Tbh I'm just basing that off a post from misuzu who directly referenced the 2.4 Ghz problem being fixed with pesa build. Link here: https://forum.gl-inet.com/t/flint-2-gl-mt6000-bug-reports-collective-thread/35608/2364
I've tried it now. It doesn't help.
Did you click the âallow VHT on 2gâ box under the 2.4g WiFi settings on his build?
Yeah, it reduced the performance. As I undertand this, it's adding QAM-256 modulation.
To be clear - the primary issue is that youâre limited to 60mbps? You state youâre having packet loss as well? My mt6000 gets between 60-80mbps on 2g, but I only run a couple outdoor cams and IOT devices on it so max throughput doesnât matter to me. Iâm not having any packet loss though, so canât comment there.
I can reach 132Mbps with my samsung galaxy s24+. There is no packet loss. There is reduced bandwidth compared to the original firmware. And if you look at the upload speed at the start of the speed test - you can see that the router "hiccup" at the start of the chart.
Stock firmware has some proprietary components, so that may be playing a role. Is 132mbps not enough for your needs? Is 5g an option? Are you attempting to solve an issue - or just comparing and curious?
Just looking for information regarding this. I thought people had an idea what this issue is about.
I also dont know the exact details but from the earlier firmware builds including vanilla snapshots and GL's own firmware, the 2.4ghz speed maxes out to 5Mbps if Im using a smartphone with no ax support. Ax210 however works fine. GL "fixed" this by using proprietary drivers as their base for their "stable" releases.
But when I tries pesa's build, the speed was the same with GL's stable firmware on 2.4ghz. Also maxes out at 60mbps. My phone can only handle at around ~70 Mbps PHY speed so I thought that it was already working fine. About the latency and hiccups, Ill try to test it on my end later.
EDIT:
There are indeed some fluctuations at the start of speedtest but I think that it is because of the interference (although I tried to make it as low as possible) and the nature of 2.4ghz itself.
is this thing able to route a 2.5G connection without hardware offload? i saw on another forum some worrying data about connection without offloading (well below 1G), is it true? quite scarying if you need to enable SQM for any reason, isn't it? and what is the top speed on wifi can be reached, with and without WED?
thanks
These are good findings. Can you post this info on the issue in GitHub so the devs can look into this? You might be on something here.
I dont really have any technical info about whats causing this and all I did was took tests between versions and the "open source" drivers performs poor on 2.4ghz while the proprietary drivers that GL got from Mediatek themselves fixed this problem. Thats all I know.
It's alright. What's the difference in performance in your case?
I havent tried the latest Openwrt snapshot for MT6000 since Im still using Pesa's build (3.1.0) but all GL's firmware that utilizes open source drivers such as their latest Op24 firmware (most probably a fork of older Openwrt snapshot) still has the 2.4ghz problem on my smartphone device (ac only). The speed only maxes out to 5-10mbps. 5mbps on most cases. However, when GL started using proprietary drivers, the speed on 2.4ghz seems to "normalize" as it can get to 50-60mbps. Again, the PHY speed on my smartphone is only 72mbps so it might be faster. Using my Ax210 this problem doesn't exist at all so maybe not being an ax device is another root of the problem?
Moving on to Pesa's build, the 2.4ghz performance is almost identical to the latest stable release of GL (which uses proprietary drivers) and even sometimes shows lowrr latency over GL's firmware. Pesa might have the tehcnical know-hows on what he did to fix this issue on his build.
Sadly, thats all I know.
For me, the Pesa's build and the latest official snapshot have the same performance in 2.4GHz.
However, they are better in bandwidth than the stable release 23.05.4 for some reason.
Of course, the original GL firmware still retains the best performance in 2.4GHz.
I am not using the wireless radios, I have a better access point, is there a way to fully turn off the wifi radios, I see in LuCI Statistics each radio at 50C in temperature, even tough they are disabled in OpenWrt, they are using a lot of energy idle to have a temperature higher than the processor, thanks
You need a custom build without wifi drivers/packages. Open a new thread.
A new thread never hurts, but maybe there is an easy way out.
Go to /lib/modules/<version>/
and delete (or better yet, move to /root in case of withdrawal symptoms) mt76.ko
. Reboot.
No driver... no device.
That won't work after next update
Why?
Sure thing, needs to be done after every flash or openwrt reset. Maybe that's what you mean.
Could be even put in rc.local: rm -fr /lib/modules/*/mt76.ko
... and sysupgrades keeping settings will keep this in place.
I still think it's a much easier way than building a custom image without wifi. And yes... this probably should be a separate thread.
Now back on this thread, could well be that I missed something and the mt76 module is going the way of the dodo (in that case whatever replaces it can take the same treatment above). If this is the case please let me know.