MT3000 price dropped to 79,73€ on Amazon DE (activate 40€ coupon)
1 Like
This is a big ask for even reasonably experienced people. There are a hundred things that can go wrong, from power issues to accidentally flashing the wrong image. I would say this calls for software improvements. Such as a client-side software that can identify the router reliably, download firmware, and do the whole flashing. In the absence of such a thing, easy recovery is equally important.
Even updates are in a similarly unfriendly state right now. I updated my two access points today, and I had to find the sysupgrade images for the right routers manually and then flash them. I was holding my breath until the routers rebooted, wondering if I had the right images. It is not too much to ask that the existing OpenWRT image must know which successor image to download and flash without involving me. Like every normal OEM firmware in the world.
I mean, a noob can hopefully get someone to help flash OpenWRT the first time, but updating should not require that.
1 Like
I completely disagree - this is a base level functionality. If we cannot guarantee that an Official firmware version installs correctly on supported hardware, it implies:
- This device/firmware is not suitable for newcomers
- There is likely something severely broken
Further thoughts: We do have software solutions that address your suggestions:
- Firmware Selector for choosing the proper firmware for a first-time user
- Attended Sysupgrade for upgrades.
We just need to screw up the courage to make these the default recommendations everywhere.
2 Likes
- Supports official STABLE OpenWrt 23.05.x. No snapshot builds
- Clear installation instructions without warnings in the Table of Hardware .(https://openwrt.org/toh/tp-link/archer_ax23_v1)
- Mass market device with case, power supply, and "standard" network connections: a WAN port, one or more LAN ports, Wi-Fi. No assembly required
- Price point below US$100 or equivalent in your currency. It must be available today (Amazon, eBay, etc.) not promised at some indeterminate date. (https://www.amazon.co.uk/TP-Link-OneMeshTM-HomeShield-Archer-AX23/dp/B09VPRWGR7/)
- Able to handle 300mbps with SQM enabled. Provide link to citation or personal experience
So we are in MT7621 territory as a minimum, so the TP-Link AX23 fits your criteria, but only the v1, as no one has confirmed if the v1.20 is compatible.
MR70X v1 would be a contender but has no toh or techdata and only exists on here:
The GL-MT3000 https://openwrt.org/toh/gl.inet/gl-mt3000 with MT7981 Filogic 820, 256 MiB flash, 512 MiB memory, 2x2 WiFi, dual core ARMv8 SoC and one 2.5 GbE uplink can be a good fit where lower costs and a smaller form factor is needed. This model is not shocking me at all. OpenWrt One will be nearly the same hardware as GL-MT3000.
My predecessor generation E8450 with dual core ARMv8, 128 MiB flash and 512 MiB memory is not shocking me, it’s still comfortable and only misses 2.4 GHz AX.
1 Like
Point take. At $80, the GL-MT3000 isn't bad for a tiny space or portable niche need.
However, with only two ports and shrunk 2x2 antennas I still wouldn't recommend it to a newcomer for a general purpose gateway all-in-one WiFi router suitable for covering a home.
4 Likes
I suggest adding a note to this recommendation up top that this device will only handle ~100 Mbps CAKE SQM, lest someone be misled it can handle ~300 Mbps SQM.
This is a MT7621AT device with dual core / quad thread 880 MHz MIPS. I used to run an ER-X gateway on OpenWrt with that SoC and it handled ~100 Mbps CAKE and ~200 Mbps fq_codel / simple SQM on a good day, and that was without needing to also support WiFi.
The E8450/RT3200 and WAX206 use the same MT7622 SoC. Seems like "the power to do SQM" should be about the same between the two?
I think it's worth remembering the volume of people for whom "home" is an apartment on a single floor, often with 1 bedroom or less. I wouldn't call that kind of space "niche", I'd probably call it "the norm" in many places.
And for such a home, the MT3000 could be genuinely excellent. Lack of ports is solved with a $20 (or less) switch and there's a good chance the 2x2 WiFi would be no impediment in practice.
For a large proportion of "casual" users, any off-the-shelf router is already overkill in terms of ports and WiFi. I know tons of families who don't have a single local client hardwired to their router.
7 Likes
I like your listing. +1 for the GL-MT6000 on the honorable mentions. It has been on sale for a while at $135 from Amazon (just picked up one with the 15% coupon). I just picked one up and decided on it over all of the other options on your listing. It is already included on 23.05 snapshot, 23.05.3 should be tagged in the very near future which would make this router truly turn key for new OpenWrt users (somebody tag 23.05, this router is great!). Firmware selector makes this router easily accessible for those that are new to OpenWrt and want luci included.
https://www.amazon.com/GL-iNet-GL-MT6000-Multi-Gig-Connectivity-WireGuard/dp/B0CP7S3117
Coming from a R7800 recently the GL-MT6000 is a big leap forward even from my NSS builds. $135 is a stone throw away from your criteria. I’d recommend the GL-MT6000 over the R7800 for a 2024 router:
2 Likes
Your implicit suspicion of my bias is spot on. You and odrt are 100% right to reel me in on it. Thank you for pointing this out.
