Blur Busters Forums

Long time lurker here. I play alot of overwatch and valorant and have been running the gsync + vsync with fps capped 3 under refresh rate for some time. It looks incredible and I love the smoothness tear free experience of these setups. I can't seem to quite figure this one issue out so maybe the pro's here can help.

I'm running a 3090 and i9-11700k on low settings with a 240 hz monitor. When I enable vsync in my nvidia drivers, with or without the 237 fps limiter on, my fps automatically caps at 224 fps. I'm assuming vsync is capping it there. So my question is why is it capping at 224 fps, and should I set my limiter to 221 fps because of this?

Also to make things more complicated after a long day of gaming yesterday I turned my FPS limiter off and it locked at 239 fps. First time this has happened and I have no idea why because it always caps at 224 fps with or without the limiter. I'm assuming vsync was still active to lock it at 239 1 fps below my refresh rate with no limiter. The game looked and felt great and I then put the limiter to 237 fps because of the 3 fps below rule. But today when I loaded in I was back at the 224 fps. Any clue why this is happening? Thanks for any input.

hockeyfanyeg97 wrote: ↑

01 Oct 2022, 11:48

When I enable vsync in my nvidia drivers, with or without the 237 fps limiter on, my fps automatically caps at 224 fps.

If you set "low latency mode" to "ultra" in the nvidia panel, you get a driver cap when you use vsync. Change it from "ultra" to "on" to disable this behavior. Or disable vsync if the game you're playing doesn't show tearing with it disabled.

The "ultra" setting really is only helpful for uncapped vsync or when GPU is being maxed out. Also, both Overwatch and Valorant support NVidia Reflex I believe? So the "ultra" setting in the nvidia panel is even less important there.

I actually have low latency mode off in the driver and nvidia reflex + boost on in overwatch. Maybe the in game setting of nvidia reflex is causing the cap? I will experiment with that.

Also in regards to low latency mode in the driver is it best to leave it off if using reflex in game?

Ok just did some experimentation and it was the ingame setting of reflex that was causing the frame cap of 224 fps. I have remove

I tried the low latency mode in the driver without reflex just for the hell of it and

Ultra - caps at 224 fps similar to when reflex is on
On - caps at 239 fps

So my question I guess is if I'm using nvidia reflex and getting 224 fps frame cap with vsync should I then set my frame limiter to 221 fps because its suggested to set it 3 under the frame rate?

hockeyfanyeg97 wrote: ↑

01 Oct 2022, 13:14

So my question I guess is if I'm using nvidia reflex and getting 224 fps frame cap with vsync should I then set my frame limiter to 221 fps because its suggested to set it 3 under the frame rate?

No idea. At this point, maybe vsync off is better. If you don't get tearing, that is.

I am going to experiment with v-sync off. I loved the experience with it on after reading the blur busters article but I guess its worth a shot.

You've been very helpful. Please clarify one more thing for me. Is it best to leave low latency mode off in the driver if using nvidia reflex in game?

hockeyfanyeg97 wrote: ↑

01 Oct 2022, 13:14

Ultra - caps at 224 fps similar to when reflex is on
On - caps at 239 fps

Both LLM Ultra and Reflex auto limit the framerate to ~224 FPS @240Hz with G-SYNC + V-SYNC, where supported. This is expected and intended.

As for LLM "On" capping to 239 FPS, it's not (in fact LLM "On" isn't involved in framerate limiting at all). Instead, it's reverting to standalone V-SYNC behavior if you're using G-SYNC + V-SYNC, which unlike an FPS limiter, throttles the framerate at the max refresh rate (240 Hz in your case), causing extra latency, which is the reason you need a -3 FPS limit in the first place.

hockeyfanyeg97 wrote: ↑

01 Oct 2022, 13:14

So my question I guess is if I'm using nvidia reflex and getting 224 fps frame cap with vsync should I then set my frame limiter to 221 fps because its suggested to set it 3 under the frame rate?

No, the minimum -3 FPS recommended limit is relative to the current maximum physical refresh rate (aka 240Hz), which is already being enforced by Reflex's auto limiter in that case, which unlike LLM "Ultra," is at the engine-level, so it should be as low latency as an in-game limiter.

Basically, you have two options in this case:

1. If you're using G-SYNC and your framerate can be sustained above your current physical refresh rate 99% of the time, leave LLM and Reflex off and set a manual in-game (lowest latency) or external (steadiest frametime) limiter a minimum of 3 frames below the refresh rate to keep G-SYNC active and within its working range.
2. If you're using G-SYNC and your framerate can't be sustained above your current physical refresh rate the majority of time (due to maxed GPU usage), use Reflex, which will 1) prevent the framerate from exceeding the refresh rate when G-SYNC is enabled (I.E. the 224 FPS auto limit you're seeing), and 2) dynamically set a limit a few frames below the currently achievable average framerate when/if it dips below 224 FPS to prevent the GPU usage from maxing, which in turn, will prevent more buffering (lag) at the render-queue-level due to an increase in pre-rendered frames.

