Hi guys,
This is my first post, so I am hope I am doing it in the correct manner, and in a way that I am not wasting anyone's time. In short: the RB5009 seems to bottlenecking my connection and I don´t understand why. My topology is as follows:

Internet --> Fritz.Box (ISP Router) --> RB5009
note that the ISP provides me with a 1 Gbps internet connection

The setup is simple. I have a bridge which connects:
* SFP 10 Gbps module, that is connected to my the ISP Router via a 2.5 Gbps connection. I validated that connection in the "Rate" tab in the Interface specification and it negotiated a 2.5 Gbps connection.
* Ether 1, that has a 2.5 Gbps port
* Ether 2, that has a 1 Gbps port

If I do an online speedtest with ookla (https://www.speedtest.net/) I get:
* Full speed on Ether 1 that has a 2.5 Gbps port
* 600 Mbps on Ether 2, that has a 1 Gbps port
I have tested this several times comparing the ports with same cable. I have also swapped cables and got the same results.

My expectation was to get a similar speed on Ether 2. My reasoning is that 1 Gbps of internet should fit through a 1 Gbps port. Am I missing something?

### Solution ###
Fritz.Box Cable 6690 has weird issues with the 2.5 GBps port. Use the 1 GBps ports or downgrade the port to only use 1 GBps in the settings.

My reasoning is that 1 Gbps of internet should fit through a 1 Gbps port. Am I missing something?

A question: do you see similar poor performance on ether2 also for upload tests?

In reality the 1Gbps data "stream" via 2.5Gbps physical connection may indeed be 1Gbps on average but can be a tad bursty ... in which case the device doing speed donwgrade (from 2.5Gbps to 1Gbps) has to buffer packets. And then it may happen that the buffer size is not enough and device thus drops packets now and then. And for TCP dropped packets are a big issue. Things may get worse if your ISP actually offers 2.5Gbps (or even faster) physical line and only shapes traffic down to 1Gbps (or slightly more) ... traffic shapers don't always provide "smooth" throughput, it can be oscillating around target value. Which will again cause issues with (small) buffer in the "speed downgrading devices".

It may well be that you'd get better performance if you used a 1Gbps port to connect to your Fritz ... because Fritz might to better job at buffering traffic. And there aren't similar problems when physical link steps up. But in this case you may see poor upload speeds when your test machine is connected to 2.5Gbps port (while your ISP line is connected to 1Gbps port).

RB and Fritz connected via SFP? Try to set fixed rate of 1G for the SFP (instead of auto).

Thanks for your thoughts! So I did some more tests and the conclusion is the following. When I force the Fritz.Box to establish a 1 GB connection, I get full bandwidth. If it negotiates a 2.5GB connection, the connection slows down to ~550 Mbps. I tested the following scenarios:

* Fritz.Box 1 GB ports, tested all, in all combinations, and I got full speed
* Fritz.Box 2.5 GB port going into a 1 GB port of the RB5009, and I got full speed
* Fritz.Box 2.5 GB port, that was set in settings to act like a 1 GB port, and I got full speed using the SFP, 2.5 GB and 1 GB ports on the RB5009

Speed goes to shit when:
* Fritz.Box 2.5 GB port goes into SFP Module or 2.5 GB port and negotiates a 2.5 GB connection.

But I am still not sure what this means and what I am doing wrong.. Should I connect to ports to the RB5009 on 1 GB to optimize the throughput?

So further investigation has proven the problem. The Fritz.Box 6690 Cable is shit. The 2.5 GB causes all sorts of issues. Here is a list of forum posts:

• https://forum.vodafone.de/t5/Ger%C3%A4t ... -p/2977493

• https://www.vodafonekabelforum.de/viewtopic.php?t=45576

• https://hilfe.o2online.de/dsl-kabel-gla ... lan-633923

TL;DR: The issues seem to no be consistent, but the ETH1 port, the only one that supports 2.5 GBps has issues connecting or keeping a connection. This is a Fritz.Box issue, not a Mikrotik issue.