My experience with the RT-AC86U

[Viktor Jaep]

Just wanted to say thanks to @eibgrad for the PoC and the concept code, and for @SomeWhereOverTheRainBow for the heavy testing and implementation... it's fantastic how you overcame this serious flaw using this method. Raising a glass in your honor! And serious thanks to @Oracle for his persistence in testing and guiding this. And to anyone else involved in this deep-level conversation - you have my thanks as well. Let's hope Asus eventually acknowledges the issue, and will actually issue a future fix, instead of us being forced to just adopt the AX line.

 

[Tech9]

I just need a hardline into it

Only if you want to stick to Asuswrt-Merlin and keep the hardware cost low. For wired router/firewall there are better options, IMO.

 

[dlandiss]

RT-AX86U is up and running, but it has weaker signal to my test AC client behind 2 walls.

RT-AC86U - 585/585
RT-AX86U - 390/390

I got this one just to play with it. It has scheduled surgery procedure in coming weeks.

And I had the opposite experience. Each step up, from N66U to AC68U to AC86U to AX86U, provided faster data speeds between the router and my favorite chair diagonally opposed to and one floor down from the router. The house is 120 years old and has large rooms and very thick walls.

 

[Tech9]

I'm getting 4dBm lower signal levels on 5GHz with AX86U compared to AC86U on both lower and higher channels. 2.4GHz band is exactly the same. The "feel" to AC clients (since AC86U is AC only) is the same. Whoever has good working AC86U and mostly AC clients and doesn't care much about software gremlins the upgrade to AX86U will bring close to nothing in performance. The best 5GHz range so far had AX88U. AX86U is better to AX clients and if scripts are run on the router, multiple VPN clients, etc.

I like the more "normal" blinking LED for 2.5GbE port used as WAN, the fact both WAN and 2.5GbE port have red no service indicator, the 2x USB3 ports. Thermal management is improved, both radios and CPU run in lower temperature range. The LEDs are too bright. The design is good, I like vertical space saving shape. It's an improved AC86U and a good direct replacement model. This one is Made in China, H/W rev 1.0, Year 2022. The price was usual about $250 or CAD330. I would say it's not expensive.

This router will be used only for fun and perhaps will be cracked open soon for internal examination. If it survives (I'm sure it will) and after I lose my interest to it, it will be donated to someone in need, like the other Asus routers I recently had in my collection.

 

[128bit]

well, i took the dive and for the most part, can agree with tech9. i put my ax86u online yesterday and i'm pondering whether to return it and bring back my tried-and-true ac86u. the ax signal strength seems to match but not exceed the ac, but where it truly shines is with intranet nas. my vids/tunes load on the ac devices at least 3 times as fast. problem is, there's only so much home-grown stuff we do. that's the end of the benes over the ac.

when using file explorer, the router itself shows up but i can't "see" the nas on my wired win 10 pc. weird cause my wireless laptops and other devices see the nas fine with either radio. i'm at a complete loss with that issue. granted, i'm running the official fw but will wait for the next merlinware release before replacing it. this issue may eventually get a separate posting.

to me, it only goes to show how well built the ac86u is. i know there's chatter about heat and cpu utilization but mine just runs even though it's behind a closed cabinet door on a warmer 2nd floor. fwiw, the house is 3,200 sq ft and the only troublesome device for maintaining its connection is the sprinkler which is on the outside of a brick wall. if the ax can maintain that connection better, well you know. and folks yes, the cabinet door position seems to be insignificant with respect to signal strength.

our fastest devices are "ac." even if we get new phones, speed is not a concern for us. we run our iphone X today on the 2.4g radio with wifi calling enabled and are plenty pleased. sorry, if that makes some of you hi-rollers puke.

 

[Tech9]

i'm pondering whether to return it and bring back my tried-and-true ac86u.

If the issues discussed here don't bother you, AC86U is a very good performing router. And very cheap new at the moment. It was available in my local computer store for $116 or CAD150. Less than half price, compared to AX86U. There is a hope newer production year AC86U routers have less hardware issues at least. There is a hope Asus will fix the software issue at some point.

even if we get new phones, speed is not a concern for us.

Correct. AC Wave 2 router to common 2-stream client can do >500Mbps throughput. What a phone/tablet will use it for?

 

[128bit]

If the issues discussed here don't bother you, AC86U is a very good performing router. And very cheap new at the moment. It was available in my local computer store for $116 or CAD150. Less than half price, compared to AX86U. There is a hope newer production year AC86U routers have less hardware issues at least. There is a hope Asus will fix the software issue at some point.



