What's new

Performance of dual AX86U in AiMesh, on Merlin 386.1

  • SNBForums Code of Conduct

    SNBForums is a community for everyone, no matter what their level of experience.

    Please be tolerant and patient of others, especially newcomers. We are all here to share and learn!

    The rules are simple: Be patient, be nice, be helpful or be gone!

AndrewJacob

Regular Contributor
Hi all.

I have recently purchased 2 AX86Us and have set them up in AiMesh. I was constantly getting disconnects (followed by a prompt that my password is incorrect) when they were connected through wired backhaul via the 2.5 GBE ports, and only on the 5ghz radio. Now that I've updated to Merlin's 386.1, the disconnects have stopped, but I'm not sure performance is where it should be.
I have run iperf3 on a wifi-6 client, with the server side running wired through a 1GBE port on the main router. I was also testing 5 streams for 30 seconds per test. On my diagram, the first figure is mbits/second for download, and the second for upload.
Please see the below results... Are these performance numbers right? I feel like I should be getting significantly better performance, but maybe my expectations are too high.
If I should update to 386.1_2, please let me know, but I feel like something else is going on here.

Thanks,

Andrew
 

Attachments

  • Home plans_2020_AX86 - Results.png
    Home plans_2020_AX86 - Results.png
    473 KB · Views: 218
@AndrewJacob there are no performance numbers shown attached. :)
 
Correct @WhiteZero , it's the circles that have xxx/xxx on the image. Thanks @L&LD . To sum it up, the numbers look to be as such... This is throughput tested with iperf3 and not a speedtest...

Room
Download​
Upload​
Office 1
802​
593​
Office 2
799​
623​
2nd bedroom extremity
790​
604​
Main bathroom
807​
241​
Ensuite bathroom
795​
279​
Master bed
673​
403​
Dining room
681​
438​
Kitchen
664​
450​
Living room
502​
478​
Near patio door
443​
504​
 
Does anyone know if these throughput performance figures are normal/good for a dual AX86U setup? I feel like they are weak, but I could be wrong... I'd like to see if there's anything in the settings that I should configure before disconnecting this setup and running the same test on the AX11000...
 
What device are you speed testing on?

I only just setup my AX86U yesterday and run a few iPerf3 tests from my Galaxy S21 going to my Ethernet connected desktop. With line-of-sight to the router (like -40db), I think I got about 650 to 750 mbits, with my S21 showing a link-rate of 1.2Gbps. Since you usually get about half the throughput of your link-rate normally, I think our speeds are pretty normal. I'll have to test with the Wifi on my desktop (MSI X570 Unify mobo with Wifi6 built-in), since when I did a quick test it showed I had a link-rate of 2.4gbps.
 
Also would like to know what device was used for the wireless testing.
A couple of other thoughts; most complaints I see about speed tests are with QOS turned on. I think it still might have some issues. Also did you confirm you were connected to the node for the living room tests? It looks like you might have still been connected to the main.
I would start by disabling all the QOS & AIProtect stuff (click "withdraw " from the Administration>Privacy page and reboot both routers (AI Mesh>System Settings>System Reboot).
Then wait a few minutes for the system to settle and re-run a few of the tests.
Having said that, the results aren't too far off what I would expect on wifi.
 
Thanks guys!
1- @WhiteZero I was speed testing using an iPhone 11 Pro. I'm sure I could get faster speeds with a laptop, but if the phone can crack 800/600, i'd hope to see those kind of numbers everywhere maybe. Just not sure
2- @vango44 I don't know for sure, but you might be right. I may have still been connected to the main router on the second floor, and maybe there's a way to force my device to connect to that node (through the app), but I'd hope for comfortable switching between the router and the node so forcing it would only force a situation that I wouldn't experience in daily use...

If these speeds are normal, my only concerns would then be:
A) The low upload speeds in the bathrooms
B) The lower speeds in the living room and near the patio, which is where we usually use a lot of Wi-Fi so I need performance there. Being so close to the node I'd hope the AiMesh would use its Ai to realize I should be connecting to the node;)
 
Correct @WhiteZero , it's the circles that have xxx/xxx on the image. Thanks @L&LD . To sum it up, the numbers look to be as such... This is throughput tested with iperf3 and not a speedtest...

