I have an RB1100AHx2. My speeds from ISP are 50meg up and 50 meg down. If I plug my laptop directly into my line from ISP I get advertised speeds. Plugging the same cable into my router gets me 50megs down and 25 megs up. I have messed with MTU settings with ISP as well as auto negotiate and no change. I have bandwidth set to unlimited. Any idea one what the issue may be?
I have same problem, did you even solve this?
Configuration and/or cabling issue.
P.S. and/or capacity (not)planning issue
Hi - I’d tend to agree with NetworkPro – that being said:
MSS settings adjusted as well or only MTU on the Routerboard? When running via the routerboard - are you the only user connected during testing?
How are you running the tests – proper iperf or similar I trust?
It has something to do with ethernetports. When i put a file on FTP server of Mikrotik I can down/uploadload it with a speed of 50/50Mbps. A PC behind the Mikrotik will not get this throughput. I tried several different PC’s and tried different settings in MSS and MTU. Witch Cisco 800 series or direct PPPoE from PC I have no problems.
This is a less than helpful observation. What someone who has this issue on their router would like to see, is I have had the experience and it was fixed by making “this” configuration change. The poster stated he tested the cable so we can assume it is a configuration issue of some sort. The million dollar question since this has not been answered is what configuration and has this been fixed.
I have a similar issue with a RB2011 now running 5.21. I was only pulling 3-5 Mbps of 50 available. I had to change the lan speed on the port to manual 100 full duplex and it helped some. I’m now pulling 40 Mbps ish on the download but only about 25 on the upload. This is however, on a basic speedtest site and not on iperf.
I still have testing to do, but it does not seem that I have all the fine tuning that I need to get the throughput available. I have seen this on a couple different routers / RB versions and finaly got frustrated enough with it to come look on the boards. I’m hoping someone has posted somewhere with performance tuning recommendations? Anyone have any ideas?
Thanks - paul
I like short posts - easy to read.
RB is tuned out of the box.
Speedtest on a website rarely fills up providers’ advertised speed.
Download many well seeded torrents and see amount on traffic on RB port.
p.s. port auto-negotiation issue is not “performance” or “tuning” issue - possibly the makers of the chips/devices on both ends did not have an interoperability lab. In the industry this issue is “owned” by whoever is setting up the connection. Congrats on making it work on 100 (or 1000?) full duplex.
Did anybody found a solution for this problem? I’m fighting with this issue since the first day I have MT router ( ROS 4.x) . I have multiple mipsbe routers and all of them shows same behaviour. The first version which shows significant improvements is ROS 6.0rc5.
I have 100/100Mbps dedicated ports on a Cisco switches in a DC. Without MT router 100/100 is not an issue to anywhere in EU, but with MT router local sites are running close to 100/100, but distant sites are running 100Mbps down /15Mbps Up. Which is very very sad. This looks like some kind of src-NAT-ing / connection tracking / packet queueing issue.
With 6.0rc5 I have 100/~30Mbps, but if I upgrade to 6.0rc6 it’s back to old “good” 100/15. That’s sucks. Web-Proxy is not a solution for me due to multiple reasons. Any creative advice are welcome. I’m considering to purchase CCR router, but if that one has same issue, I will laugh an go with a basic Cisco ASA for “slightly” more money… 
http://forum.mikrotik.com/t/slow-upload-but-with-proxy-enabled-it-is-fast/61614/8
Some speedtest results. Same config different ROS version:
Budapest - Belgrade
ROS 6.0rc5: http://speedtest.net/result/2415260981.png
ROS 6.0rc6: http://speedtest.net/result/2413658236.png
Budapest - Los Angeles:
ROS 6.0rc5: http://speedtest.net/result/2415233995.png
ROS 6.0rc6: http://speedtest.net/result/2415065789.png
Budapest - Frankfurt:
ROS 6.0rc5: http://speedtest.net/result/2415238823.png
ROS 6.0rc6: http://speedtest.net/result/2413982012.png
Budapest - Amsterdam:
ROS 6.0rc5: http://speedtest.net/result/2415244939.png
ROS 6.0rc6: http://speedtest.net/result/2413533030.png
Budapest - Sydney:
ROS 6.0rc5: http://speedtest.net/result/2415259065.png
ROS 6.0rc6: http://speedtest.net/result/2414344763.png
Well Australia is issue regardless of ROS version. Without MT router it’s about 20/15.
