I am running into an issue with my newly bought Dir-860l Rev B1. It is showing very poor wifi performance. Both AC and N are very poor, but let's focus on N first:
This is at 40mhz wide channels in an area with relatively little noise. I just moved to a new appartement. I my previous appartement I was using an Archer C7 in a very noisy environment (40+ APs visible) and I was able to get 70-100 Mb/s at 20mhz channel widths. More importantly, bufferbloat was under control, both pre-airtime fairness patch and after this patch was applied. The above test was done with just one device connected to the access point, so there was no traffic other than that. My connection is 500 Mb/s / 500 Mb/s, which I am able to achieve just fine through an ethernet cable. My ISP uses PPPoE in case that makes any difference. Does anybody happen to know how I can get the wifi bufferbloat under control?
Update: The poor wifi performance can also be observed on D-link stock firmware. Throughput is much higher and much more stable on the stock firmware, though.
I get around the same speeds but without the bufferbloat, see the second part of this post.
To take wired connection problem out of the equation could you do a speedtest with a pc directly connected to the modem and one with a pc directly connected to the router?
Also, do you have SQM QoS enabled?
I don't have a separate modem. The router is directly connected to the FTU (Fiber connection) through PPPoE. This is a test with the PC directly connected through a cable to the router:
As you can see, bufferbloat is fine wired (I am not using SQM, since that crashes my router, even with fq_codel, it's an outstanding issue with Mediatek socs). And WAN speeds are more than enough for my wifi connection.[quote="diizzy, post:3, topic:2612, full:true"]
Please do a wireless survey of your 5Ghz surroundings.https://play.google.com/store/apps/details?id=com.netgear.WiFiAnalytics is a useful tool for instance if you have a device running Android.
[/quote]
The issue is with 2.4ghz. Sorry for the confusion. 5ghz is performing just fine for me. Below you will find a screenshot of our wifi surroundings:
The two SSIDs at channel 1 with width 20mhz are our wifi connections. A primary connection and a secondary connection for guests. The tests posted in the OP were done with just one SSID set up, just to make sure that wasn't causing the issues.
I managed to find the culprit of my bufferbloat. I remember that I am running dslreports.com speedtest with non-default settings, namely 32 upload and 32 download streams. So what I did, I started a continuous ping (from the client) to the router to measure latency while stressing the connection through multiple iperf3 tests. The results are presented below:
As we can see, performance is only very poor when wifi (I only tested 2.4ghz for now) uploads are involved (which was also the case with dslreports) in combination with a high number of connections. Could somebody please try to replicate these results by doing similar iperf tests, or alternatively by running the dslreports speedtest with a high number of upload connections? And furthermore, do you guys believe this is a bug that can be fixed?
Edit: Not sure if it is related, but I also found the higher pings in my wired tests when the client was uploading interesting. I would expected similar results in either wired direction, since both are using ethernet connections.
For me, it turned out windows is to blame. My previous tests with the Archer C7 used a wan connection with just 40 Mbit/s upload speed, so the wifi connection was never the bottleneck, so the bufferbloat issues never showed. My current 500 Mbit upload connection does show the bufferbloat issues on wifi, since wifi is obviously the bottleneck there.
Repeating the exact same wifi tests on the exact same hardware with Ubuntu showed an A bufferbloat and much higher and more importantly stable speeds. Around 90 Mbit down and 70 Mbit up. Very happy with that performance.
That was the same for me. Please try running the same test with a recent Linux distribution running on the client. Do you still experience the same issue then?
Took me some time but I was finally be able to do some quick testing. Only chose to do the WLAN to LAN tests, so WLAN > LAN with 1 stream and WLAN > LAN with 32 streams.
Other variables: router running my r3636, laptop with an QCA61x4 connecting at 144 Mbps, 4 neighboring 2.4 GHz networks.
The quick and dirty preliminary results are:
WLAN > LAN, 1 stream:
Ping statistics for 192.168.0.1:
Packets: Sent = 157, Received = 157, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 2ms, Maximum = 540ms, Average = 22ms
WLAN > LAN, 32 streams:
Ping statistics for 192.168.0.1:
Packets: Sent = 97, Received = 96, Lost = 1 (1% loss),
Approximate round trip times in milli-seconds:
Minimum = 2ms, Maximum = 306ms, Average = 124ms
Why these statistics lack the standard deviation always baffle me, since that would give an idea about the spread of the data. Anyways, my results seem different the the results of @Mushoz. The 1 stream test is in line but the 32 stream test is not. I also find increased latency as opposed to the 1 stream test but not in the 2000-3000 range. Interesting results which need further investigation before a conclusion can be drawn. When I have the time I will investigate further.
What kind of improvements are you seeing? Are the improvements on the 2.4g band, the 5g band or both? This might be an unrelated question, but what is the maximum tx power that you can use on the 5g band? And I don't mean the maximum tx power that you can set in luci, but the maximum txpower that is actually usable. This can be seen in Luci:
