Posted: Sat Jan 26, 2019 7:25 Post subject: R9000 Wi-fi Issues
I have got a new R9000 this week and installed the latest available Kong build (r38100M). My LAN performance is quite good actually with reaching the max of 300 as my ISP provides but my Wifi doesn't work too well (find my settings attached).
On the 5Ghz band I get a max of around 160 and on the 2.4GHz one a max of about 90, whereas the rate is highly fluctuating between 20 and the max mentioned. In fact, when downloading a file, it mostly starts off quite OK and suddenly after a few seconds drops, after which it slightly increases and drop again. On my phones the average performance on 5GHz is about 55 and on 2.4Ghz more or less the same.
As for ping times without any load I get a stable 17 via LAN but on Wifi I get on average about 25-30 and during normal website usage this occasionally increases to 50 and even surges up to over 200 or with more load given also dropped packages.
Since the LAN performance is quite stable I assume there's a misconfiguration of my Wifi - could you please suggest better settings I can try?
Last edited by muhlba91 on Sat Jan 26, 2019 14:56; edited 1 time in total
@muhlba91
You may try enabling the Short Preamble...Enabling in this case doesn't mean required. _________________ NETGEAR R9000 DD-WRT v3.0-r45192 std (12/29/20) (ROUTER)
NETGEAR ORBI Stock FW V2.7.2.102 (MESH)
WRT3200ACM (BS Build Usually) (BACKUP)
I applied your settings and received rates in the 2.4GHz channel of a max of 140 now; however, the fluctuation is still high as in rates are dropping every few seconds to 30 as well as ping packets getting lost.
When it comes to the 5Ghz channel, I cannot connect to the network anymore. Also, I tried VHT80 as this was the width where I can connect but have a connection of about 7mbps and kinda like 90% dropped packages. It also seems as if the 5GHz restarts every few minutes - at least it disappears and reappears after 1 minute again in Wifi explorer tools.
As a note, I have disabled SFE and enabled QoS using HTB and FQ_CODEL.
Joined: 21 Jan 2017 Posts: 1782 Location: Illinois Moderator
Posted: Sun Jan 27, 2019 18:16 Post subject:
muhlba91 wrote:
@msoengineer thank you!
I applied your settings and received rates in the 2.4GHz channel of a max of 140 now; however, the fluctuation is still high as in rates are dropping every few seconds to 30 as well as ping packets getting lost.
When it comes to the 5Ghz channel, I cannot connect to the network anymore. Also, I tried VHT80 as this was the width where I can connect but have a connection of about 7mbps and kinda like 90% dropped packages. It also seems as if the 5GHz restarts every few minutes - at least it disappears and reappears after 1 minute again in Wifi explorer tools.
As a note, I have disabled SFE and enabled QoS using HTB and FQ_CODEL.
First, and foremost, what type of client is having these issues...chipset, drivers, settings? Are you running the most up to date driver on the client side if it's windows?
If you're running an iOS/Apple device then you're screwed. I boycotted Apple back in 2013 when they started charging ridiculous prices for repairs and try to upsell you to a new device rather than repair your existing device, not to mention the whole battery throttling issue and just the cost to repair their crap which is poorly designed from the get-go. Also, Kong is pretty fed up with Apple and they're shit-tastic driver support. Most Apple devices use broadcom chipsets and don't play nice with qca chipsets. their latest devices are using Intel chipsets and have issues as well. if you're asking about an Apple device and why you're having issues it's because it's Apple; change your router to a broadcom based chipset.
How far from the router are you trying to test things?
Your connection rate will drop down as a function of signal strength.
Does the 5ghz show up in the router GUI status, or does the dBm come up as zero? If it's showing up at 0, then you have a conflict in the channel you selected. You'll need to look at my signature to see channels and their allotted transmit power. _________________ FORUM RULES
Ah, if there‘s an issue with Apple devices in general it could be the cause because it‘s affecting all devices in our household but they are all Apple devices actually. In fact, only Broadcom routers will deliver the performance for Apple devices? Any good recommendation for this?
Since I‘m just about to test all functionalities, I test everything quite close to the router - between 10cm to max. 1m distance and even with about 4m distance I still get a connection strength of about 98% for 2.4GHz and 86% for 5GHz.
