Quote: "if you mean regular single core CPU - with a single CPU on the mother board total... I disagree."
I mean one single core CPU on a bord, which *does not* support mutiple hardware threads. And I must insist on what I said.
Of course, if there were only one job to do which were poorly paralelizable (strong dependency between successive instructions), you would be right. However, mostly this is not the case.
The problem is that you are simplifying the computer to CPU only, not taking in account other core stages or even peripherals. For example if you have one job which needs a lot of access to main memory, its processing time can be severely jotted by memory latencies. A task switching has a big overhead, but you can hide it behind these latencies. Actually, you can hide other code (thread) execution behind it. That becomes even more significant when accessing out of core devices.
So while it is true that it needs more instructions to do two tasks in parallel than it would take doing them sequentially, it is *not* true, that doing these tasks in parallel necessarily takes more time! And that's the important fact. And it is even more important for games, which are running in exclusive mode where you can manage threads to utilize most of the processing power of the target machine.
