Release Asus RT- AX58U Firmware 386.49674 (7-21-2022)

[Deleted member 66500]

Not really. You seem to think anything under 300.00 bucks is junk. News flash for you 90% of people will never spend that for a router it's simply not necessary. Another news flash 90% of Asus router owners never heard of RMerlin. Just stop being negative about lower priced routers there is nothing wrong with them they get the job done period. I seem to remember Tim Higgens warning you about your attitude on these forums it's getting old.

Thanks for talking sense, and its a problem with quite a few (many?) voluble posters in this forum - they always express opinions biased by their personal beliefs/preferences, rather than facts and measured guidance. It doesn't affect me, but it can be detrimental to new members who aren't familiar with the nitty-gritties. There was a recent thread about RP-AX56 AImesh node where such rather biased opinions were expressed as if expert advise (even though RMerlin had exactly advised against such personal biases expressed in that thread)

 

[Tech9]

Does anyone have a guesstimate on when the AX-58u will get the Wireguard Client feature

If this router is not running on 4.1.x kernel, most likely it won't be getting WireGuard VPN option in 388 code firmware. Also, WireGuard Server/Client on currently supported models is incompatible with NAT acceleration and has negative impact on the entire network WAN throughput. On HND ARMv8 routers OpenVPN can reach 200-250Mbps, but the devices not using the VPN tunnel are unrestricted. WireGuard can reach 300-350Mbps, but the devices not using the VPN tunnel also become restricted. I can test and verify that on my AX86U router, but a bit later after official 388 is released.

 

[Deleted member 66500]

If this router is not running on 4.1.x kernel, most likely it won't be getting WireGuard VPN option in 388 code firmware. Also, WireGuard Server/Client on currently supported models is incompatible with NAT acceleration and has negative impact on the entire network WAN throughput. On HND ARMv8 routers OpenVPN can reach 200-250Mbps, but the devices not using the VPN tunnel are unrestricted. WireGuard can reach 300-350Mbps, but the devices not using the VPN tunnel also become restricted. I can test and verify that on my AX86U router, but a bit later after official 388 is released.

Thanks that's very helpful to know. This way, I can track the releases to estimate the possibilities and weigh the pros and cons of perhaps getting an AX86U - I only have/need 300 Mbps on my WAN, and the AX58U is delivering a max of about 65 Mbps with OpenVPN on surfshark, a ceiling that would nice to raise (interesting the max speed has risen from 40 Mbps over the last couple Merlin releases).

 

[Tech9]

AX58U is delivering a max of about 65 Mbps with OpenVPN

This is about right. AX86U has ARMv8 + AES CPU and can do >200Mbps on OpenVPN, but it also depends on the server you connect to. I haven't checked lately the OpenVPN performance, but I remember 384 code going as high as 260Mbps on AC86U router. 386 code was never going above 170Mbps for some reason. Still, it will be >2 times faster compared to what AX58U can do. Careful with all-network VPN idea though, because it may limit your own Internet experience to unacceptable levels. Most if not all commercial VPN servers are well known.

 

[L&LD]

Not really. You seem to think anything under 300.00 bucks is junk. News flash for you 90% of people will never spend that for a router it's simply not necessary. Another news flash 90% of Asus router owners never heard of RMerlin. Just stop being negative about lower priced routers there is nothing wrong with them they get the job done period. I seem to remember Tim Higgens warning you about your attitude on these forums it's getting old.

Master of assumptions, huh?

 

[Deleted member 66500]

Careful with all-network VPN idea though, because it may limit your own Internet experience to unacceptable levels.

Yes, I use VPN Director (yet another reason to use merlin) to only use the VPN on my personal laptop and iPad (while corporate laptop uses its own VPN anyway) - while the streaming and other less-sensitive data flows through the ISP per Director's policies I set up. So not using all-network VPN, and with 65 Mbps ceiling with OpenVPN, things are chugging along fine for me on the laptop to me honest - so even raising that ceiling is arguably a "nice to have", but an aspiration I do have with Wireguard.

As a side note, I see that many folks don't realize that latency is much more important network QoS criteria for general experience than raw bandwidth, once the bandwidth is above ~ 100 Mbps or 200 Mbps, or so. I solved that latency problem with a RP-AX56 node in my home.

 

[Stpa67]

I had to do a power button reboot today because the gui went unresponsive when i looked at the Traffic Analyzer. Router worked fine it was just the gui that seemed to be locked up so i cut power to reboot it.

