Blur Busters Forums

Wooohah wrote:Are frametime spikes just inherently a design flaw of Gsync when it comes to FPS drops or fluctuations in FPS?

Frametime spikes originate solely from the system, not the display. So it's less a design flaw of the display, and more a delivery limitation of the given system running the given game, G-SYNC or no G-SYNC.

Frametime spikes = complete absence of new frames sent from system to display for x amount of time = display can't do a darn thing about it (sans repeating last delivered frame) until delivery of new frames from system resumes = unavoidable stutter.

jorimt wrote:

Wooohah wrote:Are frametime spikes just inherently a design flaw of Gsync when it comes to FPS drops or fluctuations in FPS?

Frametime spikes originate solely from the system, not the display. So it's less a design flaw of the display, and more a delivery limitation of the given system running the given game, G-SYNC or no G-SYNC.

Frametime spikes = complete absence of new frames sent from system to display for x amount of time = display can't do a darn thing about it (sans repeating last delivered frame) until delivery of new frames from system resumes = unavoidable stutter.

Not sure how unavoidable it is. Using gsync OFF and vsync ON does not exhibit this issue. Going from 0FPS to 60FPS with gsync OFF vsync ON produces a perfect, instant transition. With gsync ON vsync ON the transition lasts about 1 second, in which you only get 5FPS or so (so severe stutter). With gsync ON vsync OFF, it's still not instant, but much quicker (transition takes maybe 200ms or so.)

An easy way to test this is running Witcher 3 and opening the inventory screen or the map screen. The games goes to 0FPS for a bit while it opens the new screen, and once it's open, it instantly resumes rendering at full speed. With gsync OFF vsync ON, this is 100% invisible. With gsync ON, you get severe hitching as gsync goes from 0FPS to full speed.

^ Yeah, I wasn't referring to G-SYNC's recovery from frametime spikes, but the general occurrence of frametime spikes themselves, regardless of the syncing method in use.

As for standalone V-SYNC vs. G-SYNC + V-SYNC recovery time from frametime spikes (or 0 to x FPS transition more specifically), I haven't tested it enough to speak on it myself.

jorimt wrote:^ Yeah, I wasn't referring to G-SYNC's recovery from frametime spikes, but the general occurrence of frametime spikes themselves, regardless of the syncing method in use.

As for standalone V-SYNC vs. G-SYNC + V-SYNC recovery time from frametime spikes (or 0 to x FPS transition more specifically), I haven't tested it enough to speak on it myself.

If you're looking for a game where this issue is really jarring, it's Lost Horizon. This game has 0FPS loading screens, and once the loading is complete, it does not do any sort of "warm up" or fade-in transition like other games do. It instantly goes from 0FPS loading screen to full FPS cut scene. This is the worst case scenario, and g-sync + vsync ON shows severe hitching for half a second up to a whole second.

For the fun of it, I did briefly test the Witcher 3 gameplay to inventory ("I" button) test you suggested, in a scene that hovered around 100 FPS @144Hz:

1. Standalone V-SYNC = virtually stutterless transition
2. G-SYNC + V-SYNC "On" = approximately 0.5-1 sec perceived transitional stutter (varied instance to instance of course, due to given scanout progression + given occurrence of user input)
3. G-SYNC + V-SYNC "Off" = varied from brief tearing during transition to being nearly indistinguishable from scenario #1

As we already know, scenario #2 behaves such as it does because the frametime compensation mechanism adheres strictly to the VBLANK, while scenario #3 does not.

Whether scenario #2 could be improved by a hypothetical G-SYNC "3.0," I don't know, as if you adhere to the VBLANK, you're at the mercy of the scanout, which is static, and thus can cause fractions of a frame to (at very worst) a single frame of added "recovery" time from spikes with G-SYNC + V-SYNC "On" (perhaps this will be less of an issue with G-SYNC + V-SYNC at above 240Hz refresh rates in the future, as the scanout time further decreases).

Also, it should be noted, there is a difference between "0 to x FPS" (a clean, linear, replicable transition) and frametime spikes (randomized occurrences/spans of complete absence of frames output to the display from the system one or more frames at a time), with the former being the rarity, and the latter causing stutter no matter the syncing method (or lack thereof) in use.