SNAPSHOT (r24498-b2f1c6ed52)
SQM - Disabled
Software flow offloading - Checked
Hardware flow offloading - Unchecked
2 Likes
@whitedd do you have irqbalance installed and set up?
On the same snapshot version as you. SQM enabled: cake / piece of cake
The difference is that I have hardware offloading enabled.
Interestingly enough, there's a huge difference in speeds and CPU load between having the hardware offloading enabled or not:
- software offloading + no sqm = 1Gbps, 30% load of 1 core (only 1 core)
- software offloading + sqm = 600Mbps, 100%+ load on 1 core
- hardware offloading only + no SQM = 920Mbps, up to 1.3% load
- hardware offloading + SQM = 920Mbps, <1.3% load
irqbalance installed but not configured. Could be that it has a proper default config.
Looks like 920Mbps is a hardware limit of my ISP fiber-to-LAN adapter.
Turns out that the hardware offloading makes a huge difference in CPU load. The cpu barely gets any load at all when hardware offloading enabled. While using the software offloading (unless irqbalance configured) creates 100% load on 1 (random) core.
P.S. I just found out that SQM is not working while hardware offloading enabled. 
Have to figure out how to configure irqbalance settings.
An update regarding SQM (cake + piece of cake, in my case):
- SQM does not work with hardware offloading at all.
- SQM works just fine when only software offloading enabled (no hardware) and when packet steering is enabled (in Network - Interfaces - Global network options). - CPU load is significant, and it's up to 60% on all cores, but SQM works perfectly.
Works well without irqbalance installed. irqbalance might help with better cpu optimization, but it's beyond my knowledge how to set it up.
So if somebody knows what how to find the proper config for irqbalance, the help is much appreciated. This router uses Filogic 830 SoC, so that might help with finding the solution.
I was suggested to try something like that, but have no clue whether that will be working:
cat /etc/config/irqbalance
config irqbalance 'irqbalance'
option enabled '1'
# Level at which irqbalance partitions cache domains.
# Default is 2 (L2$).
option deepestcache '2'
# The default value is 10 seconds
option interval '3'
# Specify excluded cpulist
option banned_cpulist '0'
# List of IRQ's to ignore
#list banirq '36'
#list banirq '69'
Which officially makes this router capable of gigabit SQM! 
2 Likes
No...i dont install irqbalance
no sfo and hwo do'esnt work with sqm
you can test 30/30 sqm with sfo and you see the value max
1 Like
I thought with irqbalance it was enable and it just works king of thing but maybe I'm mistaken.
if you install sqm simply disable software offloading as it bypasses sqm
Installed today by using trx and then going to the latest snapshot sysupgrade. Very basic configuration for now (needed to get things working). Also, only 1Gbit WAN/LAN (ISP only gives up to 1 Gbit) wiring for the moment. This is main router after ISP router, which works as Fiber to Ethernet modem + TV.
Quick speed test results:
- ~950 Mbit via ethernet
- ~800 Mbit via 5Ghz wifi (WPA3/80 Mhz, because 160 Mhz settings failed radio startup), this was consistent between 4 different clients all being very close to the router.
Originally I wanted to test ASUS firmware as well, but understood that will have some time for this only next year and really wanted to change the router earlier. So, no comparison, just some results I got.
2 Likes
For info: I updated the wiki page. The tables are now showing the amount of 1 GbE as well as 2.5 GbE ports under the hardware highlights section. I extended the description. And I select "Snapshot" as the current supported release for the TUF-AX6000. Furthermore, I also introduce a new unsupported feature "LAN port LEDs".
Also the hardware table at the bottom of the wiki page, which is using a standard meta template, I extended to display the 2.5 Gbit/s ports. I think more and more devices will come to the market with 2.5 GbE. Eventually I personally believe the 100Mbit column can be remove from this meta template table in the future, since that will no longer be relevant for most people.
1 Like
Hi, thanks for your update but this is not fully true: https://openwrt.org/unsupported/lan_port_leds
We don't have full support only for LAN 2.5Gb LED . ( its on when link is down ). All other LEDs indicate status correctly including 1Gb LED.
1 Like
That unsupported page need to be a very broad and general description of the problem. Since it can be used by other routers as well when there is something not working with the leds.
The page already mentioned that either the leds are partially or fully (in this case the former) not working.
btw. If I remember correctly all LEDs on the ax4200 works correctly. And I think we have to add description what exactly doesn't work giving end user clear statement. For example add to the table detailed information. In the current shape "the led page" mislead end user.
If you think AX4200 works with all LEDs, please feel free to remove all remarks from that wiki page. Incl. the unsupported option from AX4200 (see datatable). It also still says "The LAN port LEDs are driven by the switch, but OpenWrt does not correctly configure the output.", if that is no longer relevant also feel free to remove that.
I don't own a AX4200 device, so I can't either confirm or deny your remark.
I updated the "led page" now to make it more clear and not misleading. Of course my intention was never to mislead users/readers, I try to use the wiki as intended by introducing "unsupported" pages rather than manually adding text on the pages about missing or not working features.
1 Like
Another happy user that got it working using the trx path. After the trx install a sysupgrade created with the snapshot image builder https://openwrt.org/docs/guide-user/additional-software/imagebuilder with some extra packages was uploaded and it is now running alright so far.
5 Likes
Support has been backported to openwrt-23.05 in https://github.com/openwrt/openwrt/commit/557a32aba0aaaae860a0abae181c42df2cf39188. We could expect official images with the next patch release of the current stable.
Kudos to @patrykk for his work on the initial support and backporting!
3 Likes
@patrykk I checked your changes and I have one question. You added it with the firmware dbdc.bin firmware same as AX4200 https://github.com/openwrt/openwrt/blob/6df6f03396a42ef2cd03b8ec757974b6afd07afb/target/linux/mediatek/filogic/base-files/etc/hotplug.d/firmware/11-mt76-caldata#L46 but in snapshot it is under dual.bin https://github.com/openwrt/openwrt/blob/6407ef8d2bcb4a0a6284de09cd77bd1868c1d6ea/target/linux/mediatek/filogic/base-files/etc/hotplug.d/firmware/11-mt76-caldata#L54
IIRC you fixed it afterwards in a separate commit. Could it be that this commit was accidentally skipped during cherry-pick?
Updated: found that commit https://github.com/openwrt/openwrt/commit/0c3f4bd85ee6067588a81843a403850ab0082270
Hi, please look at the commit:
I introduces changes for build where new packages removed files and the file was replaced by copy from flash.
best regards,
1 Like
Thank you for the prompt reply and confirming my findings of the commit. My question remains the same mostly. Should this commit be added to 23.05 branch?
1 Like
And why are you in such a hurry? There are also problems:
It also turned out that there are problems with MAC addresses on a similar device:
3 Likes
I'm not. Was just excited to see this. Sorry if it looked like I'm in a rush.
1 Like