FYI/A: Bufferbloat 101

xiaofan · Oct 13, 2023

xiaofan said:
Wireless result today is bad. Main problem is with upload.

Same test setup as before: no heavy internet usage, only two kids using Youtube.
Singtel ZTE ONT -- Asus RT-AX86U router
Wireless client -- Acer Swift 3 Windows 11 laptop (2021 model) with built-in Intel AX201 WiFi 6 adpter, about 4m away from router

Hmm, wireless test results are not consistent.

Now I am getting A.
https://www.waveform.com/tools/bufferbloat?test-id=330b8146-15da-4f08-9757-b53c32679086

This is immediately after a very good OOkla wireless speed test result.

xiaofan · Oct 14, 2023

Wired test results are more consistent. Today I got an A+ with the Asus RT-AX86U router and Acer Windows 11 laptop using external Ugreen USB 3.0 Gigabit Ethernet adapgter.

phayze · Oct 15, 2023

Sometime i get A, sometime got B. I think due to load balancing. I always see that the ping for M1 is much higher than starhub as indicated in the pfsense gateway when under load.

https://www.waveform.com/tools/bufferbloat?test-id=6f788b5c-f859-4e18-a247-0defc8b68d98

xiaofan · Oct 15, 2023

People may get curious how a S$10 router can achieve with bufferbloat).

Here is the result using Linksys EA7500 v2 running latest OpenWRT 23.05.0 release. I actually got it free as someone threw it away the dustbin area (without power supply but my Linksys WRT1900AC power supply works fine with EA7500v2).

But we need to take the result with a gain of salt as the speed from waveform.com is slow for the wired connection, compared to OOkla Speedtest using the same test setup (wired conection, Linsys EA7500v2 router, Acer Swift 3 Windows 11 laptop 2021 model).

xiaofan · Oct 15, 2023

phayze said:
Sometime i get A, sometime got B. I think due to load balancing. I always see that the ping for M1 is much higher than starhub as indicated in the pfsense gateway when under load.

https://www.waveform.com/tools/bufferbloat?test-id=6f788b5c-f859-4e18-a247-0defc8b68d98

There are only a few people using Dual WAN Load Balancing/Failover with two ISPs in this forum. Thanks for posting the result and share your results.

In terms of raw performance, I believe Starhub is a bit better than M1.

lolkikkiklol · Oct 19, 2023

Hi all, been reading through the input here but unfortunately not quite familiar with the routers in general. I have been recording similar results for a long time.

I am currently on M1 1Gbps plan > Huawei Echolife HG8240H > Asus RT-AX55. Tested on a wired connection to my PC. M1 support says that the results are normal.

Enabling QOS to limit to 500Mbps does alleviate the issue but not ideal either, at best a B reading.

Appreciate if there are any inputs on why the readings are as such.

hwzlite · Oct 20, 2023

lolkikkiklol said:
Enabling QOS to limit to 500Mbps does alleviate the issue but not ideal either, at best a B reading.

Do you mean using asus "Adaptive QoS" ?
From your 1Gbps plan, how about from "Adaptive QoS > Bandwidth Setting > Manual Setting" , test play around from 95% (950Mbps) , 90% (900Mbps) and lower...etc to see which setting start to get better results.

xiaofan · Oct 20, 2023

hwzlite said:
Do you mean using asus "Adaptive QoS" ?
From your 1Gbps plan, how about from "Adaptive QoS > Bandwidth Setting > Manual Setting" , test play around from 95% (950Mbps) , 90% (900Mbps) and lower...etc to see which setting start to get better results.

No point trying any type of QoS with RT-AX55 (which does not support Merlin) or other Asus consumer routers with stock Asus FW.

Once you enable QoS or Adaptive QoS or Bandwidth Limiter, the line speed will drop below 400Mbps. I have tried on my more powerful Asus RT-AX86U and it is the same.

Some say Merlin FW may provide workable Adaptive QoS options. But I have not tried it myself.

Asus stock FW QoS is not useful for 1Gbps users in Singapore, probably marginally useful for 500Mbps users.

xiaofan · Oct 20, 2023

lolkikkiklol said:
Hi all, been reading through the input here but unfortunately not quite familiar with the routers in general. I have been recording similar results for a long time.

I am currently on M1 1Gbps plan > Huawei Echolife HG8240H > Asus RT-AX55. Tested on a wired connection to my PC. M1 support says that the results are normal.

Enabling QOS to limit to 500Mbps does alleviate the issue but not ideal either, at best a B reading.

Appreciate if there are any inputs on why the readings are as such.

I see the only option is to upgrade to more powerful Asus router and not to use QoS at all. You should be able to get A with RT-AX86U, RT-AX86U Pro and GT-AX6000 (or similar Asus routers with faster CPU).

xiaofan · Oct 20, 2023

If I read the posts in SNB Forum, Asus QoS is basically broken, or at least not so useful for 1Gbps connection or even 500Mbps connection.

In fact, in this forum, whenever people mentioned that their Asus routers could only reached less than 500Mbps for 1Gbps plan, my first suggestion is to ask them to disable any type of QoS/Adaptive QoS/Bandwidth limiter.