I've been seeing a lot of questions about Reflex and G-SYNC lately in both my G-SYNC article comments and the forums, and it's all heavily misunderstood. LLM and Reflex are not miracle workers, have nothing directly to do with G-SYNC, and primarily only apply in GPU-bound scenarios.

hockeyfanyeg97 wrote: ↑

01 Oct 2022, 14:05

Is it best to leave low latency mode off in the driver if using nvidia reflex in game?

Reflex is intended as a replacement to LLM and will override it whether LLM is set to "On" or "Ultra." Reflex is superior to LLM, being at the engine-level instead of driver-level, and should always be used in its place where available.

Thanks for taking the time to reply I have a better understanding of a few of the mechanics between reflex, vsync and framerate capping.


I'm now experimenting with reflex off, vsync off, capped at 237 fps. Because I'm not GPU bound playing an old game(overwatch) with a 3090 is it better to keep reflex off while capped? I know this is the old way of doing it. Just cap 3 below refresh rate. Am I correct in my thinking that because I'm not GPU bound reflex isn't doing much. That's what I take from your post. But I see most pros and streamers recommend to keep reflex on which is part of my confusion.

hockeyfanyeg97 wrote: ↑

01 Oct 2022, 14:45

I'm now experimenting with reflex off, vsync off, capped at 237 fps. Because I'm not GPU bound playing an old game(overwatch) with a 3090 is it better to keep reflex off while capped?

It's not that it's "better" to keep it off, it's that Reflex won't do what it's meant to do whenever the GPU usage isn't maxed.

hockeyfanyeg97 wrote: ↑

01 Oct 2022, 14:45

I know this is the old way of doing it. Just cap 3 below refresh rate.

I'm the one who came up with the "3 frames below the refresh rate." And it's not "3" frames, it's a "minimum of" 3 frames below; 3 frames is a safe starting point in G-SYNC scenarios for the average framerate limiter (be it in-game or external).

Said recommendation only applies to G-SYNC, is relative to the max refresh rate, and is intended to keep the framerate within said max refresh rate, since G-SYNC only functions with framerates inside the refresh rate.

As for why it's -3 FPS instead of -1 FPS with G-SYNC, it's because the frametime performance of the system and limiters aren't perfect and can cause fluctuations that momentarily allow the framerate to exceed the refresh rate if the limit isn't set low enough.

And to be very clear, the -3 FPS "rule" does not apply to non-G-SYNC scenarios, be it standalone V-SYNC or no sync.

In fact with no sync, you typically want the highest framerate above the refresh rate as possible without maxing the GPU for the lowest possible latency on the given display in that scenario.

hockeyfanyeg97 wrote: ↑

01 Oct 2022, 14:45

Am I correct in my thinking that because I'm not GPU bound reflex isn't doing much. That's what I take from your post. But I see most pros and streamers recommend to keep reflex on which is part of my confusion.

Yes, whenever your GPU usage isn't maxed, settings like Reflex and LLM (barring the auto FPS limit they institute with G-SYNC to keep it within its range, which is secondary to their main function) aren't doing anything to reduce latency.

As for "pros and streamers," I doubt many of them know what Reflex actually does, and simply enable it because they read it has less latency, which = better in their minds, the end.

When thinking about Reflex and LLM, take sync settings and framerate limiters out of the equation. Those are secondary to what settings like Reflex and LLM actually attempt to address.

Both attempt to reduce a sustained increase of pre-rendered frames in the render queue, which primarily happens when the GPU usage is 99%+. When the GPU is maxed, it means it's too busy rendering the current frame to take the next/latest frame information from the CPU, at which point the CPU will begin to make frame information for future frames (pre-rendered frames) while it waits for the GPU to be ready again.

These pre-rendered frames increase latency, as by the time they are handed off from the CPU to the GPU, and the GPU renders and displays them, they are a frame or more behind the user's latest inputs if the GPU hadn't been maxed in the same situation.

LLM "On" tries to address this by reducing the amount of pre-rendered frames that can be produced at once in each instance, LLM "Ultra" (from what can be gathered) tries to address this by doing what LLM "On" does, plus waiting to deliver the latest frame "just in time" for when the GPU is ready again, whereas Reflex addresses this by preventing the GPU from being maxed by setting a dynamic FPS limit (at the engine-level) just below the currently achievable average framerate, meaning it prevents extra pre-rendered frames from being created in the first place (hence why it's the best out of the three).

Basically, there's sync latency, which is what G-SYNC addresses, and there's render queue latency, which is what settings like Reflex and LLM address; you can experience either or both at the same time depending on the given scenario and your settings.

Thanks for taking the time for the informative reply it's much appreciated.