New router test for bufferbloat Belkin RT3200 Linksys E8450 AX

I might be missing the obvious, but ~80/20 MBit/s shouldn't put either of your routers to its limits, far from it.


yeah, agreed. These two devices should both have lots of cycles to spare at these speeds. However power saving could be an issue. Also @Dopam-IT_1987 did you use the same SQM speed settings?

I notice quite a bit of difference with the rpi4 compared to the ipq806x... even on 50/15... so guessing a fair bit is tied up in interrupt latencies et. al./scaling-as-above? rather than raw bandwidth... but thats just a random guestimate/2c...


hello, yes i have use the same settings at one details to armv8 i'm use just wan for sqm
but same value speed @dlakelan

i think i will test your script now for see ...

hello everybody i have test fiber optic with rt3200 but with sqm is so poor with sqm enabled , why

this is my result ...

box only
rt3200 no SQM
rt3200 with SQM

On dslreport you are using dsl link
You have to use fiber.

The dslreport's settings like dsl or fiber basically just control how many parallel streams are used. Typically individual TCP tends to 'saturate" before line rate is achieved, so using more parallel streams allow to reach higher throughput; it should not affect the bufferbloat measurements though.
@Dopam-IT_1987 what do you consider poor in these tests, the thoughput (then follow @EXREYFOX 's advice) or the bufferbloat?

@Dopam-IT_1987 what speed do you have on your fiber ISP. 1gig download 1gig Upload

1 Like

yes my mate has 1gi &gi but is my fault for the test :slight_smile:
1first i have not all connected on the router and not placed to dmz

2nd like you say : i have not see than use the link dsl :confused:

i remake a test for see soon :slight_smile:

good evening everybody

this is my new test with linksys E8450 (rt3200) and is a good news without SQM

and now with SQM

@moeller0 @EXREYFOX

1 Like

That's a huge amount of bandwidth loss compared to without SQM

yes @dlakelan but it's still suitable, now you have to know if you choose stability or speed :slight_smile:


with fq codel + simple qos

better result than sqm

Do you mean cake

yes fq codel + simpleqos work better than cake and piece of cake

config queue 'eth1'
        option enabled '1'
        option interface 'wan'
        option debug_logging '0'
        option verbosity '5'
        option qdisc 'fq_codel'
        option script 'simple.qos'
        option download '62000'
        option upload '18000'
        option linklayer 'ethernet'
        option overhead '44'


and this is my iptables for game video

iptables -t mangle -N dscp_mark
iptables -t mangle -F dscp_mark

iptables -t mangle -A -j dscp_mark

iptables -t mangle -A POSTROUTING -p udp --dst -j DSCP --sport 30000:45000 --dport 3074 --set-dscp-class CS4 -m comment --comment "Dopam-IT_1987-UDP-1"

iptables -t mangle -A PREROUTING -p udp --src -j DSCP --sport 3074 --dport 30000:45000 --set-dscp-class CS4 -m comment --comment "Dopam-IT_1987-UDP-2"

iptables -t mangle -A POSTROUTING -p tcp --dst -j DSCP --dport 50000:65535 --sport 3074 --set-dscp-class CS4 -m comment --comment "Dopam-IT_1987-TCP-3"

iptables -t mangle -A PREROUTING -p tcp --src -j DSCP --dport 3074 --sport 50000:65535 --set-dscp-class CS4 -m comment --comment "Dopam-IT_1987-TCP-4"

2.160 is my console for games

Note, I do not think simple.qos does anything special to CS4...

1 Like

ok, you suggere use what for better traffic in ps5 please ? thanks

Not sure, but IIRC EF is put into the highest of simple.qos three priority queues.

Ok i'm playing in wired is important i think

So if i résumé cake piece of cake cs4 and

Fq codel with simple.qos ef

And simplest.qos équivalent to piece of cake

Testing to cs4 or ef my gameplay at the moment is very slow i don't understand why :confused:

Has anyone actually line tested a 1Gbit fiber connection with this router running OpenWRT without QOS?
I'm interested in whether the device can get line speed. I have 1 Gbit symmetric fiber and am considering this device.
Also, since there seems to be some preliminary support for wifi 6 (802.11ax) mode which would be very interesting to me as well considering that I have about 3-4 AX devices in the mix (phones and Pcs).
Also curious to know the WiFi throughput on 5GHz in VHT80 mode (or AX mode). With the R7800 I currently own, I'm literally getting pissed off by the random latency hikes to the WiFi (this is a known driver issue in ath10k and ath10k-ct). I'm running a build which enables NSS and can theoretically get to about 600mbps on 5GHz VHT80 mode on my AX clients. However, given that the R7800 is only now getting some love with NSS cores, stability is still being ironed out.
Given that this [Linksys & Belkin] is a Mediatek hardware underneath, I feel OpenWRT should be running much better on these compared to the R7800 because it doesn't have proprietary blobs in the middle.
Any feedback on the wired and wifi performance would be appreciated, even if it is iperf/, etc...

1 Like