QOS features cutting speeds and making no difference in bufferbloat?

[Yowie]

I have XFINITY gigabit, ASUS RT-BE86U, Merlin firmware, hard wired to PC, Download speeds are 1200mbps and 40mbps upload.

Whenever I pick a QOS option, they all halve my internet speeds to around 500mbps download on speed tests? I've even tried Cake with default and cable docsis selected and the bufferbloat tests don't even improve, I get a C rating from https://www.waveform.com/tools/bufferbloat with the settings in the image. I see the download speed peak at 1200mbps for 1-2 seconds, then it immediately drops to 500mbps, but I still get a C rating with ping spiking to 500ms, like the QOS is delayed in kicking in.

I tried a basic bandwidth limiter across the entire DHCP ip range, same issue.

Is it because QOS disables hardware NAT and I'm forced to 500mbps max? Can I fix bufferbloat without 50% cost of my speed?

 

[dave14305]

You’re screwed. Only Adaptive QoS keeps HW acceleration enabled, but Adaptive QoS is believed to be broken by Asus on BE routers.

 

[Yowie]

You’re screwed. Only Adaptive QoS keeps HW acceleration enabled, but Adaptive QoS is believed to be broken by Asus on BE routers.

Which routers does it work on? I can still return this router. Does it work properly on RT-AX86U?

 

[dave14305]

Which routers does it work on? I can still return this router. Does it work properly on RT-AX86U?

If you want reliable QoS/SQM for gigabit, you should instead look for an x86-based machine to be the router (e.g. running OpenWrt with SQM), and use a wireless router as an access point only.

Asus lost their way with QoS some time ago. You really only need QoS for your asymmetric upload speed.

 

[GJJ]

Have you run the speed tests without QOS enabled to get a baseline?

I don't know if it is still done, but in the past, some cable connections would be allowed to run full speed before being throttled by the provider. I don't know if it is true or not, but the idea on the internet was to make speed tests look good and keep customers happy. I think the truth is for a better customer experience for general browsing and small downloads, rather than max speed of multi-GB downloads. Max cable speeds have improved over time, and I'm guessing the test sizes for speed tests have increased in size to keep the test running for a suitable amount of time for a more accurate result.

I'm running an RT-BE92U and switched Adaptive QOS on after the latest FW release. I can't truly tell you that it works, because my network isn't that busy, but I can say that it hasn't crippled my network. My kids use video chat and online games and no one has complained of any issues. I just do boring things where latency doesn't matter, like web browsing and streaming. My internet connection is 250/25 fibre, so much slower than yours and half the speed to which your connection is getting throttled to.
After switching it on, I did notice that browsing was slightly dull. I didn't measure it, so I could be just imagining it. Other than that, I've not noticed anything else.

I ran the Bufferbloat test you linked and got an A, with a download speed at 252 and an upload speed at 23, which is to expected for my connection.

Next week my connection is getting upgraded to 750/50, still not as fast as your gigabit connection but closer! I'll re-run the test after the upgrade to see if I get throttled to 500Mbps.

 

[Yowie]

Have you run the speed tests without QOS enabled to get a baseline?

I don't know if it is still done, but in the past, some cable connections would be allowed to run full speed before being throttled by the provider. I don't know if it is true or not, but the idea on the internet was to make speed tests look good and keep customers happy. I think the truth is for a better customer experience for general browsing and small downloads, rather than max speed of multi-GB downloads. Max cable speeds have improved over time, and I'm guessing the test sizes for speed tests have increased in size to keep the test running for a suitable amount of time for a more accurate result.

I'm running an RT-BE92U and switched Adaptive QOS on after the latest FW release. I can't truly tell you that it works, because my network isn't that busy, but I can say that it hasn't crippled my network. My kids use video chat and online games and no one has complained of any issues. I just do boring things where latency doesn't matter, like web browsing and streaming. My internet connection is 250/25 fibre, so much slower than yours and half the speed to which your connection is getting throttled to.
After switching it on, I did notice that browsing was slightly dull. I didn't measure it, so I could be just imagining it. Other than that, I've not noticed anything else.

I ran the Bufferbloat test you linked and got an A, with a download speed at 252 and an upload speed at 23, which is to expected for my connection.

Next week my connection is getting upgraded to 750/50, still not as fast as your gigabit connection but closer! I'll re-run the test after the upgrade to see if I get throttled to 500Mbps.

Yes, I ran baseline tests, there's no difference.

I'm just going to return the BE92U, as the QOS options don't even work and it's slower to load pages than my edgerouter x running openwrt. Edgerouter with openwrt loads pages nearly instantly for me, the asus router takes 1-2 seconds more to load the same page, even with merlin. I tested this with page load extension in chrome. And for whatever reason, the BE92U has some intermittent issue that it gets stuck loading pages sometimes for me, will sit spinning for 10 seconds, then finally load the page. Happens like 1 out of 30 page loads on a bunch of different domains, tried a new DNS, same issue. Think it might be something to do with the dns caching on the router.

Kind of sad how a $50 router released 10 years ago outperforms a modern day $250 router, lol. The only real benefit of the ASUS router was the wifi signal is stronger and faster than my access point setup.

Edit: Seems Chrome was the culprit for the slower load times and page stalling. Clearing cache seems to have fixed it. After testing both again, the ASUS router is just as fast as the Edgerouter. Probably going to keep it now.

 

[Spartan]

Yes, I ran baseline tests, there's no difference.

I'm just going to return the BE92U, as the QOS options don't even work and it's slower to load pages than my edgerouter x running openwrt. Edgerouter with openwrt loads pages nearly instantly for me, the asus router takes 1-2 seconds more to load the same page, even with merlin. I tested this with page load extension in chrome. And for whatever reason, the BE92U has some intermittent issue that it gets stuck loading pages sometimes for me, will sit spinning for 10 seconds, then finally load the page. Happens like 1 out of 30 page loads on a bunch of different domains, tried a new DNS, same issue. Think it might be something to do with the dns caching on the router.

Kind of sad how a $50 router released 10 years ago outperforms a modern day $250 router, lol. The only real benefit of the ASUS router was the wifi signal is stronger and faster than my access point setup.

I never enable QoS on my BE92U nor on any router for that matter. We live in a world where most people have 500+ MBPS speeds so even if someone was downloading a big file there is still enough bandwidth for other devices to enjoy fast internet.

I use NextDNS and my page loading speeds are instantaneous.
On a side note, what page load extension for chrome do you use? I see there are 2 when I searched online.

 

[Yowie]

I never enable QoS on my BE92U nor on any router for that matter. We live in a world where most people have 500+ MBPS speeds so even if someone was downloading a big file there is still enough bandwidth for other devices to enjoy fast internet.

I use NextDNS and my page loading speeds are instantaneous.
On a side note, what page load extension for chrome do you use? I see there are 2 when I searched online.

Page load time - Chrome Web Store

Displays page load time in the toolbar

chromewebstore.google.com