What's new
  • 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!

Beta Asuswrt-Merlin 3006.102.4 Beta is now available

I'm running beta2 on my BE96U as the main with one AiMesh node (BE92U) running stock fw.

I've been having issues where the wifi networks all drop ever 12-24hrs. In the AiMesh settings, I'm running Wifi7 with MLO. The most instability seems to happen when I enable 'MLO Fronthaul for Clients' which I'd like to use if it were more stable.

I've also upload the syslog file (BE96U) for when the wifi network all drops off, the timestamp to start looking at the wifi dropping event is "Apr 22 19:10:14".
When I've run this syslog through some LLMs, it seems to point to the Broadcom driver due to error messages such as "WLC_SCB_DEAUTHORIZE error (-30)" and "Previous authentication no longer valid (2)".

Maybe this will be helpful in tracking something down. Thanks Merlin.
I believe there’s currently nothing that can be done to address Broadcom-related errors or issues with Multi-Link Operation (MLO), as these are proprietary to Broadcom. That said, I own the same router and, in my experience, there is no tangible advantage to using MLO over simply enabling a 320 MHz channel width on the 6 GHz band.

MLO operates by splitting traffic—one connection for upload and another for download. However, there are no consumer-grade adapters on the market today that can handle more than two simultaneous connections, not unless you wish to spend $300-$600 only on a Wi-Fi adapter. I recommend disabling 2.4 GHz as an available MLO band. The biggest issue I’ve noticed is that MLO can actually reduce performance. One direction (either upload or download) often underperforms, creating an imbalance. For example, your device might be downloading over 6 GHz at 320 MHz and achieving speeds around 1.8 Gbps at 25 feet. But if upload is routed through the 5 GHz band at 160 MHz, your connection might show an estimated 2 Gbps link speed—yet, real-world throughput drops to around 1 Gbps or less due to overhead.

In my opinion, MLO is not ready for widespread use. I suspect that each MLO link is not operating in half-duplex mode, effectively cutting the available bandwidth in half still. I also believe the advertised performance claims are misleading, especially when encryption overhead is factored in—though I haven’t yet tested this specifically.

There has been some discussion about eventually allowing any band and any channel width to participate in MLO. If that becomes possible, we could potentially see two separate 6 GHz connections each using 320 MHz, which might allow a 2x2 client to achieve a full 5 Gbps in both directions. That would be a game-changer.

Lastly, don’t rely too heavily on your device’s reported link speeds. In practice, you’ll typically get about half of the estimated throughput. If you want a more accurate measure, check your router’s wireless logging page. It displays real-time bandwidth allocation per client—but again, expect to see only about half of those numbers in real-world use.
 
Last edited:
If you rely on multiple scripts like Diversion or Skynet, I’d recommend using an SSD instead of a USB stick. I went through several USB sticks before switching to one of these enclosures and using a 128GB SSD drive from an old computer. I haven't had any issues since I made the switch about a year ago.
I'll second this motion: I got a 512GB m.2 SSD and appropriate enclosure for $50-60 and it has been ridiculously good in comparison to previous hardware I had used
 
Asuswrt-Merlin 3006.102.4 Beta is now available. There are a number of significant things to note with this release:

1) BCM4912 models are now migrated to this 3006 code base. This includes the following device list:
  • GT-AX6000
  • zenWifi Pro XT12
  • GT-AX11000 Pro
  • GT-AXE16000
  • RT-AX86U Pro
  • RT-AX88U Pro

2) New addition - the RT-BE92U is now supported. Note that the GPL used for this initial support suffers from some CPU-related issues, so support for this model at this time should be considered EXPERIMENTAL. If you lose the ability to access the router, then power cycle it, that usually solves it. Beta 2 updated that device's GPL to 102_37526, which should be stable.

3) Beside the addition of new models, the focus of this release is the merge of GPL code (I currently need to manage three separate GPL versions for this release.....), updated components, and a number of fixes.

