RT-AX86U PRO page now up ...

[Spartan]

You're trying to improve your wired latency?

Not going to happen. That's up to your ISP.

Right, thanks for letting me know before I waste my money.

 

[bdub76]

the only reason I wanted to get the RT-AX86U is that you keep saying your latency is lower. This is my latency now (using the best Monoprice LAN cables BTW if you remember I had asked you before for a suggestion, don't waste your money, the latency didn't improve with those cables)

View attachment 44471

I initially wanted to get the RT-AX86U based on your recommendation and low latency but that all changed with the announcement of the PRO model. Still not sure if it's worth the upgrade.

You need to manage bufferbloat with fq codel or cake.

What Can I Do About Bufferbloat? - Bufferbloat.net

www.bufferbloat.net

Here's the test for it:

Bufferbloat and Internet Speed Test - Waveform

Is bufferbloat causing issues with your internet connection? Run this test to find out.

www.waveform.com

I get an A+. But I'm only running the AX86U as an AP. My APU2 is running fq codel.

 

[bdub76]

Without testing, there is no way to predict if the RT-AX86U Pro (or not) is superior to the RT-AX88U. Same for the GT-AXE16000.

Doesn't sound like you need to change your hardware today. You are not stating anything wrong with the current setup.

2x of the identical model may prove worthy though. But, at those prices, it should.

True. I got the r7000p, and it was marginally better than the r7000. I should have bought the r7800.

 

[Tech9]

You need to manage bufferbloat with fq codel or cake.

He doesn't need to do anything with this ISP connection.

Here's the test for it:

None of those online tests are accurate and showing you the real need for QoS. While trying to get better numbers with this bufferbloat online test you may hurt your everyday Internet experience.

 

[hancox]

QoS, as it stands from ASUS, on the AX86 (and elsewhere) is much MUCH more hampered by broken code, than hardware. At this point, i would lean towards not suggesting it for anyone, unless you have a true, verifiable use case.

Evidence: https://www.snbforums.com/threads/f...pt-for-adaptive-qos.75620/page-10#post-765976

 

[Tech9]

This AdaptiveQoS with NAT acceleration enabled is a mystery box. Cake is better, but needs CPU.

 

[bdub76]

None of those online tests are accurate and showing you the real need for QoS. While trying to get better numbers with this bufferbloat online test you may hurt your everyday Internet experience.

This is nonsense. The online tests are good enough for most home users. And his every day use case isn't going to come close to his ISP speeds unless he's seeding torrents.

He can run the more robust tests such as:

Overview — Flent: The FLExible Network Tester

But it takes a bit to properly set it up.

His problem is his upstream latency is ridiculously bad. He's going to have to give up some speed to mitigate it, perhaps 10-20%. This isn't going to hurt his Internet experience. I cut my upstream of 10 to 8 without any problems.

QoS on this wireless router won't fix this problem. He needs cake or fq codel. His modem supplied by his provider should have PIE on it. But it's not going to move it much.

 

[Tech9]

This is nonsense. The online tests are good enough for most home users.

You know better. They are good enough to confuse home users. Especially ones who don't know what bufferbloat is and when it happens, eventually. Most run QoS with no need as a result.

 

[bdub76]

You know better. They are good enough to confuse home users. Especially ones who don't know what bufferbloat is and when it happens, eventually. Most run QoS with no need as a result.

Blindly turning on QoS without knowing what you're doing is going to make things worse. However, Cake is well done. Turning on Cake shouldn't be a problem.

I messed with the OFDMA types this week, and it made my latency much worse. I went back to the default value, and it fixed my latency problem.

 

[RMerlin]

Cake is well done. Turning on Cake shouldn't be a problem.

Turning on Cake requires disabling NAT acceleration, which will seriously limit your maximum throughput if you have an Internet connection faster than 300 Mbps.

 

[Tech9]

Turning on Cake shouldn't be a problem.

