G-Sync feels way worse than plain V-Sync in CPU-heavy scenarios

[Finnen]

I've been running into a VERY weird issue with G-Sync in certain games - mainly Cyberpunk 2077 and The Last of Us Part I, and it's been bugging me for a while. Decided to come here, since I believe this is the only place I can get an actual, educated answer. And I really want to dig into this deeply in a technical way and understand what's going on under the hood here.

My setup is classic G-Sync configuration made with BlurBusters's guide:

• G-Sync: ON
  V-Sync: ON in NVCP
  Low Latency Mode: On
  FPS capped via RTSS (usually 138 for high FPS, or 60 for CPU-heavy games like CP2077)

• Monitor: AOC 27G2U - FreeSync/G-Sync Compatible, connected with DisplayPort
  PC: Ryzen 5800X3D, 32GB RAM DDR4 2666, RTX 3080, MSI B450M-PRO-VDH-MAX, PSU SEASONIC G12 850W

This setup has worked fine for most games. But in Cyberpunk, when I cap the FPS to 60 (because the game is CPU-heavy in certain areas, especially with crowds, so I decided to aim for stable 60 FPS across the board), I get very weird microstutter or jitter when G-Sync is on.

It’s most noticeable in areas where the CPU usage goes high, and I slowly pan the camera with a controller - looks like subtle "frame skipping" or jittering, but the RTSS frame graph shows a perfectly flat 60 FPS line. It's very similar to traditional VSync ON stutter when you can't hit the refresh rate. Almost as if... G-Sync was disengaging for milliseconds? I've noticed this issue in The Last of Us Part 1 as well, but less prominent, and affected a small amount of areas where CPU usage went up.

If I unlock the framerate instead of capping it, the issue changes a bit - it's less of a typical stutter and more like a weird "fast/slow/fast" camera movement effect. Still super visible when CPU is bottlenecking the GPU and GPU usage drops. And if the CPU hit is hard, then it can also produce the same type of jitter as with FPS cap.

This happens regardless of how many frames I can actually output beyond that FPS cap. Even if I adjust the settings to have around 120 FPS, it still happens when the area is CPU heavy and it drops to 70-90. Still happens wether FPS cap is on or off in that scenario. It's actually very simple - CPU usage becomes high, G-Sync seems to go crazy?

Reducing the CPU usage, for example by lowering crowd density or turning "HDD Mode" in Cyberpunk helps, but doesn't solve the issue entirely, just reduces the impact of it, because we're simply changing the settings in a way that put less strain on the CPU. Which leads me to believe that there is some correlation between high CPU usage and G-Sync.

But here's the weird part: if I disable G-Sync entirely, switch monitor to fixed 60Hz, and turn on regular V-Sync, it's perfectly smooth, even in these areas that are extremely CPU-bound (as long as I keep above 60 FPS that is, obviously, otherwise V-Sync side effects kick in).

It all leads me to feel that G-Sync is the problem when the games get CPU-bound.

To summarize the weird behavior in CPU-heavy scenes:

• G-Sync ON + V-Sync ON + Unlocked FPS = smoother than capped, but has "slow/fast" pacing feel

• G-Sync ON + V-Sync ON + 60 cap (RTSS) = jittery/stuttery feel, especially during camera pans

• G-Sync OFF + V-Sync ON + 60Hz = buttery smooth, even in heavy scenes, better than above

• G-Sync OFF + V-Sync ON + 60Hz + 60 cap = pretty much the same smoothness as without the cap, the cap helps the input lag though

I tried everyting in the past couple of weeks. Different NVCP settings, turning off Low Latency mode, adjust Reflex settings in the game, trying to inject different framelockers or settings with SpecialK. Nothing helps. The only thing that helps is turning G-Sync OFF and playing the game with traditional V-Sync.

So now I’m wondering:

• Is G-Sync just super sensitive to CPU spikes or erratic frame delivery?

• Is it breaking/disengaging when the GPU isn't fully fed by the CPU?

• Am I just really sensitive to this, or is this expected behavior with G-Sync Compatible monitors?

• Would a real G-Sync monitor with a hardware module help with this kind of thing?

Would love to hear if anyone else noticed this or has a better fix than just disabling G-Sync entirely.

[jorimt]

It all leads me to feel that G-Sync is the problem when the games get CPU-bound.

