none fast enough to serve as a router.
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.
Exactly - seems like the only big thing going for it
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.
It still means you're permanently hassled with sticking a switch next to it.
Yes it's not a requirement per se to include an ethernet switch chip on the board. But at a $300 price tag I'll say it does become a failing.
VLANs would appear to defeat the ease of use aspect here. Plus that means you need managed switches, and know how to use them.
> 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.
> No, it isn't. Not even by far.
It is. You typically either have only cellular connectivity or you have fiber, with very little in-between. And fiber provides 1/10G capability.
Is it _used_ universally? No. But the capability is there.
> Anything higher than 1gbps would ramp up the cost today.
This is not going to be a cheap device _anyway_.
Working in this field - yeah don't underestimate what's available in "poor" countries. Most do better than poorly governed "rich" countries. And don't use it as an excuse to aim lower. CPE class devices do have a pretty long lifetime.
> 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.