@Spartan has an Asus home router. From what's available in Asuswrt (Merlin) the best setting for Gigabit ISP line is no QoS. Turning on Cake will cut WAN-LAN throughput to 1/3 and fq_codel is not available anymore (since 386 code base, if I remember correctly), except in custom script experiments. Adaptive QoS is struggling with Gigabit line and doesn't work very well in general. Again, custom script was an attempt to fix it. Traditional QoS and Bandwidth Limiter are also incompatible with NAT acceleration. When you give advice to someone make sure it applies to the situation and is possible on the hardware in question.

 

[Spartan]

You need to manage bufferbloat with fq codel or cake.

What Can I Do About Bufferbloat? - Bufferbloat.net

www.bufferbloat.net

Here's the test for it:

Bufferbloat and Internet Speed Test - Waveform

Is bufferbloat causing issues with your internet connection? Run this test to find out.

www.waveform.com

I get an A+. But I'm only running the AX86U as an AP. My APU2 is running fq codel.

here is my score

 

[hancox]

His problem is his upstream latency is ridiculously bad. He's going to have to give up some speed to mitigate it, perhaps 10-20%. This isn't going to hurt his Internet experience. I cut my upstream of 10 to 8 without any problems.

QoS on this wireless router won't fix this problem. He needs cake or fq codel. His modem supplied by his provider should have PIE on it. But it's not going to move it much.

How is his speedtest / bufferbloat test reflective of "ridiculously bad" latency?

 

[Tech9]

@Spartan just trusted someone changing the router will give him lower latency.

 

[L&LD]

Nobody has stated that. Reading the entire thread is important.

 

[bdub76]

Turning on Cake requires disabling NAT acceleration, which will seriously limit your maximum throughput if you have an Internet connection faster than 300 Mbps.

Now, that's a problem.

 

[bdub76]

here is my score

View attachment 44521

Looks good to me. I wouldn't worry about it.

 

[bdub76]

@Spartan has an Asus home router. From what's available in Asuswrt (Merlin) the best setting for Gigabit ISP line is no QoS. Turning on Cake will cut WAN-LAN throughput to 1/3 and fq_codel is not available anymore (since 386 code base, if I remember correctly), except in custom script experiments. Adaptive QoS is struggling with Gigabit line and doesn't work very well in general. Again, custom script was an attempt to fix it. Traditional QoS and Bandwidth Limiter are also incompatible with NAT acceleration. When you give advice to someone make sure it applies to the situation and is possible on the hardware in question.

I misread his post. I was wrong. I'm not a fan of QoS other than latency mitigations with fq_codel, cake, and PIE. PIE now comes with DOCSIS 3.1 cable modems, so off the line, I see much lower latency. The worse latency I have experienced in the southeast with Comcast. High latency (50-100ms) off the line with a low 60/10 setup between 2014-2018.

If he's already getting A (<30ms) latency on the wire, then he can't do much without offloading to another device with very good CPUs (2+ GHz) and good nics based on other responses. And based on what's being said for Merlin, it's not worth using the ax86u to try to get it down a little less. I'm not sure it's worth the cost. Might be easier to run ethernet cable.

Consumer gear in my experience doesn't play well with Gigabit speeds. It's why I upgraded and then downgraded. It's getting better. But it's not there yet unless you basically turn almost everything off. And in my tests, I'm getting more latency on top of my line from my ax86u running the latest OEM firmware than my r7000p running the last Kong DD-WRT release. And this is in AP mode for both. I'm getting almost +10ms with default settings on the ax86u and less than +5ms on the r7000p. But I'm getting much better speeds in the room farthest away from my AP. I'm on the fence as to whether the upgrade was worth it, but I only have 4 ax devices, so it might be too early to say.

 

[muzykcb]

RT-AX86U PRO
Processor BCM4912,
Radio 5GHz BCM6715,
Rangeboots+,
USB 3.0 / 2.0

RT-AX86U
Processor BCM4908,
Radio 5GHz BCM43684,
Rangeboots,
USB 3.0 / 3.0


Is it worth changing to a newer model?

 

[Tech9]

If you already have an AX86U - no.