Reference: just two examples, there are many more.
https://www.snbforums.com/threads/asus-qos-does-nothing-help-wanted-rant.77065/page-2
https://www.snbforums.com/threads/should-i-use-qos-on-a-500mbps-connection.84649/

lolkikkiklol · Oct 21, 2023

Thanks to both for your input. I have tried playing with all three QoS settings before, and xiaofan is right. Any setting will cause effective speed to be capped at around 400-500Mbps range, even when the cap is set to a higher speed.

Funnily my friend receives an A result on an old Linksys router without QoS, but perhaps it is because he is on M1 500Mbps plan instead. Shall get a stronger router to test when the chance arises.

cyberet · Nov 19, 2023

imho I wouldnt be overly concerned over bufferbloat.
too many factors will affect the result.
Further to that, in our local context, not many ppl can overwhelm a 1gbps connection under normal usage

wavefoam test servers may not be able to saturate our 1gbps connection.
if your wifi speed is lower than your internet speed, you wont be able to saturate the connection either.

Most consumer routers are not powerful enough to support 1gbps QOS as it isn't hardware accelerated.

hwzlite · Mar 3, 2024

Update on my WIRELESS adventure into using fq_codel/ATF/AQL optimized MediaTek mt76 mac80211 driver on Linksys EA8100(OpenWRT) , configured as Dumb AP, and tweaked with "aql_txq_limit = 1500" setting :

Here are the wireless benchmarks (M1 500Mbps, DFS channel 64 @80Mhz, Wall = old HDB type):

elmariachi · Mar 5, 2024

Netduma R3 with 62.5% congestion control during DPI traffic.
Starhub 1Gbps plan.

hwzlite · Mar 5, 2024

elmariachi said:
Netduma R3 with 62.5% congestion control during DPI traffic.
Starhub 1Gbps plan.

Cool. How's the CPU utilization stats under load?

elmariachi · Mar 5, 2024

hwzlite said:
Cool. How's the CPU utilization stats under load?

Earlier firmware had massive cpu loads but the new 4.0.41 firmware have fixed all that. Still have alot of stuff to be fixed but the router has been working very well since the last update. Have had no issues with gaming and hitreg.

xiaofan · Mar 5, 2024

One of the problems of the testing using Waveform.com is that the server may plays a part as well. It is a pity we do not have capable servers like OOkla SpeedTest servers (ISPs and other service providers have good servers). So we may look at the results with a grain of salt.

The other issue is how mearningful the wireless test results are. Usually wired tests are recommended. However, many users will use wireless mainly for the clients (eg: mobile phones, tablets and laptops).

xiaofan · Mar 5, 2024

cyberet said:
imho I wouldnt be overly concerned over bufferbloat.
too many factors will affect the result.
Further to that, in our local context, not many ppl can overwhelm a 1gbps connection under normal usage

wavefoam test servers may not be able to saturate our 1gbps connection.
if your wifi speed is lower than your internet speed, you wont be able to saturate the connection either.

Most consumer routers are not powerful enough to support 1gbps QOS as it isn't hardware accelerated.

I think this is quite a valid comment. No need to be overly concerned about bufferfloat test results.

Newer generation of higher end ARM CPUs used by consumer WiFi Routers (like BCM4916 and Qualcomm IPQ9570 or better) can do 1Gbps QoS but then they are often paired with 10G capable routers. So they may be good for 1Gbps QoS but for sure not 10Gbps QoS. I do not know which CPU can do 10Gbps QoS...

On the other hand, I still carry out the tests from time to time to see if there are differences between different routers/APs used. Just for reference purpose.

xiaofan · Mar 5, 2024

xiaofan said:
On the other hand, I still carry out the tests from time to time to see if there are differences between different routers/APs used. Just for reference purpose.

1. Wireless bufferbloat test result: A

Singtel 1Gbps plan, OpenWRT + ZTE BE7200 Pro+ wireless AP, Acer Windows 11 laptop with Intel AX201 WiFi 6 adapter, 4m away.
https://www.waveform.com/tools/bufferbloat?test-id=2ca99903-2c3d-4198-859e-159ac6ad4535

2. OOkla SpeedTest done just before the bufferbloat test.

hwzlite · Mar 5, 2024

xiaofan said:
One of the problems of the testing using Waveform.com is that the server may plays a part as well. It is a pity we do not have capable servers like OOkla SpeedTest servers (ISPs and other service providers have good servers). So we may look at the results with a grain of salt.

The other issue is how mearningful the wireless test results are. Usually wired tests are recommended. However, many users will use wireless mainly for the clients (eg: mobile phones, tablets and laptops).

So far based on observation here from those who benchmarked from 10Gbps plan, seems that Waveform.com still has good enough bandwidth to cater those =< 2.5Gbps plan.

IMHO , wireless test will be still meaningful for someone who appreciates consistent real-time buffering/stuttering-free video/chat/stream/vod sesssions ( eg. Teams, Zoom, Google Meet, Whatsapp ..etc any1?), at least can be certain 99% is not localized issue if any of my home dweller nags about it.

FYI/A: Bufferbloat 101

High Supremacy Member

High Supremacy Member

Supremacy Member

High Supremacy Member

High Supremacy Member

Member

Master Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

Member

Senior Member

Master Member

Supremacy Member

Master Member

Supremacy Member

High Supremacy Member

High Supremacy Member

High Supremacy Member

Master Member