We are upgrading our 1Gbps upstream connection to a Level3 10Gbps and Cogent 10Gbps.
Should I go with a x86 based ROS router or the CCR1072-1G-8S+? Or should I be looking away from MT.

I love MT but I wonder if it can handle the traffic on the edge bgp router or even as the core router with simple queues reaching 500 to 1000.

Pretty sure the CCR1072 will easily do what you want it to...it can handle up to 80 Gbps, as verified by other members on this forum. Might even be considered overkill...

I was just concerned with the full routing table loading on a single 1200MHz cpu versus a 3500MHz x86 cpu. Thanks

the experiences shared on the forum point toward a virtualized implementation

someone shared a experience of a routeros virtualized on vmware esxi or vsphere with very good performance

meantime someone shared experiences of problems with ccr1036 an ccr1072 on specific scenarios of big queues

bgp convergence is unbeatable on a fast x86

ccr1072 is a very young product, give it time to be mature to this critical production scenarios

look at this

http://forum.mikrotik.com/viewtopic.php?t=96548

With regard to CPU,
CCR1072-1G-8S+ has 72 core cpu.

1 GHz x 72 = 72 GHz
So this would basically work better.

CCRC1072 can handle the task of 10G Easily.

With regard to CPU,
CCR1072-1G-8S+ has 72 core cpu.

1 GHz x 72 = 72 GHz
So this would basically work better.

CCRC1072 can handle the task of 10G Easily.

Unsure what you wanted to say with your comment on this thread 2 years later. But the info contained in your comment is inherently wrong too. Mhz =/= Mhz especially between architectures and because of how RouterOS works, the 1Ghz can actually be a very limiting factor for non-multi-threaded traffic. Also there are inherent single-threaded processes within RouterOS which will load one core to the max and won't scale at all.

So unless you can provide detailed personal testing for several scenario's, no. The CCR1072 can perform some functions very well but in other cases it will fall flat on it's face. Even in a home situation doing inter VLAN transfers I run into these limitations, a single stream tops out well below a Gigabit, let a lone 10Gbps. The same on a x86 box is no issue whatsoever, especially on a higher class platforms where there is a lot of IO bandwidth available.