RouterOS v7.2rc1
Router's model RB5009UG+S+in

I have a 1000Mbps downlink connection to my ISP.
When connected via the 2.5G link on port 1 to the ISP cable modem, the connection speed maxes out at about 500Mbps
When removing the 2.5G negotiation speed on ethernet1 inside RouterOS and connecting to the cable modem, the speed is around 950Mbps

Steps to reproduce the issue;
Enable "2.5Gbps Full" advertise on Interface Ether1 -> Internet speed test goes to ~500Mbps
Remove "2.5Gbps Full" advertise on Interface Ether1 -> Internet speed test goes to ~900Mbps

Please see picture: Configuration export (use '/export hide-sensitive' command); This problem is affecting multiple users of RB5009, just a couple of examples:
viewtopic.php?t=179145
https://www.reddit.com/r/mikrotik/comme ... than_1gbe/
https://www.reddit.com/r/mikrotik/comme ... 25g_vs_1g/

Thank you.

Have you reported this on https://help.mikrotik.com/servicedesk ?
if not, I recommend you doing it, as via the forum they may not see it.

I can replicate this on 7.1.1 and the newly released 7.2rc3.

did you tested the same cable modem connected 2.5g computer or router?
the issue may be with the cable modem itself.
I had issues recently with an wireless access point that had a bug when connected to 2.5g port I had really slow speeds, that did not happen when connected at 1G.
Do you know if that particular cable modem have a Broadcomm SoC?

I have the same issue and there are a ton of others out there having it as well. In fact I haven't seen anyone claim that the 2.5G port works for them. It seems to be an issue when you go between speeds.

2.5G port - WAN
SFP+ port - S+RJ10 connected to PC at 2.5G

In this setup the 2.5G PC can get the full 1400mbps from the WAN connection. But any 1G device on the network suffers from the 500mbps download speeds. If the speed settings on the 2.5G port are changed to only advertise 1G then all devices are able to reach 900mbps.

I have the same problem with the 2.5g port but in an opposite configuration. I made my own post about it:

viewtopic.php?t=183165

In my configuration my WAN port is on a 1g port and my computer is on the 2.5g port on LAN bridge. Since my setup is opposite, my lower speed is on upload rather than download which indicates that the problem is 2.5g port sending. My behavior is exactly the same, if I remove the 2.5g advertisement and force the computer to connect at 1g my upload speed returns to normal. By the way my computer on the 2.5g port has an Intel I-225V NIC built into the motherboard.

I also had another problem with the 2.5g port that I'm not sure is related but may be. I noticed that at 2.5g if I upload large amounts of data to the internet (this occurred during my nightly remote backup), the 2.5g connection would go down. Windows would show a cable connected but no internet, renewing the IP address hangs forever, can't ping or connect to anything. All other devices on the RB5009 were fine and had internet, just my computer on the 2.5g port loses connection and never returns. Only thing I found that fixed it was to reboot the PC, never had to reboot RB5009 to fix it. It seems like this is a related problem because it only occurs when my computer uploads large amounts of data, downloading had no problem. Since removing the 2.5g advertisement and running at 1g, I no longer have this problem. For this reason the 2.5g port is unusable at 2.5g to me at the moment which is disappointing.

I have the same problem with the 2.5g port but in an opposite configuration. I made my own post about it:

viewtopic.php?t=183165

In my configuration my WAN port is on a 1g port and my computer is on the 2.5g port on LAN bridge. Since my setup is opposite, my lower speed is on upload rather than download which indicates that the problem is 2.5g port sending. My behavior is exactly the same, if I remove the 2.5g advertisement and force the computer to connect at 1g my upload speed returns to normal. By the way my computer on the 2.5g port has an Intel I-225V NIC built into the motherboard.
[...]

Hi,

I also have a MB with the Intel I225v, are you aware of it's hardware bugs? there is multiple offers to "fix it" but notthing actually effective. there is an fixed HW ver, but most of us got stuck with the bugged one. the only way to know if you have an affected one, is looking on the chip itself, under the motherboard heatsink.

I got mine working "okay" after a combination of a older driver and a nic firmware update. if you let windows update the driver you will get slow speeds, I actually had issues even when in 1g mode.
P.S: This test is a local intra-vlan to my DMZ that is in a server connected to the SFP+ port at 10G.

