GL-MT3000 - poor local performances

I have the Beryl AX and:

  • One iPhone 14 pro (Wifi 6)
  • One MacBook Air M2 (Wifi 6)
  • One laptop with an Intel AX210NGW (Wifi 6E)

All the devices are connected to the 5Ghz Wifi, Bandwidth 160Mhz, channel Auto, WPA3-SAE.
When I try to run iPerf between

  • The Mac and the iPhone: 401Mbps/up and 394Mbps/down
  • Between the laptop with the intel card and the iPhone: 393Mbps up and 445Mbps/down

I cannot reach the advertised speeds of the router (2402 Mbps).
But what annoys me the most isn’t just the speed but the latency.
If I ping between the Mac and the router ( I get ~8ms latency, which is fine.
If I repeat the operation between the Intel card and the router, the same thing happens, ~8ms latency.

But as soon as I ping the Intel card from the Mac I get anything between 80 and 250ms (which is pretty wild for a local network!).
Sometimes it goes as low as 5ms but it’s very rare, I don’t understand why there is so much variance in latency when I have a consistent ~12ms latency pinging
I’ve also tried it on my GL-MT1300 and I have the same latency issue (the speeds are even lower tho).

Any ideas?

The iperf test can try using -P to use multithreaded tests.

I have tried both of my phones and the delay is within 10ms. Is this problem only on certain devices?

I have bad news for you. The advertised speeds of routers are theoretical maximums based on “ideal circumstances” and are not representative of real world experiences. Here’s a link that explains where those numbers come from and why you won’t see them in the real world:

I suspect the wild ping latency when pinging the intel card has something to do with windows networking quirks. I get the same kinds of wild responses when pinging my xbox consoles on a completely wired network.

As for the wifi speeds, it’s generally not recommended to test wireless performance between two wireless devices connected to the same AP as you’ll generally halve your single client speeds at best. The Beryl-AX is a Wifi 6 2x2 access point which means it can handle two streams of data at a time. This means that it needs to juggle which device is receiving data. On a good day, in a bunker underground with no RFI floating around between a wired server and wifi client you might be able to get 1400Mbps on iperf3 with multi-stream enabled. In a normal apartment building on a normal day I’m excited to get above 500 Mbps on my enterprise AP let alone this relatively dinky $100 router/AP.

I’ve done some testing of mine and I was able to get around 700Mbps from wifi client to wired server under as ideal conditions as I could get and that’s about as good as one could hope to get realistically.

1 Like

Yeah that was my thought too. Expecting 2.4Gbps out of a $100 travel router…ain’t gonna happen. WiFi 6E is finicky and only reaches “theoreticals” in white room controlled labs…and only gets about 1/2 that at very short distances in very ideal conditions outside the lab. In a typical home/apartment you are right and I think 400-500Mbps is pretty darn good for these little routers. His expectations are WAY too high.

The thing I’m impressed by is that that test is wireless to wireless. That little router is pushing nearly 800Mbps over wifi after overhead. I’m impressed.

1 Like

I have tried both of my phones and the delay is within 10ms. Is this problem only on certain devices?

All of my devices are affected.

I can understand the speeds but I’m more concerned about the latency. It makes no sense to be able to ping and get a consistent 11/15ms from all my devices while pinging local devices gets me anywhere from 100ms to 400ms and very occasionally drop packets.

My Intel AX210NGW is running Ubuntu, pinging the MacBook of my girlfriend also results in the same latency issue so I don’t think the issue is coming from the Wifi cards.
I’ll try in a different environment while being closer to the router.

Can you try a test of latency from the Mac on Wifi to the Ubuntu laptop on Ethernet? Wifi send and receive “take turns” transmitting and receiving on the same band, resulting in higher latency and lower speed, especially if there is wireless congestion with other access points on the channel.

I do not work for and I am not directly associated with GL.iNet

Yes, with the Ubuntu laptop on Ethernet the latency issue is gone, but it isn’t really what I want to achieve.
Doing this also works in regards to latency:

  1. Ubuntu laptop on LAN on my MT-1300 (second router)with the MT-1300 set to internet relay with MT-3000
  2. Mac connected to MT 3000 via Wifi

That works but… it’s not ideal. I didn’t know the latency would start to break down locally that much with just 2 devices connected to the same Wifi, especially since Wifi 6 as I believe that “turned based congestion” has been addressed on that version of the protocol.

Most Wifi cards a setup to connect to the strongest signal/stable connection and not for speed by default
How to Increase WiFi Speed: Learn How-to Boost WiFi Speed, Signal and....
How To Optimize the Wireless Network Performance When Using...
Recommended Settings for 802.11ax Connectivity.

I tried on both the 2.4Ghz and 5Ghz bands, I get the same results.

The Ubuntu laptop on Ethernet test was only intended as a test to eliminate alternate transmitting and receiving on the same band that inherently causes higher latency due to switching delay.

I am unclear regarding your “internet relay” setup, but if the GL-MT3000 and GL-MT1300 are connected over Ethernet, then the Mac and Ubuntu laptop would be connected to separate wifi signals that eliminates the switching delay.

Another option on the GL-MT3000 is to connect to the 5GHz band for the 1st device and the 2.4GHz band for the 2nd device, although the 2nd device would get slower speed on 2.4GHz.

I also have an Intel Intel AX210NGW on the laptop I am using right now and I agree that it is not a wifi adapter issue.

Is packet steering across all processors enabled in LuCi?