Hello, I am interested in expanding my network, but I'm unsure whether mt76/ath11k and ath10k would cooperate effectively together using WDS/802.11s.
My current network consists of EA8500 and WRT1900AC. Using WDS between them works, but it's quite unstable due to mwlwifi driver. I also had the tri-band IPQ401x to experiment with, but the speeds were half of stock using WDS, thus it went back to stock. Because of this experience, I'm uncertain whether I can achieve a stable mesh with openwrt.
I've also read that ath11k has bugs atm and a bit unstable compared to mt76. Is it worth pursuing openwrt for mesh? If so, then which device is recommended? I'm just want some stability.
Just to give a data point as I don't think it altogether answers your question(s) - I use 3x RT3200's connected via WDS over two separate SSIDs (one WDS over 2.4 Ghz and another WDS over 5 GHz) and it's 100% stable.
What bandwidth are you getting with iperf3? Which topology are you running? I don't think im buying another QCA router although the EA8500 has been rock solid.
Circa 700-750Mbit/s at opposite ends of large house / between floors, but this is with stud walls / thin floors in modern internal construction.
Absolutely perfect for my needs - with these devices my WiFi signal is fine all over three floor house and surrounding garden.
The RT3200 is certainly stable and performant CPU-wise, but I've heard from @patrakov and read elsewhere that the WiFi on these devices isn't all that strong compared to other similar devices.
So if very strong WiFi range and performance is important there may be better alternatives.
As mentioned, a 3 pack of WSM20 devices can be yours for less than the cost of one equivalently spec router. I have no idea how they are making money on them, and with OpenWRT they are rock solid (months of uptime and 100s of GB routed).
The AX support on 2.4GHz may be helpful in giving you a longer range WDS backbone with 5GHz available for clients at either end.
These look really cool. I already have a Zyxel NR7101 and if seems very well designed. Do you have these and what's their WiFi range and performance like?
I bought a three pack and use two (gave the third to the in-laws).
The WiFi range is good - I get 3 bars of 2.4GHz on my iPhone at the bottom of the garden (20-25 metres) through a double skin brick wall. 5 GHz goes through single skin brick walls at 8-10 metres with 300-400 Mbps.
That price is the same as I paid from ebuyer.com. The three pack is a steal - I think some countries have it for €80!
I've recently bought a BananaPi BPI-R3, and I have set massive antennas on it. I can't see any commercial routers that fit perfectly when you want to dangle with antenna versions, so I clearly needed a development board to move straight to the top of signal power.
This said, RT3200 outperforms WSM20. I would like to see though, the test between WAX206, because I don't really have a RT3200 ATM. But WSM20 nowadays can cost 1/3 of the RT3200 so it's not a fair comparison.
How's your experience been with the BPI? I'm probably gonna get one too, cause I can add a dedicated wireless backhaul. @VA1DER which PCIE card would you recommend in that case?
At first I was super frustrated because I saw Verizon CR1000A and I saw, raw-power wise, it was demolishing, but now after I have it in my hands and the power of having a dev board in all sense, definitely blasts any other alternative I've found so far. Super-happy with my acquisition (although not cheap for me because they charged me an extra and unexpected $40 in customs VAT, 20% in my country)
RT3200 definitely has a faster CPU, but they both are “limited” by gigabit LAN/WAN, so the WiFi speed difference isn’t particularly meaningful.
Thanks to the updates in 23.05.x most 7621 devices (WSM20 included) now have full 2Gbps LAN/WAN routing capability if you enable acceleration. I’d say,therefore, that unless you’re looking to run VPN or SQM the extra performance is moot. If you do want those software features you could run the RT3200 as the router and still use WSM20s as dumb APs for better coverage at reduced cost.
In the end, one would have to do a 1:1 comparison, but...
The rt3200 has two 4x4 radios, compared to 2x2 (one 4x4 radio cut in half via DBDC), while you're right about the wired side being a bottleneck for 4x4, this should still improve beamforming (especially with multiple clients) and bringing the throughput over the range (more margin).
Yes, 802.11ax is an improvement for 2.4 GHz, but anything performance sensitive (especially the interconnecting wireless backhaul) 'should' use the 5 GHz band anyways.
That said, we're increasingly seeing filogic 830 devices on the market (and getting added in master) - which solve both 'issues'... Yes, they are more expensive (130+ EUR), yes most of them will not make it for 23.05~, but they are worth considering.
mt7621a+mt7915DBDC is still a pretty good budget option for wired backhauls (adding one more *wired* satellite might be better over all than the 'better' rt3200/ filogic options).
Because they get a free 4x4 802.11n radio for 2.4 GHz integrated into the mt7622bv SOC, they just have to add a(nother) 4x4 5 GHz radio (mt7915/ 802.11ax) to complete their product. They could add two and ignore the in-SOC 802.11n 2.4 GHz radio, but that would be 'wasteful', they could add one mt7915DBDC radio (same as done on mt7621a, which doesn't contain any radio hardware in the SOC) - but that would just get them on par with the cheap mt7621+mt7915DBDC devices (not matching their mid-tier target) - or they could switch to filogic 830, but that's reserved for the high-end devices (some of which may add 2.4+6 GHz DBDC configurations).
It's all a question of what the hardware (SOC) already gives you (for 'free') and what specs/ price point you want to target. For the designer/ manufacturer either would be possible, if they can sell it to the masses.
SOC integrated radios are cheaper, less components, fewer traces on the PCB, less layers on the PCB - but it limits the flexibility. …and Mediatek customers are used to retrofitting newer radios to older SOCs, as that speeds up their time to market and cost (the old SOC designs are cheaper, good enough for quick low-end offerings) - that often cuts into their mid- to high-end offerings (which may never show up on the market in numbers, as the old hybrid designs are already there, cheaper).
