I have a setup like attached to this post.

From left to right:
- line is my internet uplink
- RB1000 [1] is my first RB1000 device
- next line is a (utp) connection between the two RB1000 devices.
- the right RB1000 [2] had a bridge group with a ether1
- On the right RB1000 [2] there is a MetaRouter (connected to the bridge by a vif)
- The meta router is connected to ether2
- Ether2 is connected to a host machine by utp.
Nothing strange so far.

If there is a 'large' file transfer, and the connected speed between the internet and the host machine will go over 10Mbit/s, the MetaRouter cpu load is going up above 40. The load on the second RB1000 [2] is also going up with the same amount (logical). The load on the first RB1000 [1] is stable between 1 and 3% cpu load. Both RB1000 and the metarouters are only setup to forward this traffic, and no filtering or logging is done on these routers.

I know that a 100% cpu load automatically means that it hasn't no more capacity to route the traffic normally, but why is load in the Metarouter at 40% cpu and on the other fysical RB1000 still at 1-3% cpu load?
And can i see what is causing this load?

this is how metarouter works. Load is caused by metarouter handling packets. initial CPU hit is high, but you can push reasonable amount of packets through due to a way packets are handled by RouterOS.