Correct. AC Wave 2 router to common 2-stream client can do >500Mbps throughput. What a phone/tablet will use it for?

so yeah, my ac86u is still here, ready to go back into service after this testing. had it since jan 2018. there was talk back then that newer models were less stable but i can't confirm this, after all it's the internet. mine has been solid. what bug are you referring to?

as i had mentioned, we enable wifi calling but essentially talk/text. we consider ourselves very average users. it just works - have not measured but it completely eliminates the poor cell connection problem. keep in mind, it's been in use for 4.5 years, k.

this blog is a superb tool for keeping up-to-date on tech but we're average. that said, if a user has wifi 5 devices, a working ac86u may be tough to beat.

 

[Tech9]

what bug are you referring to?

Better description and mitigation options here:

Stuck commands

Has anyone else observed any "stuck" commands on their router since upgrading to 386.x? Below is a frequent occurrence of the wl command getting stuck being spawned by conn_diag (this process was created at 4:10 AM). But I've also seen it happen with some nvram get commands (I once had Skynet...

www.snbforums.com

this blog is a superb tool for keeping up-to-date on tech

You don't need the latest and greatest. Follow @Smokey613 for weekly upgrade ideas.

 

[Tech9]

Just a small update: AX86U 5GHz radio works better on channels 149-161 in my environment. AC86U radio was better on 36-48. Not very important in my case, but worth mentioning observation.

 

[capncybo]

Just a small update: AX86U 5GHz radio works better on channels 149-161 in my environment. AC86U radio was better on 36-48. Not very important in my case, but worth mentioning observation.

Could you please define what you mean by "better"... I think you are referring to the Signal Strength being Higher on the AX than the AC within the 149-161ch range?
EDIT: But you also mean, within ch 36-48 the signal strength is actually higher on the older AC-model, Correct?

 

[capncybo]

I'm asking because I "currently" have my RT-AX86U set on automatic channel (Which has selected 36).
Yet I'm fairly certain I had better signal strength via my older RT-AC-68U when placed in the same location.
Hence I might try the manual setting within the 149-161ch range when I get home from work.

 

[Tech9]

I'm asking because I "currently" have my RT-AX86U set on automatic channel (Which has selected 36).

Here is what I see, AC86U and AX86U compared side by side to test AC clients at the same locations - same room 3m away, behind one wall 10m away and behind 2 walls 15m away, plus some more differences:

- AX86U shows 4dBm lower signal strength from clients on all channels, perhaps less sensitive radio.
- AC86U can do >60Mbps real transfers on channels 36-48, but no more than 40Mbps on 149-161.*
- AX86U can do >60Mbps real transfers on channels 149-161, but no more than 40Mbps on 32-48.*
- 2.4GHz band is exactly the same on both, same dBm reported and same transfer speeds.
- despite the dBm differences on 5GHz band, I would say both routers have identical performance to AC clients, just need to be set differently.
- I tried to crash AX86U with my test transfer to USB attached SSD and it survived, AC86U doesn't. I've seen this on AX88U though - 1GB RAM.
- AX86U LEDs are much brighter than the LEDs on AC86U - a bit uncomfortable, if the router is used in a dark room.
- AX86U WAN LED blinks differently when WAN port is used and when 2.5GbE port is used for WAN. I like 2.5GbE port LED better.
- Asus App can't restore AX86U LEDs after turned off, no LAN LEDs coming up. Works in Asuswrt-Merlin GUI though. Minor inconvenience.

* - Wi-Fi environment specific

I have the router for few days only and had no much time to play with it. The case is a bit cheap plasticky looking, the antennas were not aligned perfectly and needed slight force adjustment. Acceptable for $250 router though with good hardware inside. I like the normal power jack and the new switch, but there is a design flaw - on/off switch is at the wrong side of the power jack and is too close to it. It's a bit hard to reach.

I had to open a new thread for this comparison information, apologies @Oracle.

 

[Tech9]

Sorry, AX86U auto selects control channel 48 here, but works best with channel 153 manual. @capncybo

And this proves my own theory "the router knows best" wrong, at least with this model router. @bbunge

AC86U auto selects control channel 40 and works best on 36-48 @80MHz wide. AX86U has more things to learn about Wi-Fi.

 

[SAL9K]

I too experimentally found that 149 just worked better on the AX86U (for me, when set to auto the router typically chooses ch [36-48], I.e. doesn't choose 149!)... is there an AX86U channel/tx-pwr mapping (based on region, I suppose)? Or is the receiver more sensitive at the higher channels (all of my devices generally see better throughput at 149)? BTW, I have a North American h/w version.

