So I recently wiped and installed Windows 8.1 on my system. Thus, I'm at the stage where I am getting things configured properly. Some things have changed on this system with a new installation, as I had went from Nvidia to AMD GPUs (from SLI to CFX) and some system settings had to change to remain smooth, like disabling HPET and whatnot.
I was mostly using GTA V as a benchmark for performance and frame pacing/stutter. Decided I should test other games to ensure I'm acknowledging that GTA V could be the only issue.
I loaded up a 64p map in BF4 and turned on the performance overlay graph. Mind you, I have framerate capped at 60fps, on a triple monitor 6048x1080p (bezel corrected) resolution, with 2x290X.
With HPET on and UsePlatformClock=True, I saw the CPU at 16.67ms in a solid line, as expected, but the GPU seemed to fluctuate quite a bit and averaged around 23ms on the graph (that's per frame, for those that are unaware).
With HPET on and UsePlatformClock=False, again the CPU remained the same but the GPU dropped closer, averaging around 18ms/frame and with seemingly smoother variations without as much jumping around.
I'm under the assumption that, with 60fps locked (and the frame counter in game continuing to display 60fps even with the graph fluctuating), the GPU should keep to around 16.67ms/frame though obviously some variance might be expected.
I feel both felt smooth, but honestly I wasn't as into the test sessions as I should have been... I was more obsessed with the graph than the actual gameplay. I know I was seeing some mild microstuttering in GTA V once I got things ironed out (which means mild microstuttering appeared smooth compared to what I saw before), but I haven't tested it again with everything I figured out for GTA V along with having the platformclock option disabled.
I truly hope to leave HPET on as that makes OS X in my dual-boot config happy (no HPET = angry, vengeful, suicidal OS X that hates life and doesn't want to continue on), but if I should truly be looking for a nearly locked 16.67ms/frame on the GPU, I'll have to investigate that. It did seem to be ever so slightly worse with HPET off in other testing I did.