Look at the product pages for benchmark testing. It all comes down to packets per second, what kind of services you are running on the board, number of firewall rules, etc. Everything that takes up CPU time reduces the number of packets a router can process.
So the better question is, what is the application of the router? Where will it be placed, and what kinds of services do you want to run on it?
seems nobody is interested in real testing, i've asked some numbers i am interested in this post http://forum.mikrotik.com/viewtopic.php ... 89#p267589 but everybody is obsessed with just switching speed. Seems that i will have to setup my own testbed. I have RB450G with 100mbit/s, 100mbit/s and 40mbit/s FTTH, doing PPPoE, NAT, firewall and mangle for PBR. It's production installation, so i'm not able to do much test, but i have seen 45% CPU utilization with 30mbit/s traffic, so i seriously doubt i can reach 240mbit/s.
it depends on what want to do. You know, mbps is just one side of the coin.
I've tested some routerboard with iperf testbench, and 450G can bridge (or route) more than 500mbps of data (with normal MTU). Small MTUs will cause much higher CPU load (so with 1500-300byte packets probably you can reach only 100-150mbps).
For 1000mbps...
1100AH (overclocked) can do 1.2gpbs (980+320) - or ~700 with conntrack on. CPU maxed
1200 (and RB800) about 900mbps, CPU maxed.