I'm getting about the same performance as jookraw using a similar setup though the 2.5G NIC I'm pointing it at is a RealTek rather than an Intel. It seems the RB5009 2.5G is ok talking to a RealTek. So far I've just been playing with it. I'm new to RouterOS and haven't deployed it yet. When I do I'll be pointing the 2.5G port at a Comcast-approved cable modem that I fortunately haven't bought yet. After seeing this Reddit post https://www.reddit.com/r/mikrotik/comme ... _than_1gbe/ I think I'll go with the Netgear CM2050V modem. It's one of my two options for 1.2G service. The other is a Motorola MT8733 WiFi cable router. I'd have more options if I didn't have phone service. I figured I'd wait to upgrade my modem until I got the RB5009 up and running, then after upgrading the modem upgrade my service.

The machine I have the 2.5G port on the RB5009 plugged into has a RealTek RTL8125 with firmware rtl8125b-2_0.0.2 07/13/20, running the stock driver included in RedHat 8.5. Test setup is basically:

RHEL8 "Server" RTL8125 2.5G - Cat6 - RB5009 2.5G - RB5009 SFP+ - 10G MMF - CRS326-24S+2Q+RM - 10G MMF - Solarflare sfn8522-plus "Workstation" Win10/RHEL8 dual boot

IPV4 IP V6 Normally the server machine uses a 10G Mellanox ConnectX4 and a DAC cable to the CRS, but I rearranged things to test the RB5009. The CRS is basically in dumb mode for now, though I'll be using VLANs once I get the RB5009 deployed.

There may be hope, Mikrotik seems to have reproduced the issue.

We reproduced a similar behavior locally. We are looking forward to improving such behavior with different link speeds in further RouterOS releases.

There may be hope, Mikrotik seems to have reproduced the issue.

We reproduced a similar behavior locally. We are looking forward to improving such behavior with different link speeds in further RouterOS releases.

Did they post this somewhere or is this in response to a support ticket?

Did they post this somewhere or is this in response to a support ticket?

Sorry, left that out. It was in response to a support ticket where we've been passing supout files, scenarios, etc. I'm eagerly awaiting the next testing branch update to scour over the change log that's for sure.

Thanks! I've been looking for updates on this issue every other day and this is promising news.

Did they post this somewhere or is this in response to a support ticket?

Sorry, left that out. It was in response to a support ticket where we've been passing supout files, scenarios, etc. I'm eagerly awaiting the next testing branch update to scour over the change log that's for sure.

I had similar problems with a Realtek RTL8125B. Turning flow control on fixed the problem.

I had similar problems with a Realtek RTL8125B. Turning flow control on fixed the problem.

For me, flow control was a completely separate issue but solved just general issues on my network with mismatch link speeds causing buffers to overflow.

I'm trying some test builds from MikroTik, will report back!

Three of the four test versions worked very well! Encouraging, I'm sure more to come soon from Mikrotik! I've restored to 7.2rc4 for now, not going to run those alpha builds

They had me test the alpha firmware as well and it definitely clears up the issue. Woohoo!

They had me test the alpha firmware as well and it definitely clears up the issue. Woohoo!

Did all work the same for you? One of them, the last one v4 seemed to still have issues on my side.

They had me test the alpha firmware as well and it definitely clears up the issue. Woohoo!

Did all work the same for you? One of them, the last one v4 seemed to still have issues on my side.

Same here, the first three were good and the last one had issues.

SergeP:
I've tested all four 7.99's. I use ether5,6,7,8 as LACP. I tried 1G and 2.5G on ether1. Looks like all builds do not increase speed above 1G in any settings. (Use different computers at the same time, It's combined speed). 1G +2G(LACP) gives 450Mbits + 450Mbits - total < 1G
V1: 2.5G - 937Mbits
1G - 938Mbits
V2: 2.5G - 935Mbits
1G - 938Mbits
V3: 2.5G - 860Mbits
1G - 935Mbits
V4: 2.5G - 425Mbits
1G - 933Mbits

7.2rc4 2.5G - 230Mbits
1G - 945Mbits

Direct cat6a 30" 2.5G NIC to cable modem 2.5G port gives 2040Mbits

heads up, from viewtopic.php?t=179145

Also, V1 worked best for me.

Same issue here. How can I request for the alpha firmware?

I can confirm I found noticed this just this morning on 7.2.1. I have gigabit cable service and my DOCSIS modem has a 2.5G ethernet port. Tests with 2.5G ethernet enabled get me ~600Mb/s. If I remove 2.5G from the ANEG profile I get 940 all day.

