so you think most consumers desktop usage fits what cinebench is testing?
As you make it sound like I am an odd one out.
I Also forgot to mention context switching which is horribly evil.
Basically everytime a task is moved from one cpu core to another, there is a penalty. That penalty can be huge.
On linux and bsd its not too bad, on windows its horrific, because windows will rapidly switch single core tasks between different cores for a reason only microsoft know. This switching adds overhead.
Using FF13 lightning returns as an example, since I did in depth testing on that game, when I parked 2 of my cores, the game ran faster and overall cpu usage was about 150% (core and a half).
With all 4 cores unlocked the game ran a bit slower and overall cpu usage was about 300% (3 cores), basically one and a half cores were tied up by kernel context switching, nothing else. This is not a big problem if the game itself is designed to assign specific tasks to specific cores, what happened is basically as I said earlier the rendering is single threaded and should just be using one core, but the windows cpu scheduler rapidly moves from one core to another, and the task manager will appear to show about 25% usage on each core but also with another 35% or so as kernel usage (context switching), as task manager is only reporting average usage its hiding the spikes.
Playing FF13 LR on a 16 core chip I dread to think what the context switching would be like