Hi Guys.
First time Mikrotik user and I really like it, I have a HEX to play around and learn before buying the CCR version, I got everything working but I can't understand 1 thing, why is speedtest.net maxing out only 1 core but for example speed test in DSL reports uses all cores? so in speedtest.net, I get slower results than I get in DSL reports.
Thanks in advance
Chris.

Hi Guys.
First time Mikrotik user and I really like it, I have a HEX to play around and learn before buying the CCR version, I got everything working but I can't understand 1 thing, why is speedtest.net maxing out only 1 core but for example speed test in DSL reports uses all cores? so in speedtest.net, I get slower results than I get in DSL reports.
Thanks in advance
Chris.

It's because speedtest is one connection count.
If you want to see what your costumers see, use TCP connection count 1 in bandwidth test.
If you want to see the throughput of a wireless link use TCP connection count >1 (20 for example), or UDP if you don't have drop input for udp on client.
I've also noticed that, if you make bandwidth test in both TCP with connection count >1 and UDP, TCP would be lower than UDP is some cases, but when your link does more throughput than TCP throughput you saw with test, your clients to slower and with more latency.
For example: TCP count >1 500Mbps UDP 700Mbps, when a link goes above 500Mbps, clients latency increase to 100ms and bandwidth is also decreased.
This behavior cames from CPU usage.

Thanks for your detailed and well explained reply.

Is this a limitation with Mikrotik? as other routers I used always distributed among all cores even though speed test is a single connection

Thanks for your detailed and well explained reply.

Is this a limitation with Mikrotik? as other routers I used always distributed among all cores even though speed test is a single connection

Maybe other routers provide only multiple TCP connection count. Did you noticed differences between speedtest.net and other routers bandwidth tests?

Thanks for your detailed and well explained reply.

Is this a limitation with Mikrotik? as other routers I used always distributed among all cores even though speed test is a single connection

Maybe other routers provide only multiple TCP connection count. Did you noticed differences between speedtest.net and other routers bandwidth tests?

Sorry did not understand exactly. As for other routers if I replace MikroTik with example a different brand I do a speedtest.net and I see all cores used. So if speedtets is 1 connection and Mikrotik will use 1 core this would be a problem as MikroTik has routers with many cores but low-frequency example 9 core but 1.2Ghz for CCR 1009, so if I do speed test on CCR 1009 I will see low speed as it maxes 1 core.

Thank you gain

Thanks for your detailed and well explained reply.

Is this a limitation with Mikrotik? as other routers I used always distributed among all cores even though speed test is a single connection

Maybe other routers provide only multiple TCP connection count. Did you noticed differences between speedtest.net and other routers bandwidth tests?

Sorry did not understand exactly. As for other routers if I replace MikroTik with example a different brand I do a speedtest.net and I see all cores used. So if speedtets is 1 connection and Mikrotik will use 1 core this would be a problem as MikroTik has routers with many cores but low-frequency example 9 core but 1.2Ghz for CCR 1009, so if I do speed test on CCR 1009 I will see low speed as it maxes 1 core.

Thank you gain

Not sure about core usage for TCP connection count >1, but UDP speedtest doesn't use only one core, or at least there's no 100% usage for one core. Maybe also TCP connection >1 doesn't use one core, never checked that, but for sure throughput is higher than TCP count 1 in both cases.

This is a known issue with RouterOS v6. Something to do with the kernel / connection tracking most likely.

See also viewtopic.php?t=131503

Thanks hope this will get fixed