[hAP ac3] 2.4GHz radio faster than 5GHz?

[markep]

I have a hAP-ac3 running 6.48.3 sitting about three feet away from my laptop. With a wired connection to the MikroTik, speedtest.net tests yield around 630 Mbps up/down. When connected to an SSID associated with the 2.4 GHz radio, I’m getting around 120 Mbps up/down. Connected to a 5 GHz SSID, I’m getting only about 80 Mbps up/down, which seems strangely slow. I ran the same tests from a second laptop and received similar results: significantly faster wifi on 2.4GHz.

Based on forum searches, I’ve made a few changes to the 5 GHz interface configuration:
-band from ‘5GHz-A/N/AC’ to ‘5GHz-N/AC’
-installation from ‘indoor’ to ‘any’
-adaptive noise immunity from ‘none’ to ‘ap and client mode’

None of the above changes improved the speed. I know there are many factors involved, but what sort of 5GHz speed should I expect given a 630 Mbps wired speed? Any tips or advice on where to start troubleshooting? Thank you.

[santyx32]

If you’re not using it on a production environment you can try 7.1 beta 6 and install the new wifiwave2 package, it provides new optimized drivers for your device.

[Jotne]

I can be that you have not configured your 5GHz in an optimal way.
Example channel width to 20Mhz instead of 80MHz.

[normis]

1. Upgrade
2. System reset
3. Configure the device using Quicket, set mode to HomeAP Dual - for test!

See what speed you get then. Normally this hAP ac3 can get much much more in both bands

[Jotne]

1. Upgrade

Upgrade to what? He do has 6.48.3, should he use a test or beta version?

[normis]

That’s generic advice for all For this case, he can skip it (or upgrade to 6.49beta54)

[rextended]

@markep we must also think about the laptop…
For example: how many antennas for 5GHz has your laptop?
If the reply is “only one”… do not expect good results, etc.

[bpwl]

When connected to an SSID associated with the 2.4 GHz radio, I’m getting around 120 Mbps up/down

Wow, this is a strong performance. For this as data rate you need something like 300Mbps interface rate, what can only be done with the 40MHz bandwidth. (What is only possible in an low contention environment)
Or … in other words … what are you looking at as rate indicator? If it is looking at the interface rate, then the 80Mbps interface rate for 5 GHz is indeed very low. You should get something like 400Mbps for 40 MHz (2 stream) , or even 866Mbps for 80 MHz bandwidth.

5 GHz doesn’t like to pass through walls, furniture or people. So what is the Signal strength, SNR and CCQ you get (as seen in the wireless registration table).
And what is the exact interface rate indicated? Something like “81Mbps-40MHz/1S”? This is (V)HT MCS 04, a reduced encoding, used because of signal distortion, reflection or interference.

[markep]

Hello all,

Many thanks for your replies–I will start working through these suggestions today and let you know how it goes.

[anav]

Im with bpwl, 120 for 2.4ghz… how?

[rextended]

Probably aren’t test but datarates… Like connect to a pc the ethernet cable and get 1G or 100M

[markep]

So, good news and bad news. The good news is that I’m seeing vastly improved speeds on the 5GHz networks. The bad news is that I didn’t change anything yet, so I’m not sure what the resolution was, or if it’s resolved long term.

From my laptop, a speedtest.net test just yielded 380 Mbps down, 330 Mbps up on the 5GHz wifi network. From what I’m seeing here, that should be about right.

One thing I neglected to do was reboot the MikroTik after making the changes referenced in my original post, but I haven’t reset it since then either. After each interface change, the 5GHz wireless networks vanished for about ten minutes per the normal “listening” stage, so I assumed the changes were taking effect. Could there have been some routine nightly job that made the difference?

@santyx32 - I will keep an eye out for v7.1: this box is in production, but I’m glad to hear about the coming new drivers for the hAP AC3.

@Jotne - channel width is set to 20/40/80MHz XXXX (I believe I’ve left this at the default)

@normis - I’m set to WISP AP, but if I find degraded speeds again I will give this test a try in the wee hours.

@rextended - Your comment led me to check on the devices I’m using. My primary laptop is good, but a Surface Go was connected to the 5GHz network using 802.11n, though the adapter is capable of 802.11ac–still looking into that. (I’m only getting speedtest.net results of 97Mbps down, 108Mbps up on it. The 2.4GHz numbers are a little better.) My phone (Pixel 4a) is pulling down 275 Mbps on speedtest.net, so I’m seeing great improvement there too.

@bpwl - For rate I was just using the speedtest.net website. Should I be using some other method? The numbers look much better now on the registration table (the rx rate does seem to bounce around somewhat compared to signal and tx rate, which are fairly stable)
tx/rx strength: -51 | tx rate: 866.6Mbps-80Mhz/2S/SGI | rx rate: 650Mbps-80MHz/2S/SGI
And yes, the 2.4GHz connections are excellent! A (very) nearby laptop shows -37 signal strength, tx of 150Mbps-40MHz/1S/SGI, and rx of 300Mbps-40MHz/2S/SGI. A Raspberry Pi 4 sitting next to the MikroTik has both tx/rx at 300Mbps-40MHz/2S/SGI.

I’m going to keep an eye on it and see if these improved results remain. I’m still curious about what may have happened. Thank you MikroTik community!

