[Thread Superseded] G-Sync 101 w/Chart (WIP)

[Chief Blur Buster]

I'm curious: Is GSYNC + RTSS cap affected by the sync behaviour you've configured for maximum GSYNC rate (VSYNC ON/OFF)? It would be interesting to see if there are lag differences -- I would guess not because if the RTSS cap prevents the maximum GSYNC rate from being hit, we wouldn't get any lag behaviour from that.

[MT_]

Guess that's the main drawback of a 60hz G-sync monitor.

Addressing the G-Sync ceiling at a 60Hz maximum is definitely trickier than at higher maximum refresh rates, especially at native 1080p with a capable system; you're going to be hitting/going over the ceiling much more often in this scenario, which means a framerate limit is almost always going to be necessary if you want to avoid v-sync behavior and additional input latency.

The lack of in-game fps limiters, and the apparent 1 frame latency RTSS adds, even with G-Sync (still have to confirm this myself), sometimes leaves you with no good option.

That said, I think I should make it clear that G-Sync + v-sync on + RTSS fps limit is still a superior experience when directly compared to standalone double buffer v-sync + fps limit (RTSS or otherwise).

In my original input latency tests, I didn't manage to get 60Hz results in, but native 60Hz scenarios will be included in my next line of upcoming tests.

Tell me about it. Quite a pain in the ass without losing too much smoothness due to lowered fps. Then again the 4K makes up for it (visually) although I suspect that even 4K adds somewhat more latency now too because 4x the amount of rendering has to go through the pipeline. Even if my 1070 GTX is only 30/40% saturated on max clocks. I dont think much else can be done about that except running everyhing on ultra low graphics wise. Don't think any more casual oriented gamer would notice this though.

I'll be set on 55 fps internal cap just to be safe, with V-sync on the experience is still smooth enough! At the end of the day having to adjust from 165hz back to 60hz took me a few days as well.

[dbeard]

I have noticed an oddity and have a curiosity to scratch from it.

If I have gsync enabled with vsync off, it seems like my FPS is basically halved once it starts getting into extreme territory.

Say I'm at a menu and it's burning away at 1000 fps, with gsync enabled that will be 500~. I was kind of expecting it to just flat out disable once it goes out of my refresh range with vsync forced off.

Not that it matters a ton in reality since I usually just cap the FPS, but just wondering if that's expected behavior. Fresh W10 install, 1080, latest drivers, Dell SG27.

[Chief Blur Buster]

Interesting behavior!

I'm not sure about this, this is somethng that merits investigation! I prefer in-game frame rate capping too, as it leads to more predictable visual/latency behaviours. Of all the options -- GSYNC/VSYNC ON (lag sudden increase) versus GSYNC/VSYNC OFF (sudden tearing) versus GSYNC/Framecapping (consistent visuals, consistent low lag) -- the latter option is still generally the preferred method ever since our discoveries a few years ago in GSYNC Preview #2.

[MT_]

So 55fps in game cap with V-sync off is still showing some occasional tears on the bottom of screen and once in a while somewhere in the middle.

Is that a guideline that I should probably even go lower if I want V-sync on with no latency? I feel there are random times where input latency differ still, I guess this happens when the tear would occur in the middle of the screen when Vsync is off.

The in-game cap just fluctuates a lot, especially when assets are being loaded or the engine stalls.
Seems 54~fps is much safer.

[Chief Blur Buster]

I think a lot of this depends on the accuracy of the game-cap and other factors. Given sufficient capping, in theory there should be no latency difference between VSYNC ON and VSYNC OFF if your GSYNC maximum rate is never hit (thanks to frame capping).

A 54fps cap probably will solve the tearline issues in your particular case.