 

[Tech9]

I have changed my mind - keeping AC86U as well. Too much work put on this one. Sentimental value. The two brothers may play AiMesh one day.

 

[capncybo]

Sorry, AX86U auto selects control channel 48 here, but works best with channel 153 manual. @capncybo

And this proves my own theory "the router knows best" wrong, at least with this model router. @bbunge

AC86U auto selects control channel 40 and works best on 36-48 @80MHz wide. AX86U has more things to learn about Wi-Fi.

I also have my doubts about relying on too many "automatic modes" but it might not even be fair to assume the router is wrong because... What if a couple of the neighboring routers are also on "automatic channel selection"... they frequency hop, we frequency hop... It sounds like a Happy Coordinated arrangement but it's unnecessarily resource-intensive & obviously problematic. Perhaps it's already written into some of the WiFi Standards but I doubt any algorithm can correctly predict where the next neighboring router is going to jump... Perhaps it's tied to some of the advanced wifi features. But even if it was... Many advanced features are NOT-100% implemented by all vendors.

Simple point (Too much jumping around by everything) makes it really hard to find an (Empty/Reliable) ch.

 

[Oracle]

Some new reference info
The nvram / wl error persists with the new firmware (386.7). After upgrade, the test loop got stuck at iteration 1861.

Here's my usage log before rebooting the router for firmware upgrade:

Usage statistics of nvram and wl overrides
Uptime: 00:22:01 up 5 days, 23:55, load average: 3.04, 2.95, 3.01
nvram use count: 19371
nvram errors count: 4
wl use count: 34430
wl errors count: 12

As you can see, in 6 days of normal operation (no stress testing, no fooling around with the commands) the router accumulated 16 errors, an average of nearly 3 per day. The overrides were invoked nearly 54 K times.

These are 16 hang ups prevented. In all 16 cases, the requested value was retrieved at first retry.
Apart from these errors, the router was very stable.

Everything else appears normal or at least I haven't noticed any problems yet.

 

[chongnt]

Some new reference info
The nvram / wl error persists with the new firmware (386.7). After upgrade, the test loop got stuck at iteration 1861.

Here's my usage log before rebooting the router for firmware upgrade:

As you can see, in 6 days of normal operation (no stress testing, no fooling around with the commands) the router accumulated 16 errors, an average of nearly 3 per day. The overrides were invoked nearly 54 K times.

These are 16 hang ups prevented. In all 16 cases, the requested value was retrieved at first retry.
Apart from these errors, the router was very stable.

Everything else appears normal or at least I haven't noticed any problems yet.

Have you try this step by @ColinTaylor?
After change pid max I have not notice conn_diag stop updating. Running CheckStuckProcCmds.sh by @Martinski found no stuck command.

My experience with the RT-AC86U

Here's some log output from running the test loop on tighter timings. It reported 5 retries, while the average is 2-3. Running time appears unaffected: 5:07 min per 3000 loop iterations. I switched to these versions because the other one using timeout falls apart if I unplug the USB drive...

www.snbforums.com

 

[Oracle]

Have you try this step by @ColinTaylor?
After change pid max I have not notice conn_diag stop updating. Running CheckStuckProcCmds.sh by @Martinski found no stuck command.

I just did it now. I'll see what the effect will be.

cat /proc/sys/kernel/pid_max
32768
echo 4194304 > /proc/sys/kernel/pid_max
cat /proc/sys/kernel/pid_max
4194304

Here's the starting point:

Usage statistics of nvram and wl overrides
Uptime: 19:13:16 up 2:24, load average: 2.63, 2.81, 2.81
nvram use count: 339
nvram errors count: not available
wl use count: 548
wl errors count: 1

In just around 2.5 hours of uptime there's already 1 failed wl command - before changing the pid_max.

 

[Oracle]

Here you are.
After only 5 hours of uptime we have a second fish in the tank. The pid_max doesn't appear to be making a difference, at least so far.

Usage statistics of nvram and wl overrides
Uptime: 21:49:05 up 5:00, load average: 3.15, 3.01, 2.91
nvram use count: 683
nvram errors count: not available
wl use count: 1168
wl errors count: 2

2 errors within 2000 calls (in just 5 hours) is relatively high. I wonder if the new firmware has anything to do with it. During the previous rounds of testing it took 3K to 5K calls for a single error to show up.