I turned TX Flow control on ether1 but it's not actually sending any TX Pause frames. I'm using a 10G SFP downstream, so looking at the block diagram I don't see anywhere that might actually call for flow control. Besides traffic from the 10G CPU interface towards the 2.5G ether1.

However looking at /interface/ethernet/switch/print stats does show many RX Pause Frames. Meaning something is overflowing and attempting to send pause packets internally to the router.

The 7.3 series works well for me with the cpu-flow-control on the switch. I'm running 7.3beta37 right now, and all is well for 10G, 2.5G and 1G hosts.

Something to note is that I had a lot of problems with my CRS328 switch in RouterOS, Mikrotik has confirmed tx-flow-control does not work correctly in RouterOS. I switched to SwOS and also tried other 10G/1G switches with flow control and I have zero issues now. I can saturate any mix of link speeds without issues on my network now.

I can confirm running v7.2.3 on RV5009.

At 2.5Gb I get 500Mbs - 600Mbs

When I turn it down to Gigabit I get full Gigabit around 930Mbs

Any chance of a fix making it into the v7.2 branch while we wait for v7.3 to hit stable?

Has the issue been fixed yet?

Has the issue been fixed yet?

I hope to be wrong but It can't be solved due of missing of cache in the switch chip, same problem of CCR2004

Does the new RB5009UPr+S+IN have the same problem or is it fixed on those new boards?

Has the issue been fixed yet?

I hope to be wrong but It can't be solved due of missing of cache in the switch chip, same problem of CCR2004

Everything seems fine here with `cpu-flow-control`

I hope to be wrong but It can't be solved due of missing of cache in the switch chip, same problem of CCR2004

Everything seems fine here with `cpu-flow-control`

Where is this setting located?

@kennethc it’s in the switch settings.

Does the new RB5009UPr+S+IN have the same problem or is it fixed on those new boards?

I'll wait until someone receives the new RB5009upr, states their usage case (WAN port vs switched port plus connected devices), and testing methodology with numbers before assuming anything is changed in the HW besides the addition of POE. I care about single stream, switched 2.5gbe port performance for both upstream and downstream but others may care about forwarding and multiple streams from one or more source ports. None of the options I tried which are exposed in ROS made a significant change including HW flow control and CPU flow control.

The fact that MT posts performance testing results is great and I hope they continue to do so. However, I don't feel their testing methodology covers my network very well.

having same problem working on rb5009 + odi DFP-34X-2C2 gpon onu device,just working on 1Gbps.
Mannul set to 2.5Gbps didn't working,test version on v7.4.1/v7.5/v7.5rc2

Does the new RB5009UPr+S+IN have the same problem or is it fixed on those new boards?

I'll wait until someone receives the new RB5009upr, states their usage case (WAN port vs switched port plus connected devices), and testing methodology with numbers before assuming anything is changed in the HW besides the addition of POE. I care about single stream, switched 2.5gbe port performance for both upstream and downstream but others may care about forwarding and multiple streams from one or more source ports. None of the options I tried which are exposed in ROS made a significant change including HW flow control and CPU flow control.

The fact that MT posts performance testing results is great and I hope they continue to do so. However, I don't feel their testing methodology covers my network very well.

Hi - I wanted to jump in here in hopes to help. I am slated to have one of these models deliver today. I bought one NIB from eBay and paid more than I wanted to. Below is my current network topology at home:
1. ISP: Spectrum Gigabit
2. Modem: Netgear Nighthawk CM2000
3. Current Router: TP-Link ER605
4. Switch: Cisco CBS350-16FP-2G
5. Wireless Access Points: Cisco Business 240AC x2 (these are only here so people can have Wi-Fi on phones/tablets. Everything else is wired.)
6. Spokes: Cisco CBS110-5T-D (These are little unmanaged 5 port switches that I have going to drops in the wall to extend out the network to multiple wired devices in different rooms)

I have a network cabinet in the basement as the "Demarc" point, where the modem/router/switch live. This is my first Mikrotik product, but I know enough about networking to be dangerous. That said, if you want to send me a DM I am happy to follow steps to get results you are looking for.

Piggybacking off of my last comment (I can't figure out how to edit the first one).

I've tried auto-negotiation on and off. I've tried flow control on and off. I've tried advertising at JUST 1000Mbps. I can't seem to even come close to 900mbps; high side is between 400-425, but upload seems unchanged, hovering between 35-40mbps.