Thinking about this mathematically... Tearlines at the bottom of the screen versus VSYNC ON, is probably only approximately a ~1ms difference in latency for that particular refresh cycle that the tearline shows up in, and tearlnes in center of screen versus VSYNC ON, is only a 8ms difference (1/2 of 1/60sec) in latency for only that particular refresh cycle that the tearline would otherwise show up in. The full huge sudden lag increase (sometimes cascading (1/60sec)+(1/60sec) if the frames suddenly starts lining up in a queue waiting to be delivered to the display. For a 60Hz refresh cycle scanning out top-to-bottom, 1ms translates to 1/(1/60th) of screen height == approximately ~1/16th of screen height. Tearline position mathematically indicates latency differences (since LCDs scan top-to-bottom in high speed video, and bottom edge of screen is the "finish line" essentially for an LCD refresh cycle -- as seen in high speed videos of an LCD) If the tearline is consistently near the bottom edge of the screen anyway, ideally there's generally little latency difference between VSYNC OFF and VSYNC ON technique for a 55fps-capped GSYNC 60Hz. If you're not a competition gamer, 54fps cap versus 55fps cap wouldn't make a noticeable difference in this particular case.

[kurtextrem]

Have you tested what "pre-rendered virtual reality-frames" and "maximal pre-rendered frames" do with G-Sync (+ V-Sync)?

e.g. does the virtual reality thing even do anything if I'm not using a VR application? And there is a lot of discussion as well whether max pre-rendered frames = 1 or 2 is better.

[MT_]

I think a lot of this depends on the accuracy of the game-cap and other factors. Given sufficient capping, in theory there should be no latency difference between VSYNC ON and VSYNC OFF if your GSYNC maximum rate is never hit (thanks to frame capping).

A 54fps cap probably will solve the tearline issues in your particular case.

Thinking about this mathematically... Tearlines at the bottom of the screen versus VSYNC ON, is probably only approximately a ~1ms difference in latency for that particular refresh cycle that the tearline shows up in, and tearlnes in center of screen versus VSYNC ON, is only a 8ms difference (1/2 of 1/60sec) in latency for only that particular refresh cycle that the tearline would otherwise show up in. The full huge sudden lag increase (sometimes cascading (1/60sec)+(1/60sec) if the frames suddenly starts lining up in a queue waiting to be delivered to the display. For a 60Hz refresh cycle scanning out top-to-bottom, 1ms translates to 1/(1/60th) of screen height == approximately ~1/16th of screen height. Tearline position mathematically indicates latency differences (since LCDs scan top-to-bottom in high speed video, and bottom edge of screen is the "finish line" essentially for an LCD refresh cycle -- as seen in high speed videos of an LCD) If the tearline is consistently near the bottom edge of the screen anyway, ideally there's generally little latency difference between VSYNC OFF and VSYNC ON technique for a 55fps-capped GSYNC 60Hz. If you're not a competition gamer, 54fps cap versus 55fps cap wouldn't make a noticeable difference in this particular case.

Yeah Im a competitive gamer I dropped all the way down to 52~ fps now and i still get occasional mid-upper screen tears. This is most likely due to the engine stalling (Which it does, loads)

So I might as well ignore all this and go 58~ where I get a steady bottom tear and dont take the assets / file streaming into account.

Another question If i might ask, how does pre-rendered frames work with G-sync? I assume going higher than 1 adds additional input latency despite everything? It shouldnt be as much I suppose, perhaps it would work things out with the engine microstutters.

[MT_]

Mhh still occasional tears even at 52 fps... i tried 8 prerendered too, no dice :/

Edit: Ok, everything is solved. First of all Windows 10 (Even LTSB) is a piece of crap, it uses less memory, starts up faster (probably by cheating and other insignificant stuff not relative to runtime) but when it comes to stable gameplay and frametimes and other stutter fests its horrible. Back to a clean and highly optimized Windows 7 without forced Compositor. It just feels so much cleaner. I have the idea that drivers for Win 10 are rushed, the system is badly coded and it will probably take 5 years for it to mature as Windows 7 has.

Strangely my ingame cap can now sustain a steady 58.8fps with 0.1 fps fluctuations rather than 2/3 frames, with Vsync off and have a nice tearline at the bottom but thats about it. Not sure If i can explain it but perhaps installing DX9 2010 Redist helped there (even if the game runs in DX11, i wouldn't be surprised if its a borked implementation or uses pieces of DX9 for other stuff.

Even alt+tabbing from full screen and back is faster O.o

[Chief Blur Buster]

Interesting finding!

jorimt will have to test with this too.

Does it improve with Windows 10 Creator's Update, with its new Game Mode?