Device tests results - minimum value of "Mbps" for a single true 1Gbps connection

RouterOS Beginner Basics
1

The Mikrotik products are published with “Test results” which show the “kpps” (Kilo Packets Per Second) and “Mbps” values for 3 packet sizes:

  • 1518 byte.
  • 512 byte.
  • 64 byte.

and various configurations:

  • Bridging none (fast path).
  • Bridging 25 bridge filter rules.
  • Routing none (fast path).
  • Routing 25 simple queues.
  • Routing 25 ip filter rules.

An example is available for the hEX device - product’s webpage URL:
https://mikrotik.com/product/RB750Gr3#fndtn-testresults

I would like to verify my understanding regarding the “Mbps” column - as follows:

I seek a device that is capable to perform “1Gbps connection” - meaning:
A user in device’s LAN sends data such that a server (from WAN side) would receive it in a rate of 1Gbps.

If the “Mbps” value in the “Test results” show a value which is under 1000, then I know that for the corresponding configuration and packet size that I would necessarily WON’T have a “1Gbps connection”.

Is that correct?

2

Yes.

3

But it extends futher: even if a device has “Mbps” value in “Test results” which exceeds (slightly?) the threshold (e.g. 1000), it doesn’t necessarily mean that router is capable of push all the data through single port. Table column header says “all port test” and I seem to remember that the explanatory remarks under table used to state, that results are total device throughput (i.e. sum over all ports). And that also implies multiple clients communicating concurrently. There is a common knowledge that single connection performance on vast majority of devices is lower (considerably so on some devices) than achievable cumulative result. The reason is that ROS processes packets of same connection using same CPU core (so that it avoids out-of-order delivery) and then it boils down to single-core performance. Devices with many cores (primarily CCR1xxx and some CCR2xxx) can thus see a large discrepancy between single connection / single client performance and cumulative performance (when there are number of connections/clients going on at the same time).

4

I appreciate the emphasis.
Indeed, for this reason I carefully formulated my question to include these scenarios.
So, my question could’ve stated also something like:
“If Mbps value is 1000 or higher, then 1Gbps connection is possible though not guaranteed.”

I’ve read somewhere that the Test results actually mean “packets leaving the device”.
So, packets entering the device are excluded from the tests results.

5

Also take into account a single true 1Gb network connection typically only goes to about 950Mbps. Never 1000 and certainly not 1024 Mbps.

6

Basically, the tendency is if you have a 1Gb fiber connection is to
a. for wifi router or home router go with hapax3
b. for business router go with RB5009

If you have something less and are just looking for a modern processor ARM, to do things like wireguard
c. recommend hex refresh.

If you have something less and are looking for a modern processor and more ports then hex refresh
d. recommend L009

7

Check also this seemingly unrelated thread (in a nutshell, it is not exact science):
http://forum.mikrotik.com/t/wan-capable-ports-on-routers/182470/1