To summarize the weird behavior in CPU-heavy scenes:

G-Sync ON + V-Sync ON + Unlocked FPS = smoother than capped, but has "slow/fast" pacing feel

G-Sync ON + V-Sync ON + 60 cap (RTSS) = jittery/stuttery feel, especially during camera pans

G-Sync OFF + V-Sync ON + 60Hz = buttery smooth, even in heavy scenes, better than above

G-Sync OFF + V-Sync ON + 60Hz + 60 cap = pretty much the same smoothness as without the cap, the cap helps the input lag though

I can't find the AOC 27G2U on the official G-SYNC Gaming Monitor list, or any mention of it being an officially G-SYNC compatible model from any other source, so I can't rule out a VRR limitation in your case, but...

You're missing a G-SYNC off + V-SYNC off scenario.

G-SYNC + V-SYNC less equals standalone V-SYNC in capped framerate (within the refresh rate) scenarios, and more equals standalone V-SYNC off + tearline steered offscreen.

I.E. VRR is going to reflect any and all frametime variances in say a 60 FPS 144Hz scenario that standalone V-SYNC at 60 FPS/uncapped 60Hz won't, even if your RTSS frametime graph is flat in some cases. VRR "jitter" is system jitter; it essentially becomes a mirror of system performance, similar to no sync.

[Finnen]

Thank you for the answer. I did some tests with V-Sync Off.

G-Sync Off + V-Sync Off (144Hz) = well, screen tearing (though not nearly as bad as back in the day on a 60Hz screen) and the whole feeling is very "shaky" all over the place (always hated V-Sync Off, can't stand it), gets worse around 70-75 fps range, gets more tolerable higher

G-Sync On + V-Sync Off (144Hz) = similar to G-Sync ON + V-Sync ON, except that the CPU intensive areas feel a bit... as with V-Sync Off?

Now that I think of it, playing with G-Sync ON and hitting these CPU-intensive areas seem to induce the side effects of current V-Sync situation.

If I have G-Sync ON + V-Sync ON, then during these CPU intensive areas I will get this jitter similar to V-Sync stutter when you don't hit the refresh rate, just not as bad. And if I have G-Sync ON + V-Sync Off, then these CPU intensive areas will produce this uneven, shaky feeling of V-Sync OFF with MAYBE some tiny, barely visible, occasional, individual screen tearing for very brief moments (milliseconds)? I can't say for sure, but definitely these CPU intensive areas feel different compared to G-Sync ON + V-Sync ON.

I play with G-Sync indicator on, and it doesn't seem to blink or disappear, but that's probably not a solid proof of G-Sync working consistently. I did, however, enable the option to show frame rate in my monitor's OSD (which obviously doesn't show the frame rate but rather current refresh rate of the monitor) and with G-Sync enabled it properly displays the changes in the refresh rate. When the scene has stable frame rate and no issues, the reported refresh rate is pretty much in line with the game's frame rate (with some small variations up and down for couple of Hz - even with the FPS locked at 60). When the scene has these CPU bottlenecks, the reported refresh rate in monitor's OSD becomes much more unstable, going up and down with higher differences and for tiny, brief moments it actually shows "144" and immediately goes back to the values around the actual frame rate.

Maybe your trope is right, maybe this is the monitor's issue of wonky support of VRR? Could it be that G-Sync might actually disengage for milliseconds when the game is CPU bound and frame pacing is becoming unpredictable, uneven? Now that I think of it, I used to have a problem with this monitor back on Windows 10 when I tried to lock a game at 30 FPS. Instead of LFC'ing it up to 60Hz, it would constantly report 50-75-50-75-50-75Hz. Sometimes turning the monitor off and on helped, otherwise the LFC would just not work. That issue was fixed after I went to Windows 11, though - now locking the game at 30 FPS properly reports refresh rate doubled by LFC.

But maybe indeed this monitor doesn't have the full support for all situations, and this CPU-bound situation is too much to handle for this specific VRR implementation? I always assumed it was G-Sync Compatible, because it's mentioned in the official website and marked as such: https://aoc.com/us/gaming/products/monitors/27g2u-bk

But that's just the manufacturer website...

EDIT

Yep, the more I read about it, the more probable it seems that it's the wonky support of G-Sync in this monitor series. Mine is AOC 27G2, but I found similar issue with refresh rate spiking to the max on 24G2, and that's often when the worst visual stutters occur: https://linustechtips.com/topic/1271223 ... ty-matter/

I guess it's time for an upgrade.