Regarding the dBm, I double checked this one and kept an eye on it: it’s steady at 30 even when my client sees the 5GHz as gone.
Thanks a lot for that info and your help! I really appreciate your effort.
Now my iOS performance is at least more or less on par with the MacBook and more stable (also 5GHz works now) one but still quite low compared to LAN: http://www.dslreports.com/speedtest/45219537.png (Bufferbloat rating increased now from B to A as well.)
Do you have any more hints on what I can try tweaking to get more throughput?
Joined: 21 Jan 2017 Posts: 1782 Location: Illinois Moderator
Posted: Sun Jan 27, 2019 20:24 Post subject:
muhlba91 wrote:
Thanks a lot for that info and your help! I really appreciate your effort.
Now my iOS performance is at least more or less on par with the MacBook and more stable (also 5GHz works now) one but still quite low compared to LAN: http://www.dslreports.com/speedtest/45219537.png (Bufferbloat rating increased now from B to A as well.)
Do you have any more hints on what I can try tweaking to get more throughput?
EDIT AT THE BOTTOM
First, try turning off Single user beamforming and see what happens.
You can also try playing around with the RTS threshold value.
BS made a change to 38381 so it's now a very low 500. You can start there and see if that helps anything- a setting that low is meant to help out if you have a lot of wifi devices (10+). I would try 1024, and then maybe 1500, and 2304. Test your speed on several devices under the same conditions for RTS value- meaning don't test one device in just one spot in the house. You have to tweak one thing at a time to see the affects and if they're positive or negative.
Also, Setup>Networking>port setup section. Change all TX queue lengths to 2. Setup>Basic Setup>Optional settings section- set the MTU to manual and 1472 to help reduce latency. If you are using IPv6, change the MTU value on that tab too to 1472. Leave MTU at 1500 on the Setup>networking tab.
Make sure to turn off Shortcut Forwarding Engine on the Setup>basic setup tab. _________________ FORUM RULES
I have disabled SFE and Single User Beamforming and got already over 200. After setting the RTS threshold value to 1024 but leaving the MTU at auto I now receive a stable throughput of 290 on MacBooks and 130 on iPhones.
The only thing I can't seem to get more stable is Bufferbloat. Using dslreports I always receive values between A (quite rarely though) and C, whereas I realized QoS settings of HFSC and SFQ were more stable for the Wifi performance than other combinations.
Anyhow, increasing the RTS threshold led to a drop for iPhones and setting the MTU made the rate drop for all devices. Also, I think the TX queue lengths of 2 are too short as ping times increased, so for now I left them at 1000 to monitor long time performance before trying to adapt them again.
Yeah I started having issues when I changed the TX Queue length from 1000 to 2. Since then I changed it back and I'm not experiencing weird drop out issues.
edit: the tx queue length at 2, and turning off short cut forwarding engine + making the MTU on the main page manual and 1472, seems the same when I do a speedtest.net test but just the overall feel, it feels "snappier." This is testing from my ethernet cabled desktop windows computer.
The reason why I was having weird drop out issues earlier was because I changed TX Queue Length from 1000 to 2 on my iphone and apparently I missed changing one of the 1000 fields to 2. Anyways all seems to be well now.
A side note, not sure if my internet was upgraded yesterday (I have Spectrum internet ultra), and I've been getting 450-460mbps download, and 18-23mbps upload via speedtest.net for the last 10 months, and just after I switched these TX queue length values and the short cut forwarding engine + manual MTU change, my download is now 465mbps, and my upload is now 48-60mbps. Coincidence?? I'm thinking my internet company upgraded my service in the area... but I don't remember getting any notification anything like that took place. Is it possible my upload speed increased 250-300% because of this settings change?
One more question for anyone...
What firmware are you on? I am still on the 12/26/2018 Kong, and I am debating on trying Brainslayers build again. Anyone try the latest one from BS? Reliable? Is updating through the web GUI working again yet?
Joined: 21 Jan 2017 Posts: 1782 Location: Illinois Moderator
Posted: Tue Jan 29, 2019 16:09 Post subject:
muhlba91 wrote:
The only thing I can't seem to get more stable is Bufferbloat. Using dslreports I always receive values between A (quite rarely though) and C, whereas I realized QoS settings of HFSC and SFQ were more stable for the Wifi performance than other combinations.