Update (19-Apr): 3006.102.4 Beta 2 is now available. Changelogs for all three branches:

Wifi 6:
Code:
fe104c9774 (3006.102) webui: updated DNS Director notes
dcef132c02 Update RT-BE96U and GT-BE98_PRO wireless dongle driver to 102_37812 builds
1cc4127a8d miniupnpd: fix building (missing header file)
85c9d353af miniupnpd: update to 2.3.8
198e89a5e8 rc: move service-event-end back within again: loop
170d2e48b5 rc: Have DNS Director "Router" mode use REDIRECT target instead of DNAT to LAN IP in iptables.
8d40fbd7f2 Bumped revision to beta 2
879ad023d4 webui: fix missing button to remove Offline Client List entries
c77742afb1 webui: cleanup sysdep Wireless pages that are no longer existant

Wifi 7:
Code:
fe104c9774 (3006.102) webui: updated DNS Director notes
dcef132c02 Update RT-BE96U and GT-BE98_PRO wireless dongle driver to 102_37812 builds
1cc4127a8d miniupnpd: fix building (missing header file)
85c9d353af miniupnpd: update to 2.3.8
170d2e48b5 rc: Have DNS Director "Router" mode use REDIRECT target instead of DNAT to LAN IP in iptables.
8d40fbd7f2 Bumped revision to beta 2
879ad023d4 webui: fix missing button to remove Offline Client List entries
c77742afb1 webui: cleanup sysdep Wireless pages that are no longer existant


RT-BE92U:
Code:
fe104c9774 (3006.102) webui: updated DNS Director notes
dcef132c02 Update RT-BE96U and GT-BE98_PRO wireless dongle driver to 102_37812 builds
1cc4127a8d miniupnpd: fix building (missing header file)
85c9d353af miniupnpd: update to 2.3.8
170d2e48b5 rc: Have DNS Director "Router" mode use REDIRECT target instead of DNAT to LAN IP in iptables.
8d40fbd7f2 Bumped revision to beta 2
9fe89ff9b6 httpd: fix CPU temperature report on BCM4916
cb50c47eff Merge RT-BE92U binary blobs from 37526
789a111d63 Merge with GPL 102_37526 (RT-BE92U)
879ad023d4 webui: fix missing button to remove Offline Client List entries
c77742afb1 webui: cleanup sysdep Wireless pages that are no longer existant


Changelog:
Code:
3006.102.x (xx-xxx-xxxx)
  - NOTE: If migrating from a 3004 firmware, only the first Guest
          Network will be migrated - any additionnal GN must be
          manually reconfigured.
  - NOTE: If migrating from a 3004 firmware, the Wireless Scheduler
          will need to be manually reconfigured if you were using
          it.
  - NOTE: As a reminder, the ROG variant of the webui is not
          supported in the 3006 firmware series, as maintaining
          two separate interfaces is too much extra work.

  - NEW: Moved RT-AX86U_PRO, RT-AX88U_PRO, ZenWifi Pro XT12,
         GT-AX6000, GT-AXE16000 and GT-AX11000_PRO to the
         3006 firmware series.

  - NEW: Added RT-BE92U support, based on GPL 102_37435.  This GPL
         seems to have some stability issues, so consider support
         for this model to remain Beta, until a fix is provided
         for it.

  - UPDATED: Merged GPL 3006.102_36521 for Wifi 6 models (Wifi 7
             GPL codebase is unchanged from 3006.102.3).
  - UPDATED: OpenVPN to 2.6.14.
  - UPDATED: miniupnpd to 2.3.7 (20250207 snapshot)
  - UPDATED: amtm to 5.2 (decoderman)
  - UPDATED: dnsmasq to 2.91.
  - UPDATED: dropbear to 2025.87.
  - FIXED: Missing hostname on Wireless Log for MLO-capable Wifi 7
           clients.
  - FIXED: CVE-2024-9143 in OpenSSL (Debian backport by RSDNTWK)
  - FIXED: CVE-2024-13176 in OpenSSL (Ubuntu backport by RSDNTWK)
  - FIXED: Guest Networks on an isolated VLAN with DNSDirector set
           to "Router" would fail to do any name resolution.
  - FIXED: Wrong tab selected when clicking on "Profile" on the VLAN
           page (dave14305)

Make sure you do read the changelog, especially if migrating from 3004 to 3006. I documented the known recommendations from Asus' own changelog, but there might be further compatibility issues arising when upgrading from 3004. In particular, I would recommend disconnecting and reconnecting any AiMesh node. Also, double check Wireless settings.

For this beta release I need test feedback for:
  • The models that were migrated from 3004
  • General RT-BE92U performance
  • The Wireless Log page should be more accurate on Wifi 7 devices, the root problem of missing/incorrect hostnames having been resolved

Please keep discussions on this specific beta release.


Downloads are here.
Changelog is here.
Upgraded AX88Pro from Beta 1 - Beta 2. Dirty upgrade with no problems.

As I reported previously under authentication mode while pull down indicates that AES-TKIP is available if you select WPA-WPA2 Personal it really isn't.

Also under authentication mode if you click on the explanation? it describes what various options are including an option not available WPA-AUTO-Personal. If possible, it might be useful to match the explanations with what is actually available in the router's firmware.

