I'm not sure what kind of speeds should i expect here, but 200MBit over 5GHz seems too low. I have Openwrt on two Archer C7, one is v2 the other one is v5. Those two routers are linked over wifi: v5 is client, v2 is AP.
I know there are some issues over NAT which i am not using here (hopefully), but is it realistic to expect close to 500Mbit+ over wifi link between those two routers?
Routers are in adjacent rooms, where distance is max 10 meters with a wall in between.
iperf3 says the following:
iperf from PC2 to router Archer C7 v5 (i installed iperf3 package on the router) - this is over wire, which seems strange:
Running iperf on a relatively low powered router (at least by today's standards) will distort the results quite a bit. My Archer C7v2 units couldn't push much more than ~400-500 Mbps through Ethernet (no NAT, no SQM, ...) before becoming CPU bound.
That's actually "a long way" for 5 GHz, depending on the construction of that wall.
@Hegabo: 5ghz would be used only for the link between the routers. big consumers would be using wired network and other wireless devices would be using 2.4ghz wifi on each router to access internet.
@jeff: i tried iperf as a server on PC1 on the attached picture, so link PC1-PC2 has around 200mbit/s bandwidth and this is what i am interested in. for the distance, i tried going 1m from the router and doing iperf with a laptop over 5ghz (thinkpad t440s) it increases speed to 300mbit/s. this could be limitation of the laptop though. when you say that your v2 was cpu bound, do you mean iperf was hitting your cpu or that v2 is really not able to reach advertised speeds? per their webpage:
Simultaneous 2.4GHz 450 Mbps and 5GHz 1300 Mbps connections for 1.75Gbps of total available bandwidth
the 5ghz link would be used only router2router link, so i guess this should suffice.
yes, you are still translating between subnets. when you use transparent bridge (WDS AP+STA) you can disable dnsmasq and firewall services completely which also drops some load off the devices cpu
That, like the similar statements of all the manufacturers out there, is nothing more than marketing hype that they now all have to follow to remain competitive with each other.
Adding up the theoretical maximum modulation bandwidth ignores important things like; channel throughput is significantly less than the modulation rate, maximum rate is only achievable under very high SNR conditions, wider channels are impossible to use in most situations, and may clients aren't able to use provided MIMO either at all, or with all provided streams.
Single-core MIPs-based SoCs come from an era when ISDN and 3 Mbps was common, 30 Mbps was fast and 300 Mbps was screamingly fast. Those SoCs can't handle moving much more than several hundred Mbps through the Ethernet interface, Archer C7 or otherwise.
@jeff when you say ethernet, you mean physical, wired ports?
so, long story short i should pick up some new equipment with 2+ cores. any recommendations?
goal is to have set up as in original post and the whole line should be 500mbit/s - 1gbit/s. hopefully cheaper than 300eur
have you ever tried to measure performance while running tp-link firmware on archers? that should get you close to advertised speeds. open drivers in combination with latest kernels not.
No, because AFAIK original firmware did not support wan over WiFi.
Edit: also, I wasn't burdened with abundance of internet speed, nor did i need to use wifi ... which now changed. And it is a pain to replace whole setup again. I did try disabling firewall, but I gained nothing in terms of performance.
Routing with NAT and the kinds of SQM that people seem to (reasonably) demand at those rates would be hgh-end ARM or x86_64.
See, for example this thread
APs, your choice (dedicated AP, all-in-one configured as AP) with ARM preferred over MIPS at rates above 400-500 200 Mbps aggregate throughput on wireless.
The Archer C7 units are solid. I've used them for many years here as APs. In close proximity, with a MacBook Pro, I get 150-250 Mbps throughput on 5 GHz to a wired host. A quick check shows the CPU running well over 50% during these tests.
unless you need routing and different subnets (which doesn't seem to be the case, at least not according to network setup diagram) put back tp-link firmware and all devices into the same subnet. that will get you bandwidth increase between PC1 and PC2. how much i can't tell since i have only one archer and my laptop's wifi card is 2x2 only
Effective wireless throughputs around (and slightly beyond) 1 GBit/s require 802.11ax gear (which isn't supported by OpenWrt yet), with range and interference limiting the throughput at reasonable (indoor) distances below that anyways.
@jeff - not NAT needed. i actually need a bridge since i would forbid usage of 5ghz network for anything but for the link between those routers. in the picture, PC1 is doing nat towards the internet. this pc1 i may replace with rpi4, but it depends on the performance. for the tests, i saw as well that iperf between pc1 and pc2 spikes cpu on the router, around 50%-60%.
@psyborg no i dont need different subnets, but is there a way to use wifi in the client/bridge (but not repeater mode) with stock firmware? i tried also wds, which you suggested, and i get between 200 and 300 Mbps between the two pc-s, so no major change.
@slh ok, so this is out of reach. i would be ok with 500Mbps -1 Gbps ... this should be achiavable with 802.11ac