[rextended]

After each interface change, the 5GHz wireless networks vanished for about ten minutes per the normal “listening” stage

Is the DFS, if you set one frequency different to 5180 or set to auto, on DFS frequency check if some radar are presents before staret transmit.
this require sometime more than 10 minutes

[bpwl]

Excellent. Check the registration rate as rate indication (that gives the wifi speed, (and 380Mbps out of a 866Mbps is normal max value, for MT with the current drivers)).
Speedtest over the internet has many other factors included in the test.

But my advice: if you leave the freq on auto, and/or use XXXX as channel selection, your results may vary after every change or restart. In fact you leave the channel selection to the MT, which has very few instant information, to make a proper selection. Whenever DFS channels come into play , there is a 1 minute “listen for radar” time, if a weather-radar freq (5600,5620,5640) is part of the selected range, then the listen time is 10 minutes, and comes with EVERY change. (cfr @rextended). (There is an option to skip the 10-min channels, even to skip DFS, but then there is little left to use.)

[markep]

@bpwl When I look at the current 5GHz interface, it is showing “5700/20-eeCe/ac/DP(21dBm)”. Is your recommendation that I manually set the channel width to “20/40/80 MHz eeCe” and the frequency to “5700” in order to lock in the settings that are currently working well? Would there be any downsides? Thanks!

[bpwl]

That would be a good idea, to stabilize the outcome. Actually one should first do several “Freq usage” and “Snooper” scans to find out what else is potentially interfering, and then make a proper selection.

A radar detection may still overrule your own setting. To be somewhat more in control, you have to fill in the “Channels” table.
Something like this … (my values irrelevant for your case). HOW RouterOS is using those tables, I couldn’t find out yet. (Regular start over from the beginning of the list, by radar detect (= false positive mostly))

[anav]

Excellent. Check the registration rate as rate indication (that gives the wifi speed, (and 380Mbps out of a 866Mbps is normal max value, for MT with the current drivers)).
Speedtest over the internet has many other factors included in the test.

But my advice: if you leave the freq on auto, and/or use XXXX as channel selection, your results may vary after every change or restart. In fact you leave the channel selection to the MT, which has very few instant information, to make a proper selection. Whenever DFS channels come into play , there is a 1 minute “listen for radar” time, if a weather-radar freq (5600,5620,5640) is part of the selected range, then the listen time is 10 minutes, and comes with EVERY change. (cfr @rextended). (There is an option to skip the 10-min channels, even to skip DFS, but then there is little left to use.)

bpwl, Using the 1/3 rule in good conditions he should be getting speeds of around 300, so he is doing very well for a close LOS connection. He may even get decent performance through one wall.

Interesting settings: "20/40/80 MHz eeCe AC, whereas I use 20/40 ce n/ac ( without worrying about end users, what is the difference)

[markep]

I ran a few frequency scans and let the snooper listen for a few minutes. After the tests, the 5GHz interface auto-selected settings of 5620/20-eeeC/ac/DP(21dBm). This channel and frequency looks clear to me (see screenshots), which I imagine is why the MT likes these settings, but I once again had poor 5GHz speeds, so I went ahead and rebooted. After reboot the MT again selected 5620/20-eeeC/ac/DP. The registration table for my laptop on 5GHz showed:
Signal strength: -55 | Tx rate: 173.3Mbps-20MHz/2S/SGI | Rx rate: 86.6Mbps-20MHz/1S/SGI
Another laptop showed similar weak numbers:
Signal strength: -34 | Tx rate: 72.2Mbps-20MHz/1S/SGI | Rx rate: 144.4Mbps-20MHz/2S/SGI
As you’d guess from the above, speedtest.net numbers were back in the 60 down/70 up range again (slower than my 2.4GHz numbers).

As discussed in an earlier post, I now went ahead and manually set the 5GHz interface to 20/40/80MHz eeCe/frequency 5700 and waited for radar detection to finish, then connected back to a 5GHz SSID. Registration numbers were strong again. Fluctuating a bit, but still ranging up to Tx: 866.6Mbps-80MHz/2S/SGI and Rx: 585Mbps-80MHz/2S/SGI. Speedtest numbers were back around 260Mbps down and 245Mbps up.
Why then would the 5620 setting yield such poor results? In building the channel table per @bpwl I likely would have chosen that one myself. Here’s a sample of two frequency scans (first on the left, second on the right):

And here’s a sample of the snooper output, sorted by channel:

I don’t see any reason to pick 5700 over 5620, but clearly the end results are wildly different. Unless I’m misinterpreting the freq/snooper data, I get the impression I will have to pursue this empirically, trying all the apparently clear frequencies for performance, then building the channel table accordingly.

[toxicfusion]

Be sure to disable radar detect. Then again try auto-frequency. But honest, it is always best to statically assign your channels/frequencies…

Results on your channel 5700 - look great
Tx: 866.6Mbps-80MHz/2S/SGI and Rx: 585Mbps-80MHz/2S/SGI

Set your frequency to 5220 and try again, Compaire 5220 and 5700.

[markep]

Hello @toxicfusion. I tried the 5220 frequency and it is a strong performer as well. Watching the registration table for a short while yielded these–Signal strength: -57 | Tx rate: 780Mbps-80MHz/2S/SGI | Rx rate: 520Mbps-80MHz/2S/SGI, with speedtest.net results of 260Mbps down, 210Mbps up. Very comparable to the 5700 frequency. I haven’t tried the auto settings with radar detect disabled yet, but will give that a try.

@bpwl Could you point me in the right direction to create the Channels table? I tried to “add new,” but the “List” field is required and empty, and I haven’t yet found where to create the Lists that would populate the field. In the screenshot from your earlier post, how did you create the “dfs” list?

Thanks all.