G-Sync and its affection to performance

[Dmitry Vorobyev]

So, i'm on a laptop with G-Sync screen, and when i launch 3dmark, it warns me with this message: "We recommend disabling G-SYNC before you run the benchmark."

Can anyone explain how altering the refresh rate of screen can affect my peformance? Testing with g-sync enabled or disabled doesn't change score at all.

[UL_Jarnis]

Two ways;

The more important way; By default G-Sync can cap your framerate to your screen refresh rate. So if you run a lightweight test that can run faster than, say, 144fps, and you have 144hz and G-Sync on, your score will be lower than it could be since framerate never goes over 144. Yes, you can force G-Sync to operate so there is no cap and so framerate can go over screen refresh rate, but last I checked that is not true by default. When playing games, there is no real benefit for going above screen refresh rate (unless you want to eke out every single millisecond of responsiveness out of a game, think "pro CS:GO players") so it makes sense for gaming, but throws off the result for a benchmark that bases the score on the framerate of the test.

The less important but subtle way: Exact same system, exact same hardware, running a test that does not go anywhere near the refresh rate of the screen (say, Time Spy running around 30-45fps) still exhibits *slight* performance penalty for enabling G-Sync. Our theory is that this is due to additional graphics driver overhead. The effect is small - at most a few percent - but can be noticed especially on DX11 tests. You can usually ignore this if you are just looking for a ballpark figure on the performance, but in all honesty, if you want to make an accurate comparison against another system that does not have G-Sync present, the recommendation is to temporarily disable it when benchmarking to take this small extra overhead out of the equation.

[Ruinous]

Strangely, I got a better result with G-Sync enabled on my first benchmark using Timespy than I did once I disabled it :-
https://www.3dmark.com/spy/2926563
vs
https://www.3dmark.com/spy/2926672

[UL_Jarnis]

May be a side-effect of DX12. I must admit when we added this warning, we were mostly worried about effects we saw on DX11. Since DX12 is way more CPU-efficient, it is completely plausible that the effect under it is "none". The difference off about 200 points between your scores is well within the margin of error, which is 3% for tests that are appropriate for the hardware being tested, so in your case about 400-450 points.

[coth]

I have that message, but I don't have G-Sync and FreeSync is off.

[UL_Jarnis]

Internal driver setting for it is enabled. You need to reset that. Either get nvidia inspector utility and use it to disable it, or reinstall NVIDIA drivers using custom -> perform a clean install that resets all settings.