Frame times dont match with frame rate, causing stuttering

[Martinengo]

Can you guys please explain me about VRR and what should I do? My monitor is PG279QZ 165hz G Sync, I have G sync enable most of the time.

[jorimt]

^ VRR (variable refresh rate) is another term for G-SYNC (and FreeSync).

Optimal settings for 100% tear-free VRR at 165Hz is G-SYNC + NVCP V-SYNC "On" + in-game V-SYNC "Off" + 162 in-game OR RTSS FPS limit. Change "NVCP V-SYNC" to "Off" if you want the system to recover from frametime spikes more quickly, though you will see the occasional tear with that.

Also, have you tried disabling fullscreen optimizations for the Fortnite exe? The effects of the setting is very game-specific (sometimes it doesn't help anything), but it's worth trying if you haven't.

And I'm sorry to say (while I'm not ruling something like a motherboard problem out), there is a possibility that there might not end up being a miracle cure for what you're experiencing; it may ultimately turn out to be that you're just extremely sensitive to frametime variances and spikes, especially if even properly configured G-SYNC isn't already at least reducing the perception of the variances for you.

Do you happen to have a recent Afterburner framerate/frametime capture from a Fortnite session to share?

I ask, because you may mistakingly be assinging multiple separate, and/or game-specific issues to everything you're experiencing system-wide, which makes it hard to narrow anything down unless we take it case-by-case, and first determine if the given issue is game-specific, system-specific, a mix of both, or simply a high sensitivity/intolerance to the sometimes unavoidable (frametime spikes).

[Martinengo]

I will do it tonight. I will do three sessions. The thing is this morning I did a session with everything minimum and it dididnt look like your screenshot, it looked similar to your sessions with fortnite with "regular settings". I even capped it at 120.

Sometimes (kind of uncommon), I have frame time spikes that are super noticeable (I would say more than 40ms), and I check the power % and it drops from 70 to 30 suddenly. This doesnt happen often, but these are the most noticeable frame time spikes.

I am aware Im really sensitive to these kind of stuff. I can tell easily the difference within 120 fps vs 165 fps. Today I played Hollow Knight in the morning, and for some weird reason the fps was capped to 164-165, as if v sync was actived. I checked the in game settings, and nvidia settings, and I believe I had to restart so the frames would go up to 400, and I think I could notice the difference between 165fps and 400ms (input wise I think?), Im not sure if that how it works. I know if my monitor is 165hz I wont perceive more than 165fps, but frametimes are lower the more fps I have, so I guess that could translate to some difference in how it feels?, but I dont know, I dont see to many people complaining about frame time spikes.

Anyway, Im not sure if you made that blurbuster article ablut g sync on + v sync on (nvcp), but Ive tried that with not significant difference compared to g sync on + v sync off. So I usually run g sync on + v sync off.

I recently started using g sync on window mode as well, not for any particular reason (sometimes I share my gameplay to friends through discord and that disables g sync on full screen for some weird reason)

And as you mention, gpu usage and power variance in Fortnite is normal in Fortnite, but is so weird to me that at some point Im 300fps and I walk around a little bit, or get close to enemies, gpu usage and power would drop and suddenly Im at 125fps, I guess this is optimization from Fortnite.

Tonight I will post those screenshots so you can give me your opinion. Thanks again.

[jorimt]

Yes, I'm the G-SYNC 101 author.

And to be clear, G-SYNC only works within the refresh rate, so, for instance, if you're not using a 162 FPS limit with G-SYNC + V-SYNC "Off," whenever the framerate exceeds your refresh rate, G-SYNC disables and you get plain old standalone V-SYNC OFF until your framerate drops back within the refresh rate.

Also, with standalone V-SYNC OFF (G-SYNC and V-SYNC disabled), there is actually a discernable difference between 165Hz at 165 FPS AND 165Hz at, say, 300 FPS. I explain that here:
https://www.blurbusters.com/gsync/gsync ... ettings/9/

Finally, before we go any further, I should have asked you to do this a long time ago. Go here and run the test exe: https://www.userbenchmark.com.