I think the uptime was since since the end of july like 45 days or or so. I also had some "shadow" entries in the Network Map client list that lingered for a week or so.

The System Log shows nothing but i had a usb-disk connected for a few days and memory has been at +90% usage.

It could be that if i only had waited for a bit longer the gui might have come back since the router worked fine, devices connected and no internet problems but i think i make a post about it anyway if anyone else encounters it.

Otherwise since this fw upgrade it has been rock solid

 

[asusrunshot]

Upgraded my RT-AX58U an hour ago, went from merlin 386.7 + flexcos so i had to install 45898 before installing 49674.

Dirty upgrade merlin 386.7 -> Asuswrt 386.45898 -> Asuswrt 386.49674 went well but i still did a factory reset (i am on vacation so i got the time ). One hour in, the only thing i noticed is the 49674 gui feels very very snappy, could be the factory reset though.

I use:

AiProtection
Adaptive QoS
DDNS
OpenVPN

Also activated all mu-mimo (ac+ax mu mimo) and ax features (dl/ul ofdma +mu-mino + wpa2/wpa3 authentication) despite having only ac clients. I also use fixed wifi channels, 11 on 2.4G (i got 15-20 2.4G wifis around me) and 36 on 5G (living 9 km from the airport).

Hello

Can you tell me if the adaptive qos is working correctly on your router? I have an ax58U that doesn't really function well.

Can you run a command through ssh and show me the output?

tc -s qdisc show

 

[Stpa67]

Adaptive qos has worked well for me. The output of the command generated 370 lines of text so i attached a textfile.

 

[asusrunshot]

Adaptive qos has worked well for me. The output of the command generated 370 lines of text so i attached a textfile.

Thank you!
Seems like you also have a lot of direct_packets_stat that bypassed the queue.

I was trying to get to the bottom of the problem and it seems to me that the trend micro module might have some issues with unmarked packets going to direct queue.

For e.g. in my case I set my upload bandwidth as 18 Mbps. Without editing anything my router gives me 23-24 Mbps which is what my ISP (Xfinity) gives me and it causes ping spike. However, when I add another filter on top I get limited to 18 Mbps with no ping spike like I expect.

Thank you for verifying that it is not a one-off hardware unit problem and is across multiple models and devices.

 

[Stpa67]

Thank you!
Seems like you also have a lot of direct_packets_stat that bypassed the queue.

I was trying to get to the bottom of the problem and it seems to me that the trend micro module might have some issues with unmarked packets going to direct queue.

For e.g. in my case I set my upload bandwidth as 18 Mbps. Without editing anything my router gives me 23-24 Mbps which is what my ISP (Xfinity) gives me and it causes ping spike. However, when I add another filter on top I get limited to 18 Mbps with no ping spike like I expect.

Thank you for verifying that it is not a one-off hardware unit problem and is across multiple models and devices.

I don't have much knowledge of the inner workings of asus adaptive qos but i read the qdisk output file and from what i can see it has the following queues:

???
---
qdisc pfifo_fast 0: dev eth0
qdisc pfifo_fast 0: dev eth1
qdisc pfifo_fast 0: dev eth2
qdisc pfifo_fast 0: dev eth3

???
---
qdisc htb 1 - x: dev eth4
qdisc sfq 2 - x: dev eth4

???
---
qdisc pfifo_fast 0: dev eth5
qdisc pfifo_fast 0: dev eth6

???
---
qdisc htb 1 - x: dev br0
qdisc sfq 2 - x: dev br0

VPN?
-----
qdisc pfifo_fast 0: dev tun21

Pfifo i guess is first-in-first-out possibly with a buffer, htb and sfq i guess are hierarchical token bucket and stochastic fair queing but do you have any idea what the interfaces, eth0...eth6, br0 coresponds to? (lan, wlan, wan, vpn ...)

 

[asusrunshot]

I don't have much knowledge of the inner workings of asus adaptive qos but i read the qdisk output file and from what i can see it has the following queues:

???
---
qdisc pfifo_fast 0: dev eth0
qdisc pfifo_fast 0: dev eth1
qdisc pfifo_fast 0: dev eth2
qdisc pfifo_fast 0: dev eth3

???
---
qdisc htb 1 - x: dev eth4
qdisc sfq 2 - x: dev eth4

???
---
qdisc pfifo_fast 0: dev eth5
qdisc pfifo_fast 0: dev eth6

???
---
qdisc htb 1 - x: dev br0
qdisc sfq 2 - x: dev br0

VPN?
-----
qdisc pfifo_fast 0: dev tun21

Pfifo i guess is first-in-first-out possibly with a buffer, htb and sfq i guess are hierarchical token bucket and stochastic fair queing but do you have any idea what the interfaces, eth0...eth6, br0 coresponds to? (lan, wlan, wan, vpn ...)

