I am wondering why my 5Ghz wireless speed is so much slower than it should be? When I setup the wAP AC device the absolute maximum I can push in a momentary spike is 399Mbps, but average is around ~200-240Mbps regardless of the device I use.
The router is setup as an AP only and all routing is handled by a Debian box which is 10Gbit capable and, using a Cisco AP, can push up to 800Mbps without issue. There is a simple default bridge with all interfaces attached.
Here is a 30-second iPerf from my MacBook to the Debian gateway (you can see how it drops off)
Additional notes:
1.) There is very little interference where I live and there are only 1 client connected.
2.) Test results are taken from 4 meters away, line of sight.
3.) I have changed channels, regulatory domains and have reset the device several times and still see the same results. Never over ~300 consistently.
I have the same issue. I have wap ac and cap ac connected to the router with capsman. Never exceed 300mbps on WiFi although I have 1Gbps connection from ISP.
I did all of this and the average did seem to increase by 20-30mbps, but still not where I am hoping to see this.
I’m starting to wonder if this AP can even handle 400-600mbps, which is what I would have really liked to see. I did notice at 280mbps the CPU is at 42%, so I don’t think the CPU is what is causing the problem.
The clients are running iPerf connecting to the gateway running iperf (debian firewall/router, connected to LAN). I’ve also tried public iPerf servers (I have 1GBit connection) and same result.
When doing the exact same tests, in the same location of the room and with the routers plugged in side by side (1 at a time, to avoid interference) I can double my iPerf results.
Thanks a lot for helping me out here - appreciate it.
CAPSMAN sets WMM, it can not be turned off (says Uldis: Why isn't WMM Support default?)
Worse, WMM is probably only overhead if there are no mangle rules to set the priority.
MCS5 and MCS7 are not the top speed for the "ac" interface rate. Check "registration" for CCQ and retransmits. (it wastes time, lowers the MCS rate).
The % bandwidth you get from an interface depends on aggregation. If there are many retransmits the A-PMDU with smaller A-MSDU should be preferred above A-MSDU at max size.
Tuning is then lowering hw_retry=3, set max VHT MCS, A-MSDU=2048, even going to 2X2 .... all steps that each lower the capacity considerably, but one of them might still give better results.
All of them together would be a very drastic reduction. Don't use them if no gain in throughput.
I’ve decided to use different hardware, but still would like to know what that maximum realistic throughput of these devices would be in an ideal situation.
Wifi performance depends on the client characteristics and match. Signal strength must be at least -65dBm, and clean. MCS9 as VHT interface rate, 802.11ac only. No CAPsMAN. No retransmits (CCQ=100). A-MSDU and A-PMDU must be active at the max.values. https://www.duckware.com/tech/wifi-in-the-us.html
It think you already reached the maximum expectable single stream TCP throughput with mikrotik wifi. You might be able to get 10-20% improvements with careful tuning of parameters, but with current software it will never reach anywhere near your Cisco’s 800Mbps.
Btw: The latest stable and long-term release made a huge improvement for single-chain clients, I am now able to get up to 250 Mbps with these which is great. Unfortunately the performance just doesn’t scale well with multiple chains/antennas.