While this isn't necessarily related to the Betas you apparently can't dismount multiple USB drives through the firmware if they are connected to an USB hub which is plugged into the single USB port on the router.

Again thanks for your work.
 
I believe there’s currently nothing that can be done to address Broadcom-related errors or issues with Multi-Link Operation (MLO), as these are proprietary to Broadcom. That said, I own the same router and, in my experience, there is no tangible advantage to using MLO over simply enabling a 320 MHz channel width on the 6 GHz band.

MLO operates by splitting traffic—one connection for upload and another for download. However, there are no consumer-grade adapters on the market today that can handle more than two simultaneous connections, not unless you wish to spend $300-$600 only on a Wi-Fi adapter. I recommend disabling 2.4 GHz as an available MLO band. The biggest issue I’ve noticed is that MLO can actually reduce performance. One direction (either upload or download) often underperforms, creating an imbalance. For example, your device might be downloading over 6 GHz at 320 MHz and achieving speeds around 1.8 Gbps at 25 feet. But if upload is routed through the 5 GHz band at 160 MHz, your connection might show an estimated 2 Gbps link speed—yet, real-world throughput drops to around 1 Gbps or less due to overhead.

In my opinion, MLO is not ready for widespread use. I suspect that each MLO link is not operating in half-duplex mode, effectively cutting the available bandwidth in half still. I also believe the advertised performance claims are misleading, especially when encryption overhead is factored in—though I haven’t yet tested this specifically.

There has been some discussion about eventually allowing any band and any channel width to participate in MLO. If that becomes possible, we could potentially see two separate 6 GHz connections each using 320 MHz, which might allow a 2x2 client to achieve a full 5 Gbps in both directions. That would be a game-changer.

Lastly, don’t rely too heavily on your device’s reported link speeds. In practice, you’ll typically get about half of the estimated throughput. If you want a more accurate measure, check your router’s wireless logging page. It displays real-time bandwidth allocation per client—but again, expect to see only about half of those numbers in real-world use.
Thank you for such a detailed and insightful reply.
So I've turned off the 'MLO Fronthaul for Clients' for the time being due to the instability it causes but I'm keeping the MLO enabled for the AiMesh between the BE96U <=> BE92U since I think that's useful??

The only reason I've tried to get the MLO working the whole way through is that my home office desktop needs whatever bandwidth I can get it on the far side of the house. I currently have 1gig ATT fiber service (up to 5gig, maybe in the future), so I want to try to get as much of that bandwidth across to that desktop. I bought this TpLink desktop adapter since it seems to be one of the few that exists which "supports MLO": https://www.tp-link.com/us/support/download/archer-tbe550e/

In it's current state, does MLO between AiMesh nodes seems useful to have for extra bandwidth or is that really causing more issues than it fixes?
 
Thank you for such a detailed and insightful reply.
So I've turned off the 'MLO Fronthaul for Clients' for the time being due to the instability it causes but I'm keeping the MLO enabled for the AiMesh between the BE96U <=> BE92U since I think that's useful??

The only reason I've tried to get the MLO working the whole way through is that my home office desktop needs whatever bandwidth I can get it on the far side of the house. I currently have 1gig ATT fiber service (up to 5gig, maybe in the future), so I want to try to get as much of that bandwidth across to that desktop. I bought this TpLink desktop adapter since it seems to be one of the few that exists which "supports MLO": https://www.tp-link.com/us/support/download/archer-tbe550e/

In it's current state, does MLO between AiMesh nodes seems useful to have for extra bandwidth or is that really causing more issues than it fixes?
Causes more issues because as I said MLO just is not ready. I would recommend turning off MLO entirely and then using 6Ghz 320Mhz width as the back haul for wireless transmission. Enable front haul use for clients connected to the mesh node so those clients can use 6Ghz. Just keep in mind that if front haul is enabled it will share the overall back haul max bandwidth, with that of the client. The other option is to link both the main node and mesh node with a wired back haul then all the wireless bands are free to use. The BE96U has two 10Gbps capable Ethernet ports just for this reason. The BE92U only has 2.5Gbps Ethernet options for backhaul because the WAN port is the only 10Gbps.

Unless you are downloading huge files consistently or running servers with a lot of customer connections 5 Gig is a complete waste. The average home user and even tech people like myself 2 Gig is more than enough. These high throughputs just really help me download the large PC games these days some of which can be a terabyte of data.