Yea br0 corresponds to the bridge of all local interfaces and eth4 is the wan. So when I was looking at your file output I looked at br0 and eth4 to see if you also had some unclassified packets that missed the HTB and SFQ thus going straight to pfifo_fast. On my routers I have a large amount of packets that go to pfifo_fast when using "speedtest.net". Thus manually setting my upload speed to 18 Mbps doesn't work since the packets going to direct_queue (fifo_fast) are not rate limited.

For now I have created a workaround with ASUS-Merlin and custom jffs script such that these direct_packets are redirected out of the fifo_fast and into htb. With this workaround my rate limited 18 Mbps setting actually holds true.

 

[Stpa67]

Yea br0 corresponds to the bridge of all local interfaces and eth4 is the wan. So when I was looking at your file output I looked at br0 and eth4 to see if you also had some unclassified packets that missed the HTB and SFQ thus going straight to pfifo_fast. On my routers I have a large amount of packets that go to pfifo_fast when using "speedtest.net". Thus manually setting my upload speed to 18 Mbps doesn't work since the packets going to direct_queue (fifo_fast) are not rate limited.

For now I have created a workaround with ASUS-Merlin and custom jffs script such that these direct_packets are redirected out of the fifo_fast and into htb. With this workaround my rate limited 18 Mbps setting actually holds true.

What happens if you run speedtest at the same time as you stream?

I was thinking if there's only one queue active there's no need for qos to kick in at all but if there are >=2 queues (a competition for bandwidth) then qos has to kick in? or am i thinking wrong?

 

[asusrunshot]

What happens if you run speedtest at the same time as you stream?

Speeds still exceed the manually set 18 Mbps limit on upload if I don't use my custom script.

On a different router AC3100, I see that packets on the upload interface are marked as download completely bypassing the queue and into fifo. But for this thread AX58U still has the problem of exceeding speed limits due to direct_packets.

 

[Stpa67]

Speeds still exceed the manually set 18 Mbps limit on upload if I don't use my custom script.

On a different router AC3100, I see that packets on the upload interface are marked as download completely bypassing the queue and into fifo. But for this thread AX58U still has the problem of exceeding speed limits due to direct_packets.

I gave it a test and set different ul/dl limits (18/18, 80/80, 90/90) in qos settings and i get around those ul/dl speeds when running speedtest from my laptop via wifi.

When i run the speedtest on the qos page on the router i get the max speed i get from my isp no matter what i set the qos limits to.

Since i get around 94.5/94.5 from my isp i set the qos limit to 90/90.

 

[asusrunshot]

I gave it a test and set different ul/dl limits (18/18, 80/80, 90/90) in qos settings and i get around those ul/dl speeds when running speedtest from my laptop via wifi.

When i run the speedtest on the qos page on the router i get the max speed i get from my isp no matter what i set the qos limits to.

Since i get around 94.5/94.5 from my isp i set the qos limit to 90/90.

I used the waveform bufferbloat test and AdaptiveQoS increases my upload bufferbloat to +100 ms. With no QoS my upload bufferbloat is +10 ms. I guess Comcast is probably using DOCSIS PIE on upload and ASUS QoS messes up somewhere on the upload.

 

[Stpa67]

I did some tests with ms edge and waveforn bufferbloat test.

First of all friday afternoon is not the ideal time to test, gf is facebooking, phones are online, appletv online, and an android tablet is also online and my isp is probably at >=100% streaming and gaming, and probably even waveform bufferbloat has many testers so everything is shaky

I did the following tests with ms edge (see attached pics):

adaptive qos off, aiprotection off
adaptive qos off, aiprotection on
adaptive qos on, aiprotection off
adaptive qos on, aiprotection on

I repeat this is the worst time to test bufferbloat, maybe the result are influenced by my own network load and most probably the of my isp and internet in general but this is what internet is at my house on a friday afternoon.

 

[Stpa67]

I just activated wireless mac filter to only allow my family's devices to connect to the wifi.

The router security features i use are:

AiProtection (Malicious Site Blocking, IPS, Infected Device Blocking)
OpenVPN (I use this on my mobile on public WiFi's)
Wireless MAC filter (Allow only my own devices on the wifi)

Anyone else use the builtin security features?