Interesting. So instead of addressing the actual datapath, we’re now collecting “friend of a cousin on Wednesdays” test cases.
Let’s keep it technical:
The statement “ether1 → CPU → SoC switch → ether2–5 = NOT OK” doesn’t hold up in practice.
I’ve tested this setup and can consistently reproduce different behavior.
If you have actual data or a technical explanation for the discrepancy, I’m interested. Otherwise, dismissing real measurements as “noise” doesn’t move this forward.
Well, if you post your highly specialized, technical, real, measurements as apodiptical anecdata, without context (noise), I cannot but treat them as such.
Your experience is different from Massinia's one, OK, and then?
Do you believe that since it is not happening to you then it isn't happening on Massinia's setup? (is he lying or pervceiving an altered reality?)
Or maybe simply the conditions (configuration/ISP/connections/whatever) are different and they lead to different results?
Hard to say with the data available on the thread.
Anyway, it seems that suddenly an SEP field was erected on this topic:
https://en.wikipedia.org/wiki/Somebody_else's_problem
I think you might be getting confused.
By “NOT OK” we mean that the upload speed is limited, not that the connection isn’t working.
Did you also use a 2.5G ONT for the tests?
Signed up to add another datapoint here. When doing speedtests with my 1G internet connection. I was getting constant ~1000Mbps down, but varying upload speeds. I was getting about 500-600Mbps upload to my ISP’s speedtest server, but about 100Mbps to everywhere else.
Initially I thought that my ISP just has terrible peering or something, but this thread made me get a random 2.5Gbs SFP module to try it out. After switching my WAN connection to that, I’m getting a consistent ~1000Mbps to speedtest servers in my country and ~800-1000Mbs to nearby ones. So 
DKT Comega 79761 ONT → cat6 → 2.5G RJ45 SFP, if it helps anyone. Keep in mind networking isn’t really my thing.
Thank you for sharing your experience as well.
Unfortunately, MikroTik support still hasn’t responded to my ticket after a month.
Maybe it’s not important for them…
If you have some free time, please open a new ticket too so they understand this is not just my issue.
Hello,
I just bought this router too, and I'm having the exact same issue.
When I use Ethernet 1, all wired and wireless devices have a limited upload speed.
If I use another port as WAN, wired devices get fast upload speeds, but wireless devices remain capped at 60/100.
I am using version 7.24.1.
The router has just been reset, I have only configured the WAN.
Do I really have to buy an SFP module?
I'm disappointed, I wasn't expecting this issue...
Thanks
Sometimes 2.5G is buggy. I'd say get a cheap dumb 2.5G switch from AliExpress or Amazon between the ZTE and MikroTik.
Where you live, does each ISP own their wires (overbuilding) or are the wires shared (unbundling)? If it's the former, try another ISP maybe?
Thanks for your reply.
My ISP's ONT is 10G, not 2.5G; the contract is 1 Gbps symmetric.
With the ISP's router I reach 940/940, but it is very limited in terms of features...
I will try with a cheap SFP module to see if the upload over wireless works properly, it costs me less than a 2.5G switch.
If you are getting the expected upload speeds on wired clients, the problem is not between your router and the ONT, rather the bottleneck is probably the wifi. If this is the case, an SFP (or a switch) is unlikely to help.
I haven't tested it again recently without the SFP module, but your message confirms that the issue is still present.
In my case as well, if I use ethernet port 2-3-4-5 as WAN, the upload limit on wireless remains.
The ticket (SUP-214001) I opened back in May hasn't received any further responses; they can't reproduce the issue.
If you have time open one as well...
I can confirm that with the SFP module upload works fine over wireless too.
Do you have a model to recommend?
Thanks
It doesn't seem like a wireless issue; the download reaches about 550/600 Mbps, while the upload is always stuck at 60/80.
If you're in the mood for experimenting a bit, (which may be a waste of time,) you could try enabling flow control (both tx and rx) on the WAN port, and set it's interface queue to default-large. Of course, this is for the situation when the WAN port was already moved to a port other than ether1.
The next step would be to increase the default-large queue depth up to 500.
This is the usual suggestion for these situations, so it would be useful to know for sure whether it helps, even if it's a negative.
I don't have default-large 
@massinia Thanks, I'll buy it.
While I wait for it to arrive, I'll try to solve the problem with queue.
Yep. My mistake. Just define a a queue type with, let's say, kind=pfifo pfifo-limit=100
The SFP module arrived and installed.
I confirm that the wireless upload now reaches 500 Mbps with the CPU at 100%.
I was expecting a bit more, but it's fine anyway.
I also get similar performance, about 550/600 with CPU at 100%.
I don't think there's anything else that can be done to improve it... only MikroTik can do something with the wireless driver.
That's pretty embarrassing. These speeds, 550-600 I get with cap ac + wifi-qcom-ac. And the iperf3 benchmark running OpenWrt on that device clearly shows that the hardware is capable of way more.
Looks like they are doing "something" with the wireless driver, just not the outcome that you are expecting. This is by no stretch of the imagination, considering the choice to compile the system around -march=armv5 instead of -march=armv7, which the SoC is very much capable, resulting in less than ideal support for container images.
Yes, I was expecting better too; with my old access point, I had the same speed.
I'm a bit disappointed...
I use it as a simple AP without routing or VLANs.