Room
Download​
Upload​
Office 1
802​
593​
Office 2
799​
623​
2nd bedroom extremity
790​
604​
Main bathroom
807​
241​
Ensuite bathroom
795​
279​
Master bed
673​
403​
Dining room
681​
438​
Kitchen
664​
450​
Living room
502​
478​
Near patio door
443​
504​

The horror, don't know how anybody could survive with these numbers :)
 
@AndrewJacob, I believe that your main and node is too close together (particularly for wired AiMesh mode).

I would be looking to 'spread' the main - node as much as possible as shown below.

AiMesh Ideal Placement


The node or AiMesh doesn't use 'its AI', the client devices do, when deciding where to connect.

Still waiting (with the others above) for the devices and methods used to get these numbers. :)


On my 5-year-old, AX210 equipped laptop with the latest drivers (see the hoops @thiggins had to jump to get them), I get speeds that eclipse what you show almost anywhere I care to use it in my home. While my other decade-old laptop is jealous of your slowest speeds.
 
Hmmm, you may very well be right @L&LD . The problem on the second floor is that the telecom panel is placed where it is... And on the ground floor, I am as far away as I can get...
Server: Core-i5 3570k, 32gb ddr4 ram, Geforce gtx1070 8gb, Samsung 870 evo 500gb ssd, windows 10, running iperf3 as server from cmd line, connected to main router via 1gbe lan port using cat6 cabling
Client: iPhone 11 Pro running the iperf app
I have laptops too, so I'll try with those. Honestly if I get 800 everywhere I'd be happy so I don't see my iphone tester as the limiting factor here. My goal is that when I upgrade to gigabit internet in a few months, that I have the best setup to take full advantage of it. imho if i'm getting 800 everywhere, i'd be satisfied. but if you think i should retest with a laptop, i can.
Also, FYI, I've upgraded to the 386.1_2 firmware from Merlin
 
361.1_2 definitely will not perform any faster...unfortunately.
On my side, AX200 Ryzen7 laptops shows wifi properties at 2402/2402Mbps, but gigabit internet shows only 500-560Mbps (LAN: 980-1000Mbps) and network map shows 300/240 (?!)
 
Thanks @Sky1111 . Before I disconnect the AX86U to test the AX11000, I'd like to know if anyone knows how to optimize the wired responsiveness of devices connected to the node? Youtube tends to lag and buffer more than usual since i've connected my AppleTV to the node. Any ideas how to resolve this performance issue?
 
Thanks @L&LD , but I forgot to mention that the lag occurs on a wired client! Any way to fix that? Wasn't happening before on the cheap router
 
Is that just your ISP being overwhelmed at certain peak periods? Other (external) possibilities?
 
Not unless that coincided with my installation of the AX86U routers... As an FYI, my desktop computer is wired to the main router and has no such issues whatsoever. I am perplexed as to what to try to resolve the issue... @L&LD
 
So I just plugged the GT-AX11000 back in with the old cheap Zyxel router and my wired performance at the node is snappy again compared to the dual AX86U.
Any possible explanations? Something I could do to the setup to eliminate that lag? I only experience the issue with wired clients on the node. @L&LD
 
Last edited:
Seems like the only thing left to check/change is the cables at this point. Or just re-crimping the ends may be enough to give the dual RT-AX86U the solid connection that seems to be missing.

Another way to test is by putting an unmanaged switch between the RT-AX86U's (one, or maybe, at each end). Of course, for 2.5GbE backhaul, this isn't a feasible, 'solution', given the cost of greater than 1GbE switches.

Does the latency disappear when using the 1GbE LAN ports instead (and maybe with the tips above)?
 
i used the exact same cables for both setups.. i feel like either the second ax86u has some issue or it's software related. i will try once more with the 1gbe ports for backhaul maybe... tbh, on wifi the ax86 is snappier, but the ax11000 seems to just work better right now. and with merlin coming to it soon, i'm wondering if that isn't the better solution... i have to return either the ax86u or ax11000 today.
pros for the ax86: snappier wifi, more range in backyard, smaller footprint and better looking than ax11000, merlin (not anymore soon), better upload throughput, more stable wifi throughput when testing (although merlin on ax11000 might level the playing field)
pros for the ax11000: better lan at node, slightly faster download speeds over wifi, triband (although not a significant upside, i like being able to have all work laptops on a separate radio).

i just want to find a way to resolve this lag on lan (it's like buffering almost), because it does affect us in our every day usage and i'll have to keep the ax11000 if i can't resolve this today:/ @L&LD
 

Similar threads

Latest threads

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top