2 Likes
Have you tried mt7621 cake and SQM capabilities since the 2 Gbps WAN/LAN NAT Routing was added in April 2023?
The TP-Link AX23 recommended in this thread has a relatively low capability dual core MT7621AT MIPS CPU. This is why it is so affordable. The MT7621AT does have some some tricks up its sleeve, but they don't help SQM performance, because SQM needs raw CPU power.
Each of the two cores in the TP-Link AX23 CPU is equivalent to the single core MIPS CPU in the Archer C7. The Archer C7 CPU is clocked a little slower, but its CPU is a little more efficient by nearly the same factor.
Symmetric 1 Gbps routing in two directions (the 2 Gbps WAN/LAN NAT Routing capability) will not help running out of CPU at 100 Mbps with CAKE (or 200 Mbps with fq_codel/simple). It should be noted that PPPoE ISP service and/or WiFi support, which load the CPU, will degrade this performance further.
The only reason the MT7621AT can route 1 Gbps in any direction to begin with is that it (and OpenWrt) supports hardware flow offload to bypass the CPU. This is a great trick, but SQM needs to go through the CPU, so SQM and benefiting from flow offload (hardware or software) are mutually exclusive. You have to choose which one you want.
The TP-Link AX23 is not the router for a gamer newcomer interested in OpenWrt for CAKE SQM if they want to make use of more than 100 Mbps of ISP service. Incidentally, this demographic covers a LOT of OpenWrt newcomers.
Ignore SQM (and anything else requiring CPU power like running OpenVPN or Wireguard on the router), and suddenly the TP-Link AX23 becomes a fantastic and affordable Gbps capable OpenWrt WiFi 6 gateway router thanks to hardware flow offloading on the MT7621AT. So it does have its place.
3 Likes
US Amazon is now selling MT3000 at $79, pretty good price now!
1 Like
I'm sorry if this crosses into "off topic" territory, but I do think it's relevant to the discussion as an important consideration for anyone "new" to OpenWrt:
The overall level of documentation and forum support for accomplishing configuration tasks in LuCI (i.e. the GUI that one can safely expect any "newcomer" to see first, be comfortable with, and rely on) is still substantially lacking, unfortunately.
I fully understand that it's a time-consuming chore to update existing documentation (and keep it updated as LuCI changes), and less convenient for the forum gurus who read configs like an operator reads the Matrix, but this will continue to be a significant barrier to entry regardless of how ideal your chosen device is for installing OpenWrt in the first place.
IF there's a desire for wider adoption, I think this problem should be a primary focus for the community. Maybe a strong push to maintain at least a relatively comprehensive LuCI-focused documentation package for each "full" stable release, and better forum support for accomplishing tasks in LuCI.
2 Likes
Well, are you volunteering? Or are you trying to assign tasks to others here?
The former will be quite welcome, the latter IMHO considerably less.
Not judging whether that is good or bad, just observing that in FOSS one is the scheduler only of one's own involvement/effort. Anything that is contingent upon others doing tedious work will not work out.
3 Likes
Well... as an outgrowth of this particular discussion, I have been screwing up my courage to produce a "Quick, Quickstart" document, because the current document runs to four (or a dozen) pages, depending on how you count.
I envision something that's one page, and just says, "Do this, you'll see that; then do this, etc..."
I'm up to my elbows in another project at the moment, so if anyone wants to snatch the reins - go for it! Thanks.
I would be very happy to assist but as a "newcomer" myself it's hard to see how much I can help. The amount of stuff that I know and could document is extremely limited.
It's possible that I could help someone else with extensive knowledge to translate that into helpful documentation, so please reach out if that's worth pursuing.
Setting that aside, this conceit in open source development that things aren't worth doing or considering unless the suggester can do them on their own is not helpful. It's just as unreasonable to expect that any random user of the product can make a technical contribution as it is for the user to make demands, except in this case, there was no demand. Simply an observation and suggestion.
1 Like
Given that background, you are the most valuable resource we have for a quickstart guide. Anything a deeply experienced user writes is almost certain to skip some tiny but crucial step that's "obvious" or something on that order, no matter how much effort is put into reviewing it. Only new eyes can see these things clearly and call them out.
12 Likes