We tested v3 RC13 and it has big problem when running 300+ PPPoE connections at the same time. CPU load is very higher than in 2.9.49, around 80%, when in 2.9.49 is about 50% on the same hardware.

Packet Loss through tcp/ip connections is non stop, measured with tool smokeping:
1.80% average and 9.08% high.. through 8 hours of test time to few IP addresses.

I suppose the problem is CPU load as when there is lower number of concurrent PPPoE connections than 150 there is no Packet Loss.

When running 500+ pppoe connections under V3 FINAL packet loss is lower than was in rc13...

You use Masquerade/NAT 1:1 or Connection tracking ?

You use Masquerade/NAT 1:1 or Connection tracking ?

No NAT/masquerade.

Of course that we use connection tracking... as in 2.9.49 version...

After my first post in this thread we done more testing.

So, conclusion: It must be problem of bad optimization for Intel CPU's. Even if CPU load is not 100%, sometimes 100% appears for flash of a second which is not the case in 2.9.49

We disabled some stuff, such as connection tracking, mangle mark packet and
after this actions packet loss is lower. CPU Load shows around 30-50% but THERE is still PACKET LOSS.. Lower for 0.2-0.3%...

Is this solved in RC14? Does somebody tested RC14 for packet loss when having more than 150+ pppoe tunnels?

It is not the pppoe - it is simple queue for each user, thats where the problem is. go to the "queue type" and increase size of the queue - it should solve the problem.

It is not the pppoe - it is simple queue for each user, thats where the problem is. go to the "queue type" and increase size of the queue - it should solve the problem.

What value you recommend?

But there is the same values like in 2.9.x and in v3 there is about 2-5% packet loss when running 200+
up to 10% when running 400+
up to 15% when running 600+ pppoe connections

When you compare to 2.9.x there is no packet loss in 2.9.50 version.. So, it is bug I suppose... Tested with RC13

Is it possible that somebody of you in mikrotik community test v3 with ping tool such as smokeping for packet loss for any traffic that goes through that router which have minimum 150+ pppoe connections...

I can confirm that there is no packet loss at maximum 130 pppoe tunnels, when it goes more than 150, there is imediatelly, even when cpu load is around 20-30%

Also, I saw v3 full demo, I ask somebody from Mikrotik Support if it possible to obtain .npk file for this full version, for testing of pppoe tunnels for packet loss?

I confirm that there is the same problem with packet loss in v3 final when having 150+ pppoe tunnels.

This is not solved.

We will downgrade to 2.9.50

If somebody is interested we can post image from ping tool as proof in v3 final where you can see packet loss, and image when running 2.9.X without packet loss on the same hardware configuration.

I have more than 200 pppoe clients and haven't noticed any packet loss . Are there any packet loss for real traffic? Probably your pingtool is calculating something wrong.

I have more than 200 pppoe clients and haven't noticed any packet loss . Are there any packet loss for real traffic? Probably your pingtool is calculating something wrong.

This is the same answer I got from mikrotik support. You didn't notice as you didn't watched for it It can be a real pain in the ass when you have streaming through this router.

Smokeping is very precise. Also, we made manual ping through this router and it is the same.

Why then there isn't packet loss in 2.9.x

When we noticed this in RC13, packet loss was very high under 600+ pppoe tunnels, and now in v3 final it is constant 5% measured by smokeping.

Also, when made manual ping with 65000 from windows, there is 50% increase in ping times with few time outs...
Notice: There isn't any increasement or time out when running 2.9.X under any number of pppoe tunnels.

3.0 final, approx. ~270 pppoe clients, packet loss is here:
Ping statistics for 194.87.0.50:
Packets: Sent = 500, Received = 408, Lost = 92 (18% loss),
Approximate round trip times in milli-seconds:
Minimum = 9ms, Maximum = 47ms, Average = 17ms

3.0 final, approx. ~270 pppoe clients, packet loss is here:
Ping statistics for 194.87.0.50:
Packets: Sent = 500, Received = 408, Lost = 92 (18% loss),
Approximate round trip times in milli-seconds:
Minimum = 9ms, Maximum = 47ms, Average = 17ms

Welcome to the Club of PPPoE Packet Loss

I ask every mikrotik expert in community to test pinging when have more than 150+ pppoe connnections, so we can help Mikrotik Support to resolve this problem as soon as it is possible.

It is not the pppoe - it is simple queue for each user, thats where the problem is. go to the "queue type" and increase size of the queue - it should solve the problem.

By default it is default-small queue with only 10 packages allowed in the queue (packet 11 will be dropped). Go to the /queue type increase this number step-by-step until you will get rid of packet loss. Why step-by-step? Each step will increase a latency a little bit.

Solution without any minuses is disabling dynamic simple queues and create queue tree using PCQ.

It is not the pppoe - it is simple queue for each user, thats where the problem is. go to the "queue type" and increase size of the queue - it should solve the problem.

By default it is default-small queue with only 10 packages allowed in the queue (packet 11 will be dropped). Go to the /queue type increase this number step-by-step until you will get rid of packet loss. Why step-by-step? Each step will increase a latency a little bit.

Solution without any minuses is disabling dynamic simple queues and create queue tree using PCQ.

Tried, Not helped.. This is not the problem. The same quaeue size is in 2.9.48 where we didn't encounter any packet loss. So, that is the problem in V3 which make this packet loss.