NVIDIA G-Sync Pulsar monitor - Asus ROG Strix XG27AQNGV

[mcnabb05]

Does anyone know of any measurements for how much input lag/delay pulsar adds?

Battle(non)sense's preliminary data

Anywhere that I can compare this latency to other displays?

[tsarri]

Does anyone know of any measurements for how much input lag/delay pulsar adds?

Battle(non)sense's preliminary data

Anywhere that I can compare this latency to other displays?

rtings has good data for input latency

[kyube]

Here's a showcase of the available VCP commands for the ASUS XG27AQNGV.
Locked down, of course... akin to other models on the market.

Also, confirmation of Mini-LED Pulsar happening (likely the Acer 32" 5K 165Hz)


Hopefully with a separate ULMB2 setting with adjustable PW and HDMI 2.1 this time
DP2.1 UHBR20 sadly won't happen due to the constraint of using the Mediatek MT9810

[mawi]

That genuinely sounds like a firmware bug then. There's no way they haven't thought of the standard GSYNC+VSYNC+Reflex set-up.
I already knew VRR+Strobing is a meme, but this.... would cement it.
I wouldn't even be surprised if ULMB2 is bugged as well, considering the recent discussion I've had with a user of the Pulsar models.

Yep... I just stumbled over the same behaviour. Since I have the Asus Pulsar monitor I was quite impressed but not as astonished as I expected... mainly playing CS2 with almost locked 324 fps (GSYNC+VSYNC+Reflex as you mentioned).
Until I played quake 2 RTX this weekend... which gave me (with all RTX features on) between 90 and 110 FPS. I was shocked how clear and amazing it looked. Just as I would expect it with a proper strobing... It is like a CRT.
Then I saw the Nvidia frame graph. It was 110 in average... AND 109 at 1% low fps. Constantly. When it came down to 90... it was still 89 or 90 at 1% low. It was sooooo consistant. And clearer than CS2 with 324 fps.

So I started searching and looking for hundreds of solutions for boosting the 1% lows... capping fps via nvcp, undervolting my RTX4070 to avoid thermal throttling (which works fine btw, I'm keeping that!). And it all came down to one... and only one thing. Reflex!
Once I turn this off... boom. 260-290 fps at 1% lows! Unbelievable. That is DOUBLE. And so much clearer on CS2, its night and day. My fps is now automatically capped at 357 (with vsync + gsync on, but reflex off). No manual cap from my side.
And the FPS do not matter anymore. If I throw 5 smokes on practice, fps drops down to 160... But both together, average AND 1% low. And visually? No differnce at all to 300 fps. THAT is what I expected from pulsar.

You should use Nvidia Low Latency mode "on". With this setting the game feels totally instant to me, no difference to reflex but has no influence on 1%low fps. Once I set the nvcp Low Latency mode to "ultra", my 1% low are back down to 130, same as reflex.

To be clear: I was happy with my aiming on my old 165Hz screen. Felt no issues there with reflex. Gsync + Vsync + Reflex might be the best setting for esport gamers. But as for now, its NOT the best setting for pulsar.

And as I feel no difference in latency using Low Latency "on" and now have the perfect clarity. I will keep it this way for now. Maybe future firmware upgrades will align better with Reflex... It should. Both are Nvidia features, this display is advertised as the ultimate esport display, so... Nvidia... lets go

[hamza_tm]

This is incredible, thank you so much for sharing! I’ll test when I get a chance

[kyube]

Frame rate is a misleading metric, which doesn't tell you what you experience in real-time. Latency is what matters.
Low-Latency Mode='On' (older Max-Prerendered Frames=1) is functionally identical to Nv Reflex, but it isn't aware of the internal game pipeline.
Nv Reflex (LLM='Ultra', but integrated on a game-level) overrides the aforementioned toggle in games which support Reflex.
It's use-case is lowering the total Render Queue Depth to a value, which is N+1 of what you've set it to.
To be more specific, lowering the potential 'GPU congestion' (commonly referred to as 'back pressure' by some programmers or 'GPU limited') total system latency penalty of specific games.
For example:
Nv Reflex == LLM='Ultra' → Render Queue Depth=max(1) (counting in programming starts from 0)
LLM='On' → Render_Queue_Depth=max(2) (should be the same as MaxPrerenderedFrames=1)
LLM='Off' → Render_Queue_Depth=max(4) (should be the same as MaxPrerenderedFrames=3)

What does this actually mean in practice?
The concept (RQD) itself is from queueing theory, an application of Little's Theorem or Little's Law
What's being compared (the ratio) is msUntilDisplayed & msBetweenPresents
To be specific — Render Queue Depth (RQD) = msUntilDisplayed ÷ msBetweenPresents
Fun-fact: No where in Nvidia's documentation for FrameView or other parts of it will you see this information, you have to infer it =)
As Quake RTX doesn't implement Reflex, your only option to minimise potential total system latency fluctuations is by using LLM='On' or LLM='Ultra'.

I suggest using FSE in CS2 and evaluating performance again.
Perhaps even trying a different title altogether, such as: Overwatch, Apex, Fortnite, PUBG,...

[mawi]

I suggest using FSE in CS2 and evaluating performance again.
Perhaps even trying a different title altogether, such as: Overwatch, Apex, Fortnite, PUBG,...

Sounds promising... if you could just tell me what FSE stands for?

[kyube]

Fullscreen Exclusive presentation model (also referred to as HW:Legacy Flip)
FSO / eFSE ('Fullscreen Optimizations' / emulated FullScreen; also referred to as HW:Independant Flip) is what most Win10 applications use by default and this is the default of all D3D12 ('DX12') games.
CS2 is D3D11 (DX11), hence why you can use FSE (default on OGL/VK, D3D7-D3D11)

[mawi]

Fullscreen Exclusive presentation model (also referred to as HW:Legacy Flip)

Ok, understood. So you mean the standard "exclusive fullscreen" mode in CS2 instead of borderless windowed.
Hm. That is a different stoy with my RTX 4070, 360 Hz pulsar, DL1.4 and DSC (screen compression).

Using real full screen creates black screens all the time. in CS2 not while gaming but each time when tabbing out. On other games also ingame.
RTX 50 users seem to have better experience with this...

So yes, I will try, but with 240 Hz then.

[kyube]

Ok, understood. So you mean the standard "exclusive fullscreen" mode in CS2 instead of borderless windowed.

The toggle in-game may or may not turn your game into HW:Legacy Flip
You need to use tools such as FrameView or PresentMon to evaluate the current presentation model.

Hm. That is a different story with my RTX 4070, 360 Hz pulsar, DL1.4 and DSC (screen compression).
Using real full screen creates black screens all the time. in CS2 not while gaming but each time when tabbing out. On other games also ingame.
RTX 50 users seem to have better experience with this...
So yes, I will try, but with 240 Hz then.

The elongated alt-tabbing behavior is caused by FSE, not by DSC.
DSC may slightly play into this, but it's mostly due to how FSE works.
I don't see a reason to use 240Hz for what I've mentioned above.