2.5 port on gt-axe11000 is subpar.

[lgkahn]

the 2.5g port on the asus gt-axe11000 appears to be sub par.

i tested with my gt-axe11000 plugged into a gocoax 2.5g adapter and am getting 2.5g connection.. i can only get 1.3 g or so when going to my nas (also on 2.5g through a trendnet managed switch, or to a gt-axe16000 i have on same switch)

so then i bypassed the gocoax and plugged both into the same switch.. again i can only get about 1.3g when going to my nas or to each other via iperf3,

no such problem on the axe16000 i easily get over 2gb everytime..

ideas?

 

[lgkahn]

its not the cpu as i ran a strange test with two ssh connections to the router one running iperf3 in server mode and the other connecting to itself.. i can get 6-7 gb/sec in this test.

 

[TheLostSwede]

Try -w 4M or -P 2 or 4 to see if that makes a difference in iPerf.
Also, you posted in the wrong subforum, as this doesn't appear to be an issue related to Asuswrt-Merlin.

 

[lgkahn]

No difference it was faster with -P 8 which I already had done. 2 and 4 are slower.

 

[jerry6]

I GET A SOLID 2.3 G on my AX1100 pro , not sure about the AXE1100 as it is in the city will test it next time I'm there

 

[lgkahn]

No diff with 2m.or 3m window. 4m will not start errors out...

 

[lgkahn]

as a side note similar issue with gt-axe16000 in ap mode (i have in another location) when using a lan connection via the 10gb port.
it maxes out at 3.3gb.

I dont think it server load as one of my 24 core windows boxes can get 6gb

cpu load does not seem to go above 27% on any core when trying this even with -P 4 so again dont think it is cpu load.. must be s/w or port configuration.

so in my estimation the ports on these routers dont seem to be able to support the speeds advertised.

 

[TheLostSwede]

Humor me, run the test in reverse, -r.

What Ethernet chips are in the devices you're testing with?
A lot of first gen 2.5 Gbps has bugs, especially Intel hardware, so it might not be the router at all.

I ran into this issue testing both Intel and Realtek hardware against the Aquantia card in my NAS. Uploading to the NAS was performing similar to what you're seeing, but downloading form the NAS was performing as expected.

 

[lgkahn]

No.I ran it in.both directions and in addition as I said the.same server can reach 6gb/sec when using my windows box instead on trying to the router and same as other location the 1600.gets.2.6gb as.I.posted.with the same server and switch.that the 1100 only.gets 1.2.so your hypothesis doesn't hold

 

[DJones]

Are you running with QoS enabled? Cake, bandwidth limiter, and flexqos disable hardware acceleration. I find nas’s p2p with iperf works better with a 9k jumbo frame enabled for bigger files like movies. 1500 is fine for smaller files.

Requires jumbo frames enabled on router, and client devices. Note for wan 1500 mtu is usually all you can get outbound. Lan can be set I think up to 16k or whatever your client device / switch supports.

 

[lgkahn]

no qos on the router .. i will will play with jumbo frames but it is a known issue with them and my hubitat hub that jumbo frames can freeze the device and i would have to enable them on the switch as well.

thanks

did it in reverse just for giggles anyway . same issue. but if it was mtu why does the 16000 get close to the full 2.5g but not the 11000 that to me points to the switch port or something in the router itself.. the cpus are simliar 4 core 1.8 vs 2 ghz.

 

[DJones]

no qos on the router .. i will will play with jumbo frames but it is a known issue with them and my hubitat hub that jumbo frames can freeze the device and i would have to enable them on the switch as well.

thanks

did it in reverse just for giggles anyway . same issue. but if it was mtu why does the 16000 get close to the full 2.5g but not the 11000 that to me points to the switch port or something in the router itself.. the cpus are simliar 4 core 1.8 vs 2 ghz.

I’ll also note I don’t believe enabling jumbo frames on router actually changes the mtu of the router it just allows jumbo frame packets to pass through. I think ASUS locks all mtu to whatever your wan can handle. Meaning even if you have some devices 1500 and some devices 9000 the router won’t care or fragment unless it hits the wan or if you’re using iperf directly on the router or you’re talking between lan devices of same or different speeds but different mtu.
My gt-ax11000 only has 1 2.5gb port so unless I’m testing over 160Mhz wifi I think I’d get inaccurate results using jumbo frames and iperf assuming perfect wifi. Derp jumbo frames are eth only so wifi would fragment back to 1500.

With mtu usually things like bridges fall to the lowest mtu your device is handling so 1500 because of wan as it won’t allow mismatching. So I assume the router uses loopback to pass packets in jumboframe as it has a mtu of 65536

Servers, or routers with SDN (software defined networks) probably can get away with different mtu’s on differently subnets because they are virtualized. But this is a way outside my wheelhouse I’m speculating at this point.

 

[TheLostSwede]

No.I ran it in.both directions and in addition as I said the.same server can reach 6gb/sec when using my windows box instead on trying to the router and same as other location the 1600.gets.2.6gb as.I.posted.with the same server and switch.that the 1100 only.gets 1.2.so your hypothesis doesn't hold

Well, sorry for trying to help then. Also, it's not a hypothesis, it's a real world, proven issue with a lot of 2.5 Gbps Ethernet controllers.

 

[lgkahn]

thanks.. anyway jumbo frames seems to be a non starter for me as the router is to service the basement of the house where (
no matter what i did, ie try router in ai mesh mode, try a different router (netgear) that allowed me to be in hybraid ap/extender mode so that i could use one 5ghz channel as backchannel and rebroadcast a different ssid for 5ghz and 2.4ghz on difference channesl, etc)
my cameras failed to stay connected.

so now i bit the bullet and am using the router in ap mode with separate ssids and a coax back channel via moca.

works much better and higlhy recommended

anyway it will be behind moca adapters in the long run and i dont believe jumbo frames work on these.


all in all it is working well and since i only have 1gb internet not a reallly big deal but i do have some devices such as my nas and main router on 2.5gb ports mostly for backups to/from the nas. and it just annoys me that even when connected to the same 2.5gbe switch i cannot achieve over 1.3gb on the 2.5gb port whereas the gt-axe16000 has no such issue.. So it must be some limitation of the port/switch internally in the router.

I guess i can buy another gt-axe16000 if it bothers me too much, or try my ax11000 pro (but that is installed in a different location)