GL.iNET Flint 2 (GL-MT6000) discussions

nahsha · 2024-12-28T17:49:27.179Z

The CPU seems fine, and sure not all adapters are built equally, that's why I got 3 usb to eth adapters to test with and unfortunately, all of them are performing the same (whether they're used on mac os or windows).

I'll check with gl.inet support, if they won't be able to resolve the issue, back to amazon it goes.. haven't anticipated these issues at all.

Bartvz · 2024-12-28T22:12:22.046Z

nahsha:

not all adapters are built equally

I meant not all router boards are built equally

slh · 2024-12-28T22:23:55.791Z

The Mediatek platform has very well optimized ethernet/ switch and packet processing engines, but that does not translate to a random USB device you connect to it. USB3 is not the main focus for this platform, as mentioned it does cause interference for the wireless cards, the device drivers of the various USB3 networking cards differ a lot in quality/ CPU requirements (e.g. ASIX is known to be particularly problematic) - and at the end of the day, the router may very well not be fast enough to service a USB card at those speeds.

But, again, this forum is for devices running OpenWrt, not the vendor firmware - results may vary widely between them.

pattagghiu · 2024-12-29T08:18:02.593Z

it looks like a wonderful ""2-5G USB Adapter"" found on Ali

what about if you force negotiation to 1G?
what about other lan ports @1G?
don't you really have any "serious" wired device at home to test?

Jack007 · 2024-12-29T11:17:49.379Z

@nahsha ; it seems you are running iperf3 on the router.
What happens in a more realistic scenario? Iperf from your macbook to an other client on the network.

qunvureze · 2024-12-29T13:07:40.141Z

You could try the following script:

MT6000 custom build with LuCi and some optimization - kernel 6.6.x Community Builds, Projects & Packages

When observing napi/phy0-10 and napi/phy0-12, an invalid configuration was found. Unlike ethernet napi, it seems that the CPU is fixed for mt76-tx. The following is the configuration after the correction. The configuration of phy0-10 and phy0-12 is a temporary setting. echo 2 > /proc/irq/120/smp_affinity echo 8 > /proc/irq/121/smp_affinity echo 4 > /proc/irq/133/smp_affinity echo 1 > /proc/irq/128/smp_affinity taskset -p -c 0,2,3 `pgrep mtk_eth-5` taskset -p -c 0,1,2 `pgrep mtk_eth-6` taskse…

It would be interesting to see what does the vendor support have to say about LAN1 speed issues.

nahsha · 2024-12-29T15:03:22.462Z

More like 3 USB adapters from Amazon (highly rated)

Edit: Forcing LAN1 to 1G or using another LAN port results in the same speeds, averaging 500Mbit.

Sadly, none of my devices have 2.5G, I was just curious how USB-A/C to 2.5G Ethernet would perform.

I know you guys mentioned that vendor firmware isn't supported, but would anyone know why, on the latest snapshot (OpenWrt build), I am not getting full WAN speeds, whereas on the vendor firmware I do?

Thanks!

nahsha · 2024-12-29T15:25:34.744Z

I've contacted the vendor by email, with all of the info/tests that I've performed, lets see what they have to say.

quarky · 2024-12-31T02:02:26.078Z

@phinn , may I check if you are still seeing lots of "integer out of range" logs for your router?

If yes, may I know if you have configured he_bss_color for your AX wireless interfaces?

If you're still seeing the logs, you may want to try configuring he_bss_color (regardless whether you have it configured originally in your /etc/config/wireless) to a random value between 1-63 and restarting your AX wireless interfaces and see if that stops the logs. You may need to try a few different values if the first few didn't stop the logs.

If you did not configure he_bss_color you can also try restarting the AX wireless interfaces. This has the same effect of assigning a random value between 1-63 to he_bss_color.

phinn · 2024-12-31T16:30:55.786Z

Yea I posted on the github bug page about it (my github account name is different) but what I also saw was disabling 2.4GHz I am not seeing the errors. I only have one old device on 2.4 Chromecast Audio plugged into studio monitors, my phones/laptop defaults to 5GHz band and don't seem to trigger that integer out of range error.

Thanks for the suggestion I'll try setting BSS color, (haven't set that before) and report back on the bug page. Between this weeks hostapd and mt76 wifi updates gotta do a new build soon currently running rc4.

mindwolf · 2024-12-31T21:25:52.344Z

Use can use SFO with this unit and most other, just not HFO.
WAN Speed: 600 x 30
Band: 5ghz
Distance: Approx 30ft through 1 wall
Driver Version: 23.100.0.4
Waveform Results for Intel AX200 <-> MT6000

NickC · 2025-01-01T10:46:54.821Z

In my testing SFO messes with the ingress part of SQM and I had to disable it to stabilize my latency.

I guess it depends on use case

mindwolf · 2025-01-01T22:14:49.232Z

On a second look, My area is congested with neigbors blasting their dbMs so it's difficult to get stable rates for consistent testing.

gbx · 2025-01-02T16:11:38.010Z

On RC4....

I notice the tx power, even if set to lower like 13 dBm stays at 20 dBm.

Is this normal? iwinfo reflects the 20 dBm tx power.

ktmakwana · 2025-01-02T21:38:29.082Z

I changed the WiFi firmware to a newer version of mediatek from 2024 and now the router cannot create a link at MCS11 1024 QAM.

The link receive speed is 1200 MB but link transmit from router doesn't even come close.

Did anyone else have a similar issue?

Bartvz · 2025-01-02T21:39:48.389Z

Which version of OpenWRT are you running since the version of the mt76 driver matters a lot.

ktmakwana · 2025-01-02T21:42:12.338Z

I'm running 23.05.5

Bartvz · 2025-01-02T21:46:16.302Z

I would try a snapshot. They are not as unstable as people think they are. If you keep a backup of your current configuration you can easily switch back.

ktmakwana · 2025-01-02T21:52:20.119Z

I'll try that later tonight. Do you know what version of firmware is default in 24.10-rc4? Is it a build from 2024 May?

Edit: can I keep settings from 23.05? Or should I start over?

Bartvz · 2025-01-02T22:29:26.325Z

Version 20240823160840

I would start from scratch since sometimes phy's can have different names.