Although not mainly a router, the CRS317 test results show that it is able to route and bridge at over 3000Mbps, but the block diagram shows only a 1Gbps connection to the cpu.
Is there some other magic that is going on, or is the connection to the CPU actually 10Gbps? Has anyone checked actual routing/software bridging throughput independently?
https://mikrotik.com/product/crs317_1g_16s_rm#fndtn-testresults
https://i.mt.lv/routerboard/files/CRS317-16splus-180109085108.png
It can switch 3000 Mbps (1500 up and 1500 down). It can’t route this much. While switching, all the hard work is done inside the switch chip - it never reaches the CPU.
Well it can switch a lot more than that between all ports, 158,906Mbps. I’m strictly taking about cpu bridging/routing, which the chart shows the CRS317 can actually software bridge 1518 sized packets at 3258.2Mbps (route at 3089.4) yet the block diagram shows a connection of only 1000Mbps.
Sorry, my mistake. The first two lines are “bridging” - which is done in hardware. But the last three lines are routing indeed. That is weird. It must be a typo, or something like this. You are right: I can’t see how this would work out.
The connection from switch chip to CPU of CRS317 is not 1Gb/s, it is an internally connected bus and the speed is not documented.
We will fix the diagram.
https://i.mt.lv/routerboard/files/CRS317-1G-16Splus-180223102556.png
Thank you Normis!