GL.iNET Flint 2 (GL-MT6000) discussions

Spent the better part of the day troubleshooting, trying different configs, etc... and just cannot seem to get anything better. -92 noise on 5ghz, -90 on 2.4ghz. Every client is 5-10dbm worse signal than the Dynalink as well. It's directly impacting performance and wifi connectivity at all ranges. I poked around on the gl.inet forum and found a screenshot of someone with the same noise numbers on their MT6000, and also stating 5-10dbm worse signal on clients than their previous ISP provided broadcomm. So many other things are great though. HFO, WED... heck even irqbalance actually works compared to ipq807x.

I just purchased gl-mt6000, installed vanilla openwrt (using firmware selector, followed by attended sysupgrade to include all the packages I so far installed). I would like to use LetsEncrypt certificates for uhttpd. I installed uacme and successfully created the certificate. However, uhttpd refuses to use it and reverts to the self-signed certificates after 90 seconds. I tried certificates used in my old openwrt router (which work there) - they do not work either. Do people know how to make this work?

I don't know much about this but it might just be inherent to MediaTek. I had -92 dBm on another MediaTek device with OpenWRT.

Interesting ... don't know if it is coincidence are not by my MT7621 / MT7905DAN / MT7975DN too has a noise level of around -91/-92 dBm

All 7 clients of mine have a noise level of 90-92 dBm as well. Some are very close and some are several rooms away. Makes me wonder if this is related to the inability to set the TX power (maybe incorrect reporting?), as I can't complain about the wifi performance, all clients have great bandwidth and stability.

1 Like

I share with you also my experience.
I also have noise of -90db- -92db.
From my knowledge of wireless signals, -90db or -92db is a very low noise, close to no noise at all, so, for me it is very good.
So, if you connect a bad shielded usb 3.0 pendrive or hard drive you should see the noise increase due to interference between usb3.0 signal and 2.4ghz wifi.
And, if you change the country to another who permit more mw you normally should see a more level of signal not noise

When I check it, for me it is important the SNR (signal noise ratio)
From my experience, when I get:

  • 5 dB to 10 dB: is below the minimum level to establish a connection, due to the noise level being nearly indistinguishable from the desired signal (useful information).
  • 10 dB to 15 dB: is the accepted minimum to establish an unreliable connection.
  • 15 dB to 25 dB: is typically considered the minimally acceptable level to establish poor connectivity.
  • 25 dB to 40 dB: is deemed to be good.
  • 41 dB or higher: is considered to be excellent.

If you look at the picture only the first one on the list is a low signal, but ok for me because it is a NodeMCU outside of my house to get the outside temperature.
immagine

Regarding irqbalance I didn't see any benefit, but I don't have a fiber connection, I have only 100mb Internet connection and I reach 90 with openvpn.

Latest days I tell to the people of this thread that I prepare a build with kernel 6.1, if somebody would like to test here it is...

Thanks and sorry for annoying post

3 Likes

Good information.

Can you explain why kernel 6.1 is important? I heard it's big for mt76 devices but not sure why.

I compile with 6.1 to test it because it will be the next LTS kernel for our openwrt devices... and I think that DSA support is better
But frankly speaking I never tried for a long time the MT6000 with 5.15.

1 Like

What are people's take on the best way to flash a new firmware on this device? We have 8GB eMMC which is fantastic, but I want to make sure all my packages are going to the latest as well and I have a ton I've been toying with. I thought about doing an attendedsysupgrade but also want to go back to testing 6.1 again too. Are people flashing sysupgrade and then manually updating packages? I want to avoid that headache. Thanks.

This is one area I miss using the WRT32X as my main which has dual-partion/dual-boot.

Compiling firmware image with all needed packages with the full toolchain...
And then normal sysupgrade. All packages are newest and in sync with eachother.
No reason to complicate things.

(Attendedsysupgrade, auc, etc. are all complications/shortcuts, as they rely on buildbot produced binaries, which are not always fully coherent, if certain packages have been just updated.)