High throughput doesn’t make your internet faster just capable of getting larger amounts of data completed quicker. The average PC game is only transferring 50/60 kbps to update character placement and other stuff. Even 4K video stream is not as demanding as most would think. Again though if you are hosting a server that’s a different story depending on a lot of factors.

Would love to see port bonding as a GUI option on these routers then you could bond two 2.5 Gbps ports on each router to reach 5Gbps wire backhaul throughput.
 
Last edited:
If you rely on multiple scripts like Diversion or Skynet, I’d recommend using an SSD instead of a USB stick. I went through several USB sticks before switching to one of these enclosures and using a 128GB SSD drive from an old computer. I haven't had any issues since I made the switch about a year ago.

IT WAS THE DAMN USB DRIVE..IMAGINE THE TIMING..FAILED JUST AS I UPGRADED THE FW..WHY I AM ALWAYS THIS UNLUCKY? :/
 
Causes more issues because as I said MLO just is not ready. I would recommend turning off MLO entirely and then using 6Ghz 320Mhz width as the back haul for wireless transmission. Enable front haul use for clients connected to the mesh node so those clients can use 6Ghz. Just keep in mind that if front haul is enabled it will share the overall back haul max bandwidth, with that of the client. The other option is to link both the main node and mesh node with a wired back haul then all the wireless bands are free to use. The BE96U has two 10Gbps capable Ethernet ports just for this reason. The BE92U only has 2.5Gbps Ethernet options for backhaul because the WAN port is the only 10Gbps.

Unless you are downloading huge files consistently or running servers with a lot of customer connections 5 Gig is a complete waste. The average home user and even tech people like myself 2 Gig is more than enough. These high throughputs just really help me download the large PC games these days some of which can be a terabyte of data.

High throughput doesn’t make your internet faster just capable of getting larger amounts of data completed quicker. The average PC game is only transferring 50/60 kbps to update character placement and other stuff. Even 4K video stream is not as demanding as most would think. Again though if you are hosting a server that’s a different story depending on a lot of factors.

Would love to see port bonding as a GUI option on these routers then you could bond two 2.5 Gbps ports on each router to reach 5Gbps wire backhaul throughput.
I software dev from home so it's mostly about deployment and large assets uploading where I need it the most. I would definitely do wired backhaul but I'm in a rental so I can't do the wiring through the house like I'd want to.

I'll give it a shot with MLO off and see where things go with your suggestions on the configuration of wireless bands between the main->AiMesh node.
 
IT WAS THE DAMN USB DRIVE..IMAGINE THE TIMING..FAILED JUST AS I UPGRADED THE FW..WHY I AM ALWAYS THIS UNLUCKY? :/
Unlucky? Only when using a USB stick. SSD in a USB enclosure will make you “lucky” ;-) And never look back.
 
Regarding guest network: I'm using network #2 at the moment with 3004 because I don't want the guests on a separate subnet (which is the case with guest #1). Is this still the same with 3006, that guest network #1 would use another ip subnet and therefore #2 should be used in my case? Would be nice if someone could tell me before upgrading to 3006. Thank you in advance.
Have you upgraded since then? I'm also using Guest #2 because of IP subnet of #1 is different. Even if it does not carry over it's easy to copy name and password.
 
Wondering if 3006 is mature enough to update or if I should stay on 3004 for next 6 months or so. What's major in 3006 that may push someone to update?
 
I can't get the guest portals working. I can activate them and let them show in my network list, but they can't connect to it. I have AGH active and Skynet. Maybe that is the problem. In the logs I see an entry dnsmasq that is giving the problem. Has anyone an solution?
 
In the logs I see an entry dnsmasq that is giving the problem.
I’m not best placed to read logs but maybe post that bit along with redacted screenshots of your GNP settings? All Tabs.
 
Have you upgraded since then? I'm also using Guest #2 because of IP subnet of #1 is different. Even if it does not carry over it's easy to copy name and password.
Yes I did, you can find the whole story within this thread. ;) Different logic now, living with different subnet at the moment. Also had to delete and reconfigure it because clients were no longer able to associate over night without changes, reboot didn't fix it. No idea why, but now the missing checkbox for main network access is available (deleted the network before without success, still a bit flakey obviously, at least on my AX86Upro).
 
