Hello, i upgrade my wireless network from 3.13 to 3.15 and i have same problems,
the first, rate limits works but random i need to restart the pppoe server because the rate limits not work properly, i suppose queue management problem.

more pppoe client die with peer not responding message but before upgrade it works fine.

help me please

Another problem is that the interface traffic that i read in with winbox or telnet is lower then the real traffic, it can be a cause of pppoe connection termination , because when i read 700 Kbps the real traffic is 1200 Kbps, then if the router ack is 1 or more bps, the interface not read it.

if it is help for you to understand my problems.
i 've downgraded all the network to 3.14 and the problems appears solved

What platform are you working on? Is it x86? If it is, it seems that clock issue on x86 platforms with v3.1x RouterOS is the reason. With traffic growth clock speeds up and everything that depends on it fails. Queue does not limit correctly, interface traffic shows smaller data rates than real etc...

MT said they solved that problem on v3.16, but in fact they did not. We tested v3.16 on several x86 platforms (PC, PC Engines Alix boards etc.) and although clock is still quicker than it should be it is "only" 2-5 times faster than realtime. On v3.14 and v3.15 it was over 10 times faster.

We even have situation on v3.16 (x86) that CPU is counting 3-4 times slower, and users have 3-4 times less bandwith. For example, MT queue stats show full 1Mbps and users cant get more then 256k downloads...
Is there hope that MT will fix this bug ?

We have been having the same issues with 3.15. We have around 3,000 uses whose BW we manage through using PCQ for both upload and download. When we used to be on 3.6 or 3.7 version, the queues worked pretty fine, and all users would get the proper bandwidth.

However, as soon as we migrated to 3.15, we were facing problems. Users would complain that they are not getting their bandwidth. A Torch on the 3.15 router would show that the users were using the full limits assigned. But, that users is actually behind another MT using 3.7 device which does not do any natting, just simple routing. Checking the Torch on the interface of the 2nd MT that the user is connected on shows that the user is getting what the user is claiming, that is less BW.

Another interesting thing to note is, during these times, the Interface would show abnormal usage, say instead of regular 40Mbps, it would show 65 or 70 Mbps. Whereas, on the Queues, it would show only 10 or 11 Mbps. Sometimes, the data would be reversed, meaning more on Queues, but less on Interface.

Is there any fix to this? Mikrotik?