Looks like it was already an issue with Mudi V1 too. And of course it was ignored.
I decided to measure in every way my Mudi V2 after the poor cellular experiences and I just realised that this router does not even capable to provide at least the half of any advertised wireless speeds.
I connected it to my main router via LAN with a 1000/200 Mbps connection. Of course Mudi V2's switch is limited to 100 Mbps. That's clear. The goal was to reach this speed at least through Wi-Fi.
It is just one problem that the 2.4 GHz band is very poor, but the peak 5 GHz speed is as useless as the 2.4 GHz. This is the maximum what I can get no matter what configuration I use. Tested in every possible way, every possible combination.
No overlapping routers in my area. All of my client devices/hardware are AX capable. Intel AX201, Intel AX210, OnePlus 11 with a plenty of antennas.... Cannot even reach the 100 Mbps. Again. This is the peak. It's awful.
May we confirm whether you have changed the bandwidth of the 5 GHz SSID to 80 MHz, and whether your other devices are able to negotiate speeds of 433 Mbps?
In our tests using iperf3 running on the Mudi V2, we observed approximately 110 Mbps download and 150 Mbps upload under Wi-Fi interference.
Please also note that the Mudi V2 is a portable 4G mobile router. In most use cases, LTE speeds of around 75+ Mbps should be sufficient to meet general requirements.
Tested with iperf and Waveform Bufferbloat test too just to be sure that this is not some kind of weird case with OOKLA's Speedtest. But all ended with same results. Then tried several settings. I also tested all encryption and all possible bandwidth settings. Just in case I tested again with the same settings as you asked:
About the cellular issue:
Did a tons of measurements few days ago. In my case ~50 Mbps is the top 4G speed with Mudi V2 while mostly all CA configurations available by the provider. In my area B3/B7/B20/B28 bands available and the coverage is perfect. Using the same card in my phone means ~400 Mbps on 4G. I have no such idea what are the exact differences between your measurements and mine could be, but I can promise that I would not complain here if everything would be fine with the device. I have a Puli and I bought the Mudi V2 because I expected better performance in key cases. I know that theoretical capabilities are far from the real world experiences. But the differences are oddly huge and unexpected. You might notice that I constantly report and investigate issues here, because I care. So I would like to figure out what cause these huge differences.
We have made some adjustments based on your test environment.
In the iperf3 test environment: iperf3 server <-- Ethernet connection --> Mudi v2 <-- Wi-Fi connection --> Client
We observed throughput of approximately 90 Mbps for both download and upload.
In the Speedtest environment: Internet <-- Ethernet connection --> Mudi v2 <-- Wi-Fi connection --> Client
We observed throughput of approximately 80 Mbps for both download and upload.
During testing, we also noted that the Mudi v2’s CPU was operating at near full capacity.
This suggests that the observed speeds are close to the upper performance limit of the device when handling traffic through both the USB network adapter and Wi-Fi simultaneously.
Considering that the Mudi v2's primary internet access method is cellular (perhaps supplemented by USB to Ethernet and Wi-Fi repeater functionality), we believe this throughput should suffice for the vast majority of requirements.
You might consider a device like the XE3000 to achieve higher Wi-Fi speeds.
Furthermore, the modules on the Mudi v2 are merely CAT 4/6, with theoretical cellular speeds capped at 150Mbps/300Mbps DL.
Comparing this to the CAT 20 found on flagship/near-flagship smartphones, which boasts theoretical cellular speeds of up to 2Gbps DL, is meaningless.
Of course, we fully understand and appreciate your concerns.
For future product releases, we will take such requirements into careful consideration, with the goal of delivering products that better meet your needs and expectations.
I did many tests. But I definitely forgot that I can test directly on the local network. However the results are only relevant if we only use the modem for local file management which is not the case most of the time.
And sadly I totally forgot to check the SoC before I bought the Mudi V2. I was a bit more confident and trusted way too much in my recent GL.iNet experiences. Now I see that this device is badly designed. Less effort for more profit. I mean, come on: you built a complex hardware on a DECADE OLD, ENTRY LEVEL, one core SoC which does not even have built in 5 GHz radio instead of using a somewhat relevant SoC. Of course I did not expect gigs of bandwidth but I would say a solid ~100 Mbps throughput in every direction in the 2020s would't be so futuristic with a modem like this. But what I got is a totally under powered hardware. This is the issue. It struggles in every ways. Hits the ceiling most of the time.
This hardware is bottle necked so hard and far from anything what the specs sheet promise. Basically it is an expensive but limited modem because you don't even tried to add a BIT better SoC. It cannot even compete with the Puli but it is marketed as a "750" Mbps router while the only relevant difference is the extra 5 GHz radio.
I don't want to argue, but it is pretty confusing how far is this device from the paper. It is usable, but I am sure that most GL.iNet users are not Netflix watching average Joes and interested in networking and expecting more reliable and future proof devices instead of a PCI-E modem which is packed well in plastic.
I am still unable to replicate the speeds what you did on WLAN and WWAN (except the direct iperf LAN > WAN speed) side and I don't know why. I cannot achieve more than ~80 Mbps on WLAN and ~50 Mbps on WWAN. The average is usually less. I guess the software NAT acceleration would not give any boost.
XE3000 is not an option for me. I'm travelling mostly on foot with a backpack. But I would really appreciate if the Mudi V2 were more polished and less buggy on software side.
