Hi,
I've got ubiquiti unifi ac lite running OpenWrt 23.05.5
Main router is tplink TL-WR1043ND v2 running OpenWrt 23.05.0.
Tplink gets online via some chinese ZLT 5G LTE router connected to its wan port (gigabit).
And one 5 port switch. (where I have a samba server with SSD disks).
All cabling is 1Gbit. (checked all ports on all devices)
When I connect with UTP either to the tplink or the switch, and copy something over samba I get wire speed (100MB/s)
When I connect via wifi, i get 21MB/s speed copying either over samba or scp.
That kinda sux, but what is expected from this dumb AP? what speeds do you get in real life? if anyone using this AP.
I read through this forum...found this remark: At 80mhz the interference is too great in the city for a reliable connection. 80mhz is a massive bw @ 5ghz
so, when I set the AC wifi to 80Mhz width i get 860mbit reported wifi speed on windows client, but samba copying is 14MB/s
When I reduce width to 40Mhz, i get the above mentioned 21MB/s.
My current internet connection is 160Mbit even on wire, so I'm good for now.
It's 4G+, but the ISP can upgrade my area to 5G 500Mbit anytime, and then what, according to those speeds on my WIFI -> LAN, I can get max ~200Mbit via this WIFI AP.
Now, I may survive slow access to my samba share, but getting 200Mbit instead of 500Mbit internet access sounds like an inconvenient waste of bandwith. Even though i have more then one device so they can share and use up the internet bw, I don't expect one single laptop to use 500mbit, but still, it's a matter of principle
Any advice on optimizing this wifi performance, or I have to change the device?
That is roughly the maximum you can expect from QCA9558, with almost no margins left. This hardware is old and slow, it is part of your bottleneck (and its wireless performance is particularly sub-par).
It is not at all clear to me if your wireless clients connect to the tl-wr1043ndv2 or the unifi ac lite here, nor if they connect via 2.4 GHz or 5 GHz. Whenever your clients pick 2.4 GHz (and the tl-wr1043ndv2 can only do 2.4 GHz), performance will be slow by design.
What's the model of the network switch?
It reads like you have a fast Ethernet limitation somewhere in your setup.
Is the AC lite powered by PoE switch or PoE injector?
Hi, sorry, I forgot to mention, tplink has wifi disabled. Its doing only routing and VLANS.
The only wifi access is ubiquiti AC 5Ghz (tried both 40mhz and 80mhz width as mentioned, 40mhz is faster).
As mentioned, if I plug myself into 1gbit wire in that small switch OR tplink, I get wire speed (100 MB per second samba copying).
Ubiquiti is on poe injector that came with it
so my question is, if I am getting reported wifi speeds of 800mbit on 5Ghz AC 80Mhz channel width or 400mbit on 5Ghz AC 40Mhz channel width, why is my file transfer 170mbit (21MB/s) if we know the network equipment can push 100MB/s via wire.
Station f0:b6:1e:92:0b:58 (on phy0-ap0)
inactive time: 0 ms
rx bytes: 3911341
rx packets: 38855
tx bytes: 1539774097
tx packets: 1008393
tx retries: 49
tx failed: 0
rx drop misc: 10
signal: -50 [-54, -52] dBm
signal avg: -49 [-53, -51] dBm
tx bitrate: 360.0 MBit/s VHT-MCS 8 40MHz short GI VHT-NSS 2
tx duration: 2204953997 us
rx bitrate: 400.0 MBit/s VHT-MCS 9 40MHz short GI VHT-NSS 2
rx duration: 0 us
airtime weight: 256
authorized: yes
authenticated: yes
associated: yes
preamble: long
WMM/WME: yes
MFP: no
TDLS peer: no
DTIM period: 2
beacon interval:100
short slot time:yes
connected time: 351 seconds
associated at [boottime]: 69420.466s
associated at: 1739700498624 ms
current time: 1739700849830 ms
reported speeds are double then what I get in real life, is this expected on this ubiquiti AP or with 5ghz AC wifi in general?
I was suspecting one of the ports in the network went down to 100mbit, but:
the speeds I get are too high, almost 200mbit.
I checked all of them, they report 1gbit.
EDIT: I managed to get ~29MB/s transfer by playing with channels. (80Mhz channel width, channel 157.
You suspect the ap is underpowered? I have had no stability issues with it...
Actually nothing is written on the injector.
Back side of it is just black surface with some preparation to be mounted on that side. Front side you see on the photo.
Interesting, thanks. I will plug my laptop to it and see if I get anything lower then 100MB/sec samba transfer which I was getting directly on the switch and router
If it is not too late, do not plug the Poe output into your computer. This could fry your Ethernet port if it isn’t expecting PoE (most computers do not), because the poe injector is passive and will put voltage on the line without any tests/negotiation.
Your poe adapter will have a label on it - possibly under a removable mounting bracket on one side. If it has a “G” in the model number, it is gigabit. I’m pretty sure your adapter is indeed gigabit based on this:
I plugged in directly to the PoE injector and got the same wire speed as on other points in the network. 100 Megabytes per second samba share file transfer.
This experiment has proven all other parts of the network except ubiquiti AP itself are not the bottleneck.
Let me explain the network topology: my laptop was connected to the poe injector, which is connected to tplink gigabit switch interface. on the other side samba linux server is connected to the gigabit switch which is connected to the tplink gigabit switch interface. so the file transfer went through every piece of network equipment with 100MB per second transfer speed. But when going over ubiquiti 5Ghz wifi, transfer speed is 30MB per second...
it was too late, I was wondering if it will fry my nic interface, but so far so good.
I briefly googled it, found a statement it is perfectly safe to connect laptops to poe switch. but this is not a switch, just a dumb injector so I guess you are right it just pushes voltage without any negotiation...
This is ~168Mbps. This sounds about right for a 40MHz wide channel on that device when you consider wifi's unpredictability (RF noise, signal strength, etc.) and all the overhead involved for both WiFi in general and SMB copies.
It's worth noting that you're reporting ~800Mbps, which is around 15% or so below the maximum rate you'd expect for gigabit copper.
The above are the symbol rates, not actual data throughput. You can expect a best case of around 2/3 the symbol rates as your actual data bandwidth. So the tx bandwidth will likely be around 240Mbps, and then we'll factor in the ~15% overhead that was observed even on the wired connection... that gets us to about 200Mbps in a realistic best case scenario. But you said your area is noisy with RF, so, drop that down a bit more and you can see that the ~170Mbps really isn't so bad.