The results will tell us whether any of your system components are underperforming when directly compared to other users with the same components.

Oh, and are you capping FPS with RTSS or the in-game limiter while testing Fortnite at the lowest settings? Because you want to use the RTSS limiter in this case for best frametime performance. And again, be sure to test Fortnite with fullscreen optimizations disabled for the game's exe.

[Martinengo]

Yes, I'm the G-SYNC 101 author.

And to be clear, G-SYNC only works within the refresh rate, so, for instance, if you're not using a 162 FPS limit with G-SYNC + V-SYNC "Off," whenever the framerate exceeds your refresh rate, G-SYNC disables and you get plain old standalone V-SYNC OFF until your framerate drops back within the refresh rate.

Also, with standalone V-SYNC OFF (G-SYNC and V-SYNC disabled), there is actually a discernable difference between 165Hz at 165 FPS AND 165Hz at, say, 300 FPS. I explain that here:
https://www.blurbusters.com/gsync/gsync ... ettings/9/

Finally, before we go any further, I should have asked you to do this a long time ago. Go here and run the test exe: https://www.userbenchmark.com.

The results will tell us whether any of your system components are underperforming when directly compared to other users with the same components.

Oh, and are you capping FPS with RTSS or the in-game limiter while testing Fortnite at the lowest settings? Because you want to use the RTSS limiter in this case for best frametime performance. And again, be sure to test Fortnite with fullscreen optimizations disabled for the game's exe.

I've tried capping with both, rtss and the in game frame limiter, to 162, 120, even 100 sometimes.

https://www.userbenchmark.com/UserRun/16109950 that's my userbenchmark run, seems normal. The CPU is overclocked (I overclock it yesterday to see if that could help, not necessarily did, but just to give it a try)

Everything seems fine according to that benchmark, so at this point my main concerns are how the game performed on my pc compared to yours, my latencymon latency, I would say some inconsistency in general, etc.

[jorimt]

[...] that's my userbenchmark run, seems normal.

It certainly does. So it pretty much rules that specific aspect out. And while I'm not suggesting there won't end up being anything wrong with your motherboard, those results definitely put a damper on that possibility.

Here is my benchmark, and keep in mind it is comparing my system to users with systems using MY components, so it's all relative here:
https://www.userbenchmark.com/UserRun/16111023

Grass is always greener and all that, but comparing our numbers here, (while it doesn't always translate to better game performance) your CPU is certainly stronger than mine (especially since I don't currently overclock and yours has more physical cores), so there's no issue there.

Your GPU, the 2080, is known to be (other than ray tracing) about on par performance-wise with a 1080 Ti, but barring differences in factory or manual overclocks, it does look it may be running a little more weakly than it should (and when I say a little, I mean very little, and I could be wrong here). Just to be safe, have you made sure in your selected Windows power plan that under "Change advanced power settings" in the "PCI Express" section the "Link State Power Management" option is set to "Off?"

I'll also point out your 2080 is still benching 10% better than the average 2080 there (that's pretty good for out of over 45,000 other 2080s tested on that site).

Either way, it could just be margin of error between benchmarks (the drift can be pretty large), or OC/brand differences between our cards (my factory OC is one of the higher available for my model, and my GPU is one of the better aftermarket 1080 Ti models). Or, if there is anything to what I am saying (short of the DPC latency spikes I saw in a couple of your captures, I'm having a hard time finding anything wrong with your system), it could even be the way the benchmark pushes the GPU depending on if you have the NVCP setting on "Optimal performance" or "Prefer maximum performance."

Moving onto your SSD, there's nothing surprising there; it looks like it is running better than most (of your model), and my 960 Evo is just stupid fast compared to non-M.2 drives. That, and it has no bearing on our situation, since I haven't used the 960 Evo once in any of the tests I've shared (most of my games, including the test games, are on my 5TB WD HDD, which is much slower than your 860).

