I'm using Windows 8 Pro x64. I've noticed that Rundll32.exe process is taking a lot of CPU (25%, which is a 100% load of a single core) then I'm not not using computer, specifically then sceensaver is active. And that process is terminated immediately then screensaver goes away (then I move mouse, or press a key on keyboard).

I've found this because I'm using custom process monitoring program, that also allowed me to collect info about process (that is last info awailable from already stopped process):

Path              = C:\WINDOWS\system32\rundll32.exe
User Name         = NT AUTHORITY \ SYSTEM
PID/Parent PID    = 4576 / 800
Started by        = C:\WINDOWS\system32\services.exe
Virtual KB        Curr = 149 872 Peak = 299 416
Working Set KB    Curr = 54 732 Peak = 54 732
Page File KB      Curr = 46 720 Peak = 46 720
System Pool KB    Paged = 208 Nonpaged = 22
Private KB        = 46 720
Handles Count     = 385
Faults Count      = 293 958
Windows           = 0
Reads             = 4 255   Read KB = 56 263
Writes            = 1   Write KB = 0
Other IOs         = 677 511   Other KB = 8 614

From that info it's obvious that process is taking a lot of reads and no writes. Also it's started by Services Control Manager. I've never noticed such behaviour on my Windows 7 prior to upgrade, and wondering if it some sort of a bug which causing , or a some legitimate system scan task?

I'm using a Blank screen screensaver, which is not using any CPU. I've actually verified that scrnsave.scr is also terminated, but does not consume CPU. CPU is consumed specifically by Rundll32.exe which is, obviously, somehow tied with screensaver DOWN events.

Also I've noticed that it start not exactly then ssaver starts, but some time after it (or maybe it have some other time interval of inactivity before it starts), but stops always then ssaver is stopped. And as I discovered it actually terminates after some time, I estimate latest encounter as about 2 minutes of work (but I'm not sure if it always takes this much time).

And (as I discovered just now) it seems it does not start again after finishing it's work. Looks like it is actually some legitimate system stuff is just going on, then system is idle. It makes sense since it takes a lot of CPU, so it won't slow down any user process (somewhat similarly to how ngen.exe is working).