I checked switch settings and CPU flow control was already enabled.

RouterOS was updated to 7.5 stable prior to being deployed. Unsure where to go at this point.

EDIT: since I'm brand new to RouterOS, I decided to completely reset and start over, following instructions on RouterOS documentation from help.mikrotik.com on a first-time configuration.
Since then, I am now pulling down between 920-924mbps with 38-40mbps upload - this is WITH 2.5g advertising enabled on WAN (note: all other advertising speeds were removed except 1000M and 2.5G. IPv6 is also disabled.). The only difference is that there was a local bridge set up this time. Not understanding what that would have to do with it, but as far as I'm concerned, I'm adopting the "if it ain't broke don't fix it" mindset.

Problem still there tried on the 5009Upr latest stable and beta with a Wireless Access Point (Netgear WAX610) and I get reduced speed and significant latency spikes compared to 1Gbps rate.

TLDR probably just a TCP problem the whole time. Disregard this post lol

I had a similar problem with my CRS317 (CRS317-1G-16S+RM) - bought back in april 2021.

Connected my EnGenius EWS377AP which provides 2.5G LAN on it. Got full 1Gbps throughput from my internet connection (gigabit service -> gigabit modem -> CRS317 -> AP -> laptop) and perfect 10Gbps throughput on local devices (desktop -> CRS317 -> local machine) but terrible 30-40 Mbps throughput on local LAN + Wifi (local machine -> CRS317 -> AP -> laptop). It drove me absolutely batty.

Fixed the problem with the help of this thread.
1) Booted the CRS317 into SwOS, since I saw mention of 2.5G and 5G being fixed in the release notes for some other CRSxxx device. Dunno about the latest ROS, maybe it has the same fixes.
2) Turned on Rx AND Tx flow control for all the affected SFP+ ports
3) In addition, my AP had STP disabled by default (configuration on the AP device). I'm not network-savvy enough to actually know what STP/RSTP truly do under the hood - but I have to assume having STP enabled on an AP is important for this use case (to span a subtree?). Then, the CRS317 detects the AP providing (doing?) STP, and sets the link to the AP as "point-to-point" instead of "edge" or "endpoint" (forgot the term). I didn't A-B test that one, but I am guessing it's important.

Edit, after some more experimentation:
When running iperf from wired machine -> CRS317 -> AP -> wifi machine, I see pause frames accumulating in their respective counters for both Rx on the one CRS317 port/interface (don't know the proper term), and Tx on the other. The throughput stays up around the proper real-world amount, matching the throughput from ISP -> CRS317 -> AP -> wifi machine.

So flow control is almost certainly working on SwOS 2.13.

However, going from wifi machine -> AP -> CRS317 -> wired machine is still slow, with wild swings in bandwidth, and some outright connection drops. There are no pause frames, as expected; we're sending data from a smaller pipe (2.5G from AP -> CRS317) to bigger pipe (10G from CRS317 -> wired machine). So unless there's something obviously wrong with my config that I don't know about (which is admittedly pretty likely), I may have issues with my AP after all. I'm going to swap it out for another one and see what happens.

Edit Edit
I was using iperf3 to test bandwidth, which defaults to TCP. Saw a bunch of retries, so I switched to UDP and the problem was fixed. (see https://software.es.net/iperf/faq.html under the section: "TCP throughput drops to (almost) zero during a test, what’s going on?") Probably some kind of fragmentation issue with my NICs/configs on client devices, and TCP freezes up/retries a bunch.

To saturate my ~866 Mbps wireless link, I needed two udp streams on two Zen2 cores, and specified the client MTU manually:
iperf3 -s -p 5201 & iperf3 -s -p 5202 && fg
iperf3 -c <server-addr> -T s1 -p 5201 -M 1500 -u & iperf3 -c <server-addr> -T s2 -p 5202 -M 1500 -u && fg

Any updates on this for newer rb5009?

Any updates on this for newer rb5009?

FWIW, it seems to have mitigated itself in ROS 7.8, but now thruput on the SFP+ port is much lower if the 2.5GbE port is also used.

I have the same problem

Any updates on this for newer rb5009?

FWIW, it seems to have mitigated itself in ROS 7.8, but now thruput on the SFP+ port is much lower if the 2.5GbE port is also used.

