[Solved] MT3000 slow wifi

Routers
1

Hi,

Just unpacked new Beryl AX MT3000 and flashed 4.6.8.

The wireless speeds I see [measured with iperf3] are disappointing:

  1. 2.4Ghz ~70Mbps both down and up;
  2. 5Ghz ~445Mbps down / ~400Mbps up.

All settings (incl. wireless) at their defaults (please see below).

Reducing the bandwidth to 20 on 2.4Ghz does not impact the result.

Increasing the bandwidth to 160 does not impact the result. Reducing the bandwidth to 40 on 5Ghz more than halves the speed (down to ~210Mbps down / ~200Mpbs up).

What is the reason I see that low speed rates? I fully understand that the notional advertised 574Mbps (2.4Ghz) and 2402Mbps are theoretic, but the actual results are only 12% (!) of the advertised 2.4Ghz max, and some 18% (!) of the advertised 5Ghz max, which is underwhelming.

Ruling out possible external reasons:

  1. There is no congestion on 5Ghz in my place, with only one 'competing' SSID broadcasting at channel 36 (-89 dBm). 2.4Ghz is somewhat more crowded, but not terribly so. The strongest neighbors' 2.4Ghz signal I see is at -70 dBm.

  2. I know that the testing setup (wifi phone as client - MT3000 - wired PC as server) can do up to 1000Mpbs, so ruling out this is culprit, too.

Why am I seeing such slow speeds? What can I do about these? What can reasonably be expected of BerylAX in real life?

Thanks.

Just in case, uci export wireless:

config wifi-device 'mt798111'
        option type 'mtk'
        option band '2g'
        option htmode 'HE40'
        option channel 'auto'
        option txpower '100'
        option country 'DE'
        option random_bssid '1'
        option disabled '0'
        option legacy_rates '0'

config wifi-iface 'wifi2g'
        option device 'mt798111'
        option mode 'ap'
        option network 'lan'
        option ifname 'ra0'
        option macaddr 'REDACTED'
        option ssid 'REDACTED'
        option encryption 'psk2'
        option key 'REDACTED'
        option wds '1'
        option isolate '0'
        option ieee80211k '1'
        option bss_transition '1'

config wifi-device 'mt798112'
        option type 'mtk'
        option band '5g'
        option channel 'auto'
        option htmode 'HE80'
        option txpower '100'
        option country 'DE'
        option random_bssid '1'
        option disabled '0'
        option legacy_rates '0'

config wifi-iface 'wifi5g'
        option device 'mt798112'
        option mode 'ap'
        option network 'lan'
        option ifname 'rax0'
        option macaddr 'REDACTED'
        option ssid 'REDACTED'
        option encryption 'psk2'
        option key 'REDACTED'
        option wds '1'
        option isolate '0'
        option ieee80211k '1'
        option bss_transition '1'

config wifi-iface 'guest2g'
        option device 'mt798111'
        option network 'guest'
        option mode 'ap'
        option ifname 'ra1'
        option encryption 'psk2'
        option key 'REDACTED'
        option ssid 'REDACTED'
        option macaddr 'REDACTED'
        option guest '1'
        option disabled '1'
        option wds '1'
        option isolate '1'

config wifi-iface 'guest5g'
        option device 'mt798112'
        option network 'guest'
        option mode 'ap'
        option ifname 'rax1'
        option encryption 'psk2'
        option key 'REDACTED'
        option ssid 'REDACTED'
        option macaddr 'REDACTED'
        option guest '1'
        option disabled '1'
        option wds '1'
        option isolate '1
2

Reducing the width will cause the throughput to be less, that's to be expected.

However you haven't stated what your client is. To me it sounds like your client is probably using 1x1 therefore is capped at a rate of 433Mbps using 80hz and 200*mbits using 40mhz width.

So again, I believe it's your client that is the bottleneck here. Check your RX/TX rates of your client to the WiFi.

1 Like
3

Thanks for the suggestion. Just to be safe I rechecked the limits for my client (iPhone 11):

image
image823×339 19.4 KB

Your suggestion led me to realize that my previous tests had been done with the phone in low power mode, which turns out limits the data transfer rates. Once in full power mode, I was able to get the following results:
2.4 Ghz, 20 bandwidth: down 110 / up 115 (>55% of the client's max)
5 Ghz, 80 bandwidth: down 765 / up 715 (>60% of the client's max)
Are these results something that can be reasonably expected in real life or there is a way to achieve better rates? No longer unhappy about the results, just curious to know if these can be improved.

4

Those results looks pretty much around the expected for your client. To get anyway near the top figures you would need a device that supports the higher channel width or a device with more streams available. Also, you are doing iperf3 tests which will be to a server on the LAN. The ports are rated 1Gb on the LAN side so your tests will always max out at gigabit speeds if hitting a iperf3 server on the LAN (WIFI > LAN)

If you are lucky enough to have over a 1gb WAN ISP connection and you have a device that can sync higher TX/RX rate then you could potentially get better speeds on WiFi however that's only via WiFi > WAN, again if you are to transfer via WiFi to LAN clients then your bottleneck is the LAN ports rated gigabit speeds.

For 2.4ghz it's always best to keep the channel width to 20mhz and choose from either 1,6,11 for the channel as they're the only non overlapping channels. Yes using a 40mhz width will double the speed however it then overlaps and can cause interference, for that reason 2.4ghz is always suggested to be kept at 20mhz.

Sometimes using 80mhz on 5ghz can also be problematic with interface and range. Choosing 40mhz also has the advantage of increasing signal range.

Again, you tested via iperf3 and are chasing LAN speeds. If you have a NAS and transfer files over LAN then setting to 80mhz makes sense. If however your ISP speeds are less than your current RX/TX device sync rates then you could get away with lowering your channel widths which in turn might actually perform better.

Unfortunately WiFi isn't a guarantee, trial and error with settings is the best way to figure out your sweet spot. But looking at your current figures I would suggest you are pretty much in the region of what's to be expected.

2 Likes
5

Many thanks for the detailed explanation and help!

1 Like
6

This topic was automatically closed 180 days after the last reply. New replies are no longer allowed.