[jorimt]

When the scene has these CPU bottlenecks, the reported refresh rate in monitor's OSD becomes much more unstable, going up and down with higher differences and for tiny, brief moments it actually shows "144" and immediately goes back to the values around the actual frame rate.

Monitor-side refresh rate meters are notoriously approximated and unreliable (I have warned of this multiple times in the forums and elsewhere), and should not be used to determine if VRR is working as intended.

Whether it's reflecting the actual limitation of your particular monitor's VRR/LFC capabilities is the question.

Maybe your trope is right, maybe this is the monitor's issue of wonky support of VRR? Could it be that G-Sync might actually disengage for milliseconds when the game is CPU bound and frame pacing is becoming unpredictable, uneven?

It entirely depends on how your monitor does or doesn't handle LFC.

That said, there are no guarantees the behavior you're experiencing will resolve, even with a G-SYNC monitor containing a module; again, VRR + V-SYNC will reflect any and all system-side frametime variances, just without tearing.

You'd have to try another monitor that is either officially G-SYNC-Compatible or has a module to rule out whether your monitor model is responsible.

[Chief Blur Buster]

I've been running into a VERY weird issue with G-Sync in certain games - mainly Cyberpunk 2077 and The Last of Us Part I, and it's been bugging me for a while. Decided to come here, since I believe this is the only place I can get an actual, educated answer. And I really want to dig into this deeply in a technical way and understand what's going on under the hood here.

Jorim's reply is great, but I wanted to add a reply for readers who are prospective GSYNC purchasers who haven't bought yet:

The lowest latency GSYNC is when the entire framerate range of a game breathes organically inside the VRR range of the display. Which is why a 480Hz GSYNC display feels much better for 100-300fps games, than a 240Hz GSYNC display.

Framerate range enter/exiting VRR range are the problem that creates compromises.

Range enter/exit problems create latency yo-yo effects (things feel suddenly laggy during framerate surges = bad). You want consistency, and consistency during VRR requires strategy during purchasing a monitor:

(A) Range top problems: Framerates higher than Hz = you gotta cap = while less lag than not capping, it is still more lag than having higher-Hz VRR. Therefore, if you are upgrading or have not bought yet, please consider a higher max Hz, to make sure GSYNC doesn't feel worse than VSYNC.

(B) Range bottom problems: Framerates lower than Hz = you have LFC stutter/lag problems = can be problematic with some games like Cyberpunk 2077 = try a higher LFC minimum (Custom Resolution Utility) such as 65fps to avoid the latency penalty of LFC thrashing on/off. If you have a very high max Hz, your LFC penalty is tiny (e.g. [0..1]/480sec penalty for 480Hz LFC at 29fps, instead of [0..1]/144sec penalty for 144Hz LFC at 29fps, whenever your 30fps goes to 29fps on a 30Hz-minimum VRR display)

Therefore, get a higher Max Hz + bigger VRR range than framerate range if you buy a VRR monitor
The VRR nirvana occurs when your refresh rate range is permanently bigger than your frame rate range. Yes, even those pesky 0.1% tops and 0.1% bottoms. And if not possble, get the best. Ginormous 10:1 VRR ranges (48..480Hz) solve a lot of 100fps VRR problems, so 480Hz makes 100fps VRR feel fantastic. Heck, even mere 5:1 VRR range feels like a big upgrade (48..240Hz), compared to the crappy 2.5:1 (48..144Hz) which is just atrocious nowadays. The GSYNC:VSYNC penalty seems to virtually disappear at about 10:1 max:min Hz, and everybody's sensitivity is different, but the higher your MaxHz is, the less likely you're able to tell part VSYNC-only and GSYNC/VRR operation.

TL;DR: Always buy more Hz in your next GSYNC monitor than you think you ever need

Now that said, Cyberpunk typically don't reach these framerates, so you can try a bunch of tweaks. However, even at 100fps, playing 100fps Cyberpunk at 480Hz is about 5ms less laggy than 100fps Cyberpunk at 144Hz. (1/144)-(1/480) = (6.94ms)-(2.09ms) = 4.85ms less laggy to play 100fps games at 480Hz than at 144Hz (video cable latency difference; ignoring display processing differences)

GSYNC chip defnitely helps (especially for LCD-based VRR), but if you do OLED, then the GSYNC chip is less important. There's other compromises but the pros may very well outweigh the compromises.