As "big and heavy" as Windows is, it's the Apps that cause most of the problems; Windows Update notwithstanding. Sure you could kill all logging and free up a percentage point or two and maybe a couple degrees of heat. OR efficiency tuning might get you similar results. While those efforts are not unworthy it's not generally where big gains will be realized. There's a significant downside as well to the stripped down approach which makes finding and resolving problems that much more difficult when they happen and they always happen.
I understand there are huge improvements in the new graphics standards but apps have to be written/rewritten to use it. However, devs aren't even using the advantages that are available to them now. Maybe someone can figure out how to optimize all the old and bad code out there and magically make it perform efficiently but I'm not holding my breath.
Windows Update does need to be given a Lap-Band operation and some significant improvements. Why is it that a so called background process kills your performance whenever its running. Very, very, sad. There could be a couple other background process that need some work as well but even when they are out of the way its the Apps that kill.
I worked on a corporate app performance swat team and most of the problems boiled down to poor and uninformed choices by developers. Even when they chose a bad tool for the job such as using VB for a critical Database Application other choices prevented successful conversion (VB was Single threaded so on a 16 CPU server 1 CPU was running at 95+% a couple others at a few percent doing other work and most of the CPUs were idle). The expert development consultants were prototyping a rewrite but only getting 20% improvement due to other bad choices. Yet there was a very simple fix when you looked at the problem correctly and in much less time than they had already spent we achieved a 4x performance improvement cutting run time from 36 hours to 8 and avoided buying new 100k servers which would have only given them a 5% improvement at best. In most cases they just write code with no understanding of the impact it has. Throw crap at the wall and see what sticks, close the door and deliver it. Don't worry, the next system will run it twice as fast and have twice as much ram.