Could You elaborate on this please or post some sources?
I've been trying to use RB5009 with 2.5GBit GPON SFP+ stick (Leox LXT-010S-H) and whilst everything should check out, i seem to be only getting about 1.5Gbit/s so it sounds about right.

Actually ROS 7.9(?) fixed the problem I'd seen in 7.8 . I get full speed on both SFP+ and Ether1 now.

They had me test the alpha firmware as well and it definitely clears up the issue. Woohoo!

Hello, do you still have those alpha firmwares? I just bought a new RB5009 and have the same problem : 2.5 ethernet port connected to 2.5 ethernet from provider, maximum speed for clients is 600-650 mbs, if I put it manually on 1 G, then they can rich 900-950 mbs.
I've tested with os 7.9, 7.12, and even 7.13rc but problem persist. I would like to test with those alphas also, otherwise ... I will need to return it back, as old HAP AC2 was doing good job.

Thank you!

Supout.rif and ticket to support might be better option.
Already did that ?

I did that also, but I don’t have time to wait for their reply … when they have time for me, as I have only 7 days for testing for free, after these 7 days I can not return it.

Supout.rif and ticket to support might be better option.
Already did that ?

hahaah now THAT is a good joke ! I opened a ticket with them over a month ago and they haven't bothered to reply anything yet.

And long would you be waiting if you had NOT send in a ticket ?
Eternally ...

You can post a question in your ticket to ask about the status.

And long would you be waiting if you had NOT send in a ticket ?
Eternally ...

You can post a question in your ticket to ask about the status.

wow can't really argue with that kind of logic.. I guess in your book it's fine to not receive any acknowledgment of an open ticket for a month because no matter when they respond, I will have waited less than infinity time. (assuming they do respond) so it's like totally fine if they get back to me in 10 years, I shouldn't complain

Make no mistake, I don't consider such a thing fine at all, far from.
However, requesting an update yourself in an already open ticket, is something you can always do.

But not filing a ticket does yield in no response forever. This is always true.
The more persons file a ticket, the more the chances increase they will actually look into it in case it got really low on a priority list.
Simple law of numbers at play.

Make no mistake, I don't consider such a thing fine at all, far from.
However, requesting an update yourself in an already open ticket, is something you can always do.

But not filing a ticket does yield in no response forever. This is always true.
The more persons file a ticket, the more the chances increase they will actually look into it in case it got really low on a priority list.
Simple law of numbers at play.

This issue was opened almost 2 years ago, and it looks like others have opened quite a few tickets and issues (viewtopic.php?t=201496) with zero resolution. I'd call that a complete failure to address the issue at hand. I've used dirt-cheap off-brand stuff off of AliExpress that had infinitely better support, there's just no world in which paying customers should put up with this kind of treatment and end up being suckers.
I for one am planning on getting rid of all my MT gear asap if I don't hear from them by the end of the year. I'd rather save up and get equipment that is somewhat supported, not this Mickey Mouse stuff. I guess this is the reason why you don't hear about MT outside of a few YT hobbyists, the company is a joke.

On 5009UPr purchased in january 2024, Tested with 7.13 and 7.13.1 and the exact same problem continues to exist.
Rx overflow counter is going up, no tx pause sent by the router and tx speeds _from_ the 2.5gbit nic are only at ~500mbps(with 2.5gbit full duplex link) and ~900mbps(not 930-940mbps like it should be, with 1gbit full duplex link). In both cases, Rx overflow counter on ether1 keeps going up.


Edit 1:

* Rx overflow counter still goes up but every thing works perfectly(~940mbps bidir speeds) if I enable Flow control on the node connected to ether1, ether1 interface on the router and remove 2.5G-base-T from auto negotiation list.
Speeds drop back down to ~500mbps the moment I enable 2.5gbit on ether1.

Gahhhh! I have this same problem... So one thing I've noticed is that on a device connected at 1Gbps everything is fine bidirectionally... UNTIL there's some delay added... Add some delay on a linux machine with `sudo tc qdisc add dev eno1 root netem delay 5ms` (replace eno1 with your test interface) and you'll see the throughput plummet to the bad speeds.

The only real solution I've found is to remove hardware offloading on the 1-2.5 Gbps interfaces...

I am now real curious why I get tons of Rx Overflows but it never sends any Pause frames upstream. That would likely fix the issue.

This is the same problem as reported in another thread.
I have explained the solution there
viewtopic.php?p=1068203#p1067463