Finally our DRAM looks pretty darn even, so nothing to see there.

Basically, there's nothing worrying in your benchmark. Everything looks above average.

I'll await your Fortnite Afterburner graphs (and, by the way, all of my Fortnite tests have been Battle Royal, Solo mode)...

EDIT: I played through another (got near the end this time and placed 4th in Solo), 120 in-game FPS limit, all max (Epic) settings. All I can say is that the performance in this game simply appears to be wildly random from session to session...



Just saying, none of my sessions (sans different match lengths) even look close to the same thus far, even those with the same settings (previously shared graphs below for comparison)...

[jorimt]

For the sake of thoroughness here, as I mentioned earlier, I'm now going to post the most relevant captures of my system configuration:

Startup entries (all disabled):



User-specific processes on desktop at idle directly after startup (I obviously run Steam [or insert-game-client-name-here] when playing games from there, and sometimes have Afterburner and/or RTSS on, but that's about it when gaming):



All "Running" "Services":



Scheduled Tasks:



Pagefile:



Windows power plan:





Windows "Settings > Gaming":







Nvidia Control Panel:







MSI Afterburner monitoring settings:





Mouse settings (I run my Logitech G Pro Wireless at a relatively low 600 DPI w/1000Hz polling rate):

[Martinengo]

This is in my house, using a wi fi adapter. Im starting to think my theory regarding the network and ethernet adapter of my motherboard having a problem. Today has been really smooth compared to other times, like yes, I do feel those 18-23ms frame time spikes, but not as bad as other days. I'll keep testing, so far I'm happy. I've managed to win 2 out of 8 matches, with 13 kills one, 11 kills the other one, and I've had good kda today. The FPS seems stable compared to the other house and/or using the network adapters from the motherboard.

That was all settings low, view distance max, anti aliasing max. I'll keep posting more updates, and I'll keep testing the motherboard and everything network related, and if I prove the motherboard being the problem, I will replace it.

Once again, thank you for everything.

[ko4]

I'm having this exact problem that you're having

I don't know if im hypersensitive to frame time spikes or if something is wrong with my pc.
I've done quite a lot of troubleshooting. Multiple windows installs and different driver combinations. Basically everything on the software side. I will keep following this thread and hopefully we will all come to a conclusion.

btw, does the afterburner frametime graph show info for all frames, or just those that occurred during the refresh period?

[jorimt]

Seeing as we've established here that Fortnite in particular appears to have a very random performance metric from session to session, regardless of configuration, that looks expected/pretty good to me.

And as the greater internet community has and continues to discuss, Fornite isn't the best or most consistent performer. Hopefully it will improve in time.

I'll keep posting more updates, and I'll keep testing the motherboard and everything network related, and if I prove the motherboard being the problem, I will replace it.

Once again, thank you for everything.

Sure thing, sounds good.

Oh, and a bit late, but I found another source that delves into frametime performance capture and the limitations of programs such as Afterburner and FRAPS in this respect (a few years old, but still relevant, as nothing about this stuff has changed since):
https://www.guru3d.com/articles_pages/f ... iew,1.html

I'm having this exact problem that you're having

I don't know if im hypersensitive to frame time spikes or if something is wrong with my pc.
I've done quite a lot of troubleshooting. Multiple windows installs and different driver combinations. Basically everything on the software side. I will keep following this thread and hopefully we will all come to a conclusion.

Regarding Fortnite, definitely game-specific where its frametime performance is concerned. Regarding frametime spikes in general? They can absolutely be reduced in some instances, but are ultimately an unavoidable occurrence in modern gaming.

btw, does the afterburner frametime graph show info for all frames, or just those that occurred during the refresh period?

Good question. Not 100% sure, but I do know Afterburner (along with others, like FRAPS) is approximate, at best, even at lower polling periods (which can also potentially introduce performance issues). You'd need something like FCAT to get more accurate frametime readings:
https://www.guru3d.com/articles_pages/f ... iew,1.html