CRS without fasttrack as a router - thats definitely cause of the issue. It simply does not have enough CPU power.
I am not sure if you don't have fast track on purpose (it can't be enabled if you want to use simple queues, ipsec and
some other features)
or if you don't have it by mistake. It definitely speeds up things a log (basically, if packet hits fasttrack rule, it will mark the connection and most subsequent packet within that connection will skip some processes = causing less cpu load).
If you have RB3011 it will definitely improve your experience a lot:
RB3011 is full-featured router with
2 cores on 1.4GHz, meanwhile CRS-125 has only
1 core on 600MHz. Simply said - it has 4.6 times more processing power. (also way more RAM which helps with some tasks) In addition, CRS125 has only single 1gbit line between cpu and switch chip, which means you have any CPU-processed data limited by hardware to 1Gbit half-duplex.
Is there any particular reason why you decided to use CRS125 as your router instead of RB3011? Although I would recommend to use RB3011 I also want to make sure, that you don't have any legitimate reason for this setup.