The Q.o.S. on all dd-wrt builds is simply flakey on ISP speeds >100Mbps for DL. Unless you have Fiber, you will never be able to get it completely perfect and get an A+.
Coax and DSL have too much latency in them to get perfect bufferbloat scores. What you do need to do is run speedtests using incognito/private mode in your browser of choice. This assures any extra extensions and other stuff is not running adding to the latency.
More importantly, if you are not getting consistent A's, it means you need to tweak your downlink and uplink speeds to be 92.5% of your actual provisioned speeds from your ISP. Also make sure all 4 tcp packet priority boxes are checked (click on my QOS link in the signature)
Realize, most ISP's will over provision the speed 15%-20% to assure you "get good numbers." I would run several tests without Q.o.S., first, to see what your ISP actually gives you. Then set your downlink/uplink based on those averages at 92.5%.
muhlba91 wrote:
Anyhow, increasing the RTS threshold led to a drop for iPhones and setting the MTU made the rate drop for all devices. Also, I think the TX queue lengths of 2 are too short as ping times increased, so for now I left them at 1000 to monitor long time performance before trying to adapt them again.
Here is an article, or two, about why to use shorter RTS threshold. But the TL/DL is that 2346 is too high a value and the RTS/CTS will never kick "ON." This is why I think BS has changed it to 500 instead...Because of all the iPhone/Apple shit not working on QCA. I am using 1024 for RTS and have no issues with connections on any of my wifi, but then again I don't/won't own Apple products >2013, so you'll have to play around and find a good value between 500 and 2304 that works for your iOS/Apple crap.
yodog wrote:
The tx queue length at 2, and turning off short cut forwarding engine + making the MTU on the main page manual and 1472, seems the same when I do a speedtest.net test but just the overall feel, it feels "snappier." This is testing from my ethernet cabled desktop windows computer.
When you run the test, take a look at the results page and the actual latency graphs. You will see that my TX Que length = 2 "hack" will lower your latency by a shit ton on Coax & DSL.
yodog wrote:
The reason why I was having weird drop out issues earlier was because I changed TX Queue Length from 1000 to 2 on my iphone and apparently I missed changing one of the 1000 fields to 2. Anyways all seems to be well now.
This issue is a phenomena of RTS/CTS being set too high on the router for iOS/Apple devices...see above comments...
yodog wrote:
A side note, not sure if my internet was upgraded yesterday (I have Spectrum internet ultra), and I've been getting 450-460mbps download, and 18-23mbps upload via speedtest.net for the last 10 months, and just after I switched these TX queue length values and the short cut forwarding engine + manual MTU change, my download is now 465mbps, and my upload is now 48-60mbps. Coincidence?? I'm thinking my internet company upgraded my service in the area... but I don't remember getting any notification anything like that took place. Is it possible my upload speed increased 250-300% because of this settings change?
Yes they updated your speeds most likely to be competetive with other local ISP offerings. Comcast did their speed bumps last fall.
yodog wrote:
One more question for anyone...
What firmware are you on? I am still on the 12/26/2018 Kong, and I am debating on trying Brainslayers build again. Anyone try the latest one from BS? Reliable? Is updating through the web GUI working again yet?
Always update using incognito/private mode to avoid the gui update hanging. I am running 38381 and it's mostly fine except for flakey DL on Q.o.S. that is a bug on IPQ and Annapurna SOC's. I think that BS 38253 was a little bit snappier. The builds in between have driver gremlins. _________________ FORUM RULES
if we use the dd-wrt like a wifi client with 2 virtual access points do we have to change some values for beter performance?
The Rts threshold value does it have to be the same in client and the virtual access points?
Joined: 21 Jan 2017 Posts: 1782 Location: Illinois Moderator
Posted: Fri May 15, 2020 12:41 Post subject:
I use a vap on 2.4 and 5 and use the same rts threshold values as the main radio. I've never tried to run the VAP with different settings from the main radio..I have no idea what might happen if you use differnt settings, but I think it's best to keeps things the same so you don't accidentally create gremlins. You want to keep WMM turned on for the VAP to make sure your clients can connect at N/AC speeds.
outside that, there aren't any other different settings that I think would cause an issue. _________________ FORUM RULES