1 Like

That's the ideal approach thanks, but I always leave out a few packages and tend to add more over time and end up with a mess after sysupgrade. Right now I have so many packages on /overlay especially this being a new target. I might just update all of them and hope for the best.

At least it appears to be consistent noise across multiple users and units… that’s reassuring to a degree. Now curious if it’s a reporting issue, or simply poor manufacturing. In my network environment, just to reiterate, it’s 14dbm worse than the Dynalink… which is significant. Especially when combined with 5-10dbm worse signal across the clients. Put them together and you’re looking at 20-25dbm of network quality regression.

It’s still also concerning why adjusting transmit power (up or down) does not appear to translate to the clients. I’m guessing it’s all tied together.

What is 14dbm worse signal, SNR or noise? please paste some result details to understand better of Mt6000 and other device...

you can use

 iwinfo phy0-ap0 assoclist

 iwinfo phy1-ap0 assoclist

remove MAC, thanks

Noise is 14dbm worse… signal is 5-10dbm worse per client. So overall SNR is 15-25 worse per client.

We use my example...

root@MT6000:~# iwinfo phy0-ap0 assoclist
MAC_1  -65 dBm / -91 dBm (SNR 26) 

(-65)-(-91)=26

I assume that you have MT6000 with this values on device [A]:
-70 of signal -90 of noise, SNR is 20.
What are the result of other device?
If your other device have SNR at 35 maybe you have noise at -76 and signal at -41. So signal is not only 5-10db worse...

(-41)-(-76)=35

Where I'm wrong?

To compare correctly use that command iwinfo phy0-ap0 assoclist and post result on same device, same position, no usb3.0 devices connected

Please re-read what has been thoroughly posted. Noise floor is -106dbm on my Dynalink 5ghz in my environment. With the Dynalink - the position of two specific clients at (iPhone 15 pro max and iPad Pro 11”) had signals of -66 for the iPhone and -74 for the iPad… so their SNR is 40 and 32 respectively. Changing to the MT6000 resulted in a noise floor of -92 on the 5ghz… and the iPhone went to -74 and the iPad to -84… resulting in SNRs now at 18 and 8 respectively. This is true of the remaining clients as well, but no - I’m not going to go through and post every single one. Every client is between 5-10dbm worse. It’s consistent.

At this point I’m beyond confirming the issue, have attempted various settings, but have not been able to close the gap. As has also been discussed - there’s concern that the antennas may not be putting out proper TX power being reported… but that’s not yet confirmed.

I'm just adding data points: reported noise for clients connected on my Reyee RG E5 and Belkin RT3200 dumb AP's (both MT7622/7915 based) is -92 dBm.

2 Likes

Thank you! The more the merrier.

I've noticed that noise level changes a lot for different channels. What's different from other routers that I have, is that GL-MT6000 on lower 2,4 GHz channels (1-5) shows much less noise than on higher. On lower it's the same as on other routers, but higher are significantly worse. My other routers seem to show pretty much the same noise across 2,4 GHz. Note that there are many APs around, including some on channel 1.

Similar story with 5 GHz. The lowest noise is on ch 36, while others vary. Even high end of spectrum is worse, even though the transmit power is very low in EU for these.

Could be some low quality DAC, or whatever is there that converts wifi chip outputs to radio waves.

Could you compare various channels as well? I'm going to redo this experiment a couple of times at different times of the day.

No change either way (between higher / lower channels)... both show noise at -92. This is why I'm really curious if it's a reporting error, because I too would expect to see some sort of fluctuation. I have yet to see someone post a better noise level than -92... almost as if that's the maximum reporting value for Mediatek? All hypothesis clearly, but still curious.

EDIT - To clarify, this is regarding the 5GHz band. I have seen the noise floor fluctuate between -89 and -92 on the 2.4ghz... but still never better than -92.

1 Like