But what was the solution, can you tell me that. I have tried many thing. I see this in the logs:

Apr 24 07:33:32 dnsmasq[1083746]: failed to create listening socket for 192.168.52.1: Address already in use
Apr 24 07:33:32 dnsmasq[1083746]: FAILED to start up
Apr 24 07:33:37 dnsmasq[1084650]: failed to create listening socket for 192.168.52.1: Address already in use
Apr 24 07:33:37 dnsmasq[1084650]: FAILED to start up
 
Unlucky? Only when using a USB stick. SSD in a USB enclosure will make you “lucky” ;-) And never look back.
Like don't SSD's fail mate? I have a 256gb nvme wd blue compact drive (dram less) i got off from the work PC that i upgraded and put in a usb 3 enclosure for m.2's, it's full of red SMART errors.

Also a 4TB new Samsung 990 Pro i bought from amazon on black friday, after 3 months developed an issue where under extended constant IOPS (specifically when i have macrium reflect do daily backup to HDD), its controller would randomly crash and crash the whole PC with it, then it would disappear in bios until you did a complete power off cycle. Spend 1 month trying to find what it was causing it, but the hint is that i would not have the whea/bsod error logged into event viewer. Amazon now sent me a replacement, only apparent difference is packaging and the replacement in made in korea rather than china. And this is on a big butt thermaltake m.2 heatsink that will not let it go over 51oC at full load.

Before that i had a 2TB Kingston KC3000, i got as soon as the Phison E18 controller the Micron B47R was released; after a year it started becoming randomly ultra slow, had a replacement as after testing and talking with the engineers at Kingston we couldn't identify the issue, only for the same issue to appear a year later. And it was at that time i found the issue, some random blocks will get stuck needing 1-3s each to read (without any chkdsk issues), and it was a Phison FW issue that was fixed..after 2 years: And this affects all Phison E18 based ssd's running this controller (so half of all pci 4.0 ssd's models released).

Not to account for the myriad of AMD CPU and GPU software bugs and silicon problems i've reported over the years.

It doesn't matter how much money you spend, all tech nowadays is unreliable trash, or i am so unlucky that the bane of my existence is to be able to fix stuff so that everything breaks on me so i need to fix it (including the women i end up with..)

The only thing is so far super reliable (knock on wood, metal and earth) is my 2 HGST HC530 (server grade) HDD's i use for storage/backup. I bought these specifically after i eliminated all other brands and consumer models as you can imagine how many HDD failures since my first Pentium 3 PC.

edit: Also fixed my cat back during covid, there is a disease called FIP (essentially it's covid alternative that only cats can get), and a variation called wet FIP that will kill the cat within 2 weeks, there was no official treatment (and still not is for most countries), big story short identified the issue and a potential solution of a drug called GS-441524 which inhibits one enzyme this virus uses, i had to source the drug immediately (as by the time i got it it was 1-2 days before the cat would pass away) and for 12 weeks i was doing daily injections and treatment that saved my cat (and a few others by helping other owners after). There is some joke somewhere about my situations, but you come up with it;p
 
Last edited:
It would be valuable for everyone reading this thread if everyone posting had a signature with their own setup. Without it, a post is pointless...
 
Dirty upgrade from 8.4 on a AX86U Pro, with diversion, unbound, scribe, vn-stat and scMerlin and 2 OpenVPN servers active. There are some hinky things going on in the addons that I'm working through, but the second VPN server is failing to start.

I noticed in the advanced settings the subnet mask is reported as: 255.255.255.0r2_nm=255.255.255.0. When I fixed that the server started normally.

I'm guessing I will need to reset to defaults and reconfigure, as in the upgrade at least this nvram variable got garbled.
 
I'm guessing I will need to reset to defaults and reconfigure
I'd go that route. You've gone from 3004 to 3006 firmwares with addons installed and running eg it's well known that YazFi doesn't play at all with the 3006 code - amongst other addons that will only work correctly if you switch to the develop branch or manually install.
 
Last edited:
I recommend disabling 2.4 GHz as an available MLO band.
This is what i was trying to achieve when enabling MLO on my RT-BE92U but enabling MLO causes the router to prompt that the 2.4Ghz radio needs to be enabled also in order for MLO to be enabled. I don't use 2.4Ghz and only 5Ghz/6Ghz bands. I previously had a TP-Link BE-550 (BE9300) router and it allowed MLO to be enabled only on 5Ghz/6Ghz bands while keeping 2.4Ghz off, would be good if this was an option also on RT-BE92U but it appears it's all or nothing. I contacted Asus with a suggestion to allow the option for 5Ghz/6Ghz only but haven't heard back. Maybe it's something Broadcom need to add support for in the driver/firmware, i am not sure.
 

Latest threads

Support SNBForums w/ Amazon

If you'd like to support SNBForums, just use this link and buy anything on Amazon. Thanks!

Sign Up For SNBForums Daily Digest

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