This material is no proof of anything. Unless your ISP makes proper engineering measurements (they know and control their network) - you cannot yet prove the issue is in the router that you have.
I have 100/100Mbps dedicated ports on a Cisco switches in a DC. Without MT router 100/100 is not an issue to anywhere in EU, but with MT router local sites are running close to 100/100, but distant sites are running 100Mbps down /15Mbps Up.
can you make an drawing when you are satisfied and example drawing when you observe the “problem” ?
Can you remember if all conditions are the same when you compare? Including hour of the day 
Don’t be sad, OK - you can figure it out I am sure.
What model router are you trying this with and with what amount of packets per second VS megabits ? 
Above speed tests are not proving that MT is good or bad, they just proves there is a performance difference between two release candidates. ISP can’t guarantee anything since they have dozens of peering partners use lot of IP transit providers and path to the destination is dynamic (lot of hops, all under different ownership) - So real Internet. There is two things we know. MT is attached to core switch which has several 10Gbps uplinks everywhere. Destination link is underutilized on first two hops (rest is unknown), but we know that it works as expected if server is attached directly to the network without MT router. It works with Cisco ASA as well, even more OpenWRT based SOHO toys like LinkSys, TP-Link, D-Link stuff works properly. So even I can’t prove it, tests shows that something isn’t as it should be in RouterOS.
Well I have RB750G, RB750GL and RB2011UAS devices. All of them shows same issue. I’m testing with RB2011UAS right now. I couldn’t make a any configuration which doesn’t shows this behaviour. So problem exists even with factory default configuration. DHCP Client on WAN, masquerade, DHCP server on LAN; nothing else - so pure out of the box configuration. There are no packet drops on interfaces checked on both switches.
Time and ROS version was different on above links. Time is an issue, but I did lot of tests and I can confirm that time of day is irrelevant in this case.
I’m using RB2011UAS right now. SpeedTest use pure http file transfer like traffic, big packets on one way and small ack/syn packets in another way. Nothing special. Router is not overloaded; cpu peaks at <60%, but this is in case it can reach 80-90Mbps. During problematic uploads CPU runs on ~10-15%. (LCD is disabled since it consume 5-8% of CPU power).
Do you have any Queues in your config ?
ISP can’t guarantee anything since they have dozens of peering partners use lot of IP transit providers and path to the destination is dynamic (lot of hops, all under different ownership) - So real Internet.
can you test again and this time include traceroute with each speedtest.net test - you will have to traceroute to the actual IP that is the actual bandwidth test server (not speedtest.net)
P.S. check the source IP address (see a whats my IP site) is the same when you do comparison tests !!!
Hi NetworkPro,
There is no queues except of Interface Queues / only-hardware-queue on all interfaces.
Thankfully I’m happy man right now, since all destinations are running nicely using my RB2011 router. Even more I have nice results to Australia: http://www.speedtest.net/result/2418348959.png, If anyone can do this better from EU I’m looking for his/hers results. 
Thanks for your assistance.
Alright what was the issue after all man ?
I have identical problem on RB951 and RB2011 , 100 Mbit Download / 50 Mbit Upload and only half upload when using Mikrotik ( 21-22 Mbit ) , when connecting directly without router I have 50-52 Mbit.
I was trying everything , different versions of ROS , totally changing all network cables to cat6 shielded, changing of MTU / MRU / MSS / TTL,connecting using additional switch between pc and router also between router and internet, changing port speed in Mikrotik as well in PC , changing ports in Mikrotik to Master & Slave as well as just bridge them all, changing network cards in PC , trying different drivers …
And nothing works
Refresh.
Refresh , problem still existing in ROS 6.1
Same problem here. Upload speed to Internet is very low when directly connected to a MT Router (CCR,x86,RB750 all the same) but perfectly fine when connected by PPPoE.
I think MSS or MTU problem.
Anyone has a solution ?
forgot to add , I am using PPPoE too.
Do you use change MSS in your PPP profile ?