Hello Moritz,
The shortest/simplified answer as to why inserts on partitioned tables is faster is that each partition is being worked upon independently. So if you have 12 partitions then 12 cores/threads will be working in parallel hence your seeing your CPU consumption. You can see this behaviour when you run "top -H" and you'll see the running threads.
I don't want to try and paraphrase the contents of a good article by john_appleby on the subject so I'll point you to the article directly which will help you better understand what's going on and some best practices for data loads into HDB.
Best Practices for SAP HANA Data Loads
KR,
Amerjit