Input lag at 60 FPS locked fighting games when used with high refresh rate G-SYNC monitors

[Kaled]

Hi,

I have a question on how to interpret input lag test result such as those done by RTINGS.COM, looking at the Asus PG27AQN's result for example

Asus PG27AQN input lag results

inputlag.jpg (54.19 KiB) Viewed 19331 times

If I'm using the monitor with G-SYNC enabled along with a 60 FPS game like a fighting game, G-SYNC will match the refresh rate to the FPS I'm getting, does that mean that I will get the input lag of 12.2 ms associated with running the monitor at 60Hz or the 2.1 ms input lag as if I'm running the monitor at 360Hz because I'm using G-SYNC?

[Boop]

Also -- "Quick Frame Transport" equivalent is already built into FreeSync/GSYNC. Variable refresh rate displays have have been doing this since 2012. Low frame rates shows large input-lag-reducing benefits on high-Hz variable refresh rate dispays, since those refresh cycles are delivered at full dotclock velocity of maximum Hz, even if you're just doing low frame rates (ala 40fps / 40Hz). The lag-reduction benefits show really clearly in the various Blur Busters GSYNC tests (including GSYNC 101). Quick Frame Transport is simply bringing these lag-reducing to fixed-Hz displays (benefits of faster scanout of low refresh rates).

viewtopic.php?p=32652#p32652

Often overlooked is G-SYNC’s ability to adjust the refresh rate to lower fixed framerates. This can be particularly useful for games hard-locked to 60 FPS, and has potential in emulators to replicate unique signals such as the 60.1Hz of NES games, which would otherwise be impossible to reproduce. And due to the scanout speed increase at 100Hz+ refresh rates, an input lag reduction can be had as well…

https://blurbusters.com/gsync/gsync101- ... ettings/13

[jorimt]

If I'm using the monitor with G-SYNC enabled along with a 60 FPS game like a fighting game, G-SYNC will match the refresh rate to the FPS I'm getting, does that mean that I will get the input lag of 12.2 ms associated with running the monitor at 60Hz or the 2.1 ms input lag as if I'm running the monitor at 360Hz because I'm using G-SYNC?

If you're running 60 FPS G-SYNC at a physical refresh rate of 360Hz, you'll get the input lag/processing latency of that physical refresh rate (aka the 2.1ms shown in the RTINGS review). You'd have to lower the actual physical refresh rate to 60Hz to get the 12.2ms.

G-SYNC controls how many times the scanout cycle repeats per second, whereas the currently set physical refresh rate determines how fast each of those scanout cycles complete.

So while 60 FPS 60Hz G-SYNC repeats the scanout cycle 60 times per second at 16.6ms per, 60 FPS 360Hz G-SYNC repeats the scanout cycle 60 times per second, but at a much faster 2.8ms per, resulting in an overall reduction of latency via faster frame delivery, regardless of framerate.

As @Boop already quoted, max physical refresh rate + lower framerates + G-SYNC effectively = a form of dynamic/auto QFT (quick frame transport) operation.

[Kaled]

If I'm using the monitor with G-SYNC enabled along with a 60 FPS game like a fighting game, G-SYNC will match the refresh rate to the FPS I'm getting, does that mean that I will get the input lag of 12.2 ms associated with running the monitor at 60Hz or the 2.1 ms input lag as if I'm running the monitor at 360Hz because I'm using G-SYNC?

If you're running 60 FPS G-SYNC at a physical refresh rate of 360Hz, you'll get the input lag/processing latency of that physical refresh rate (aka the 2.1ms shown in the RTINGS review). You'd have to lower the actual physical refresh rate to 60Hz to get the 12.2ms.

G-SYNC controls how many times the scanout cycle repeats per second, whereas the currently set physical refresh rate determines how fast each of those scanout cycles complete.

So while 60 FPS 60Hz G-SYNC repeats the scanout cycle 60 times per second at 16.6ms per, 60 FPS 360Hz G-SYNC repeats the scanout cycle 60 times per second, but at a much faster 2.8ms per, resulting in an overall reduction of latency via faster frame delivery, regardless of framerate.

As @Boop already quoted, max physical refresh rate + lower framerates + G-SYNC effectively = a form of dynamic/auto QFT (quick frame transport) operation.

Thanks a lot, I have read your article before and it helped me a lot. I wanted to make sure I understood this correctly because I'm using this info with my purchase decision.

So, could we say that in practice while using G-SYNC with 60 FPS locked games the 360Hz IPS Asus PG27AQN could achieve lower input lag of the 240Hz OLED LG 27GR95QE-B (input lag results below) due to the higher refresh rate of the Asus monitor? (Asus 2.1 ms vs LG 2.9 ms)

[Chief Blur Buster]

Remember 60fps 60Hz (laggy) and 60fps 360Hz (less lag) are different lag.

The "60Hz" lag doesn't apply to 60fps 360Hz.

They may look the same (Good 60fps at 360Hz VRR can look like 60fps 60Hz) but 60fps 360Hz is far less lag.

They two different latency beasts.

[jorimt]

Thanks a lot, I have read your article before and it helped me a lot. I wanted to make sure I understood this correctly because I'm using this info with my purchase decision.

So, could we say that in practice while using G-SYNC with 60 FPS locked games the 360Hz IPS Asus PG27AQN could achieve lower input lag of the 240Hz OLED LG 27GR95QE-B (input lag results below) due to the higher refresh rate of the Asus monitor? (Asus 2.1 ms vs LG 2.9 ms)

RTINGs is testing specifically for monitor processing latency there, which can be separate/additive of system-side, sync, refresh rate, and/or framerate latency.

In other words, a 240Hz monitor can have less processing latency at 240Hz than a 500Hz monitor does at 500Hz and visa-versa, depending on the given monitor's display scaler, firmware, etc.

But yes, if the LG 27GR95QE-B has 2.9ms display processing latency (aka how much the monitor adds) at 240Hz, and the PG27AQN has 2.1ms display processing latency at 360Hz, then the AQN will have lower overall latency in your 60 FPS max Hz G-SYNC scenario in like-for-like scenarios, again, with the added benefit of a faster scanout/frame scan-in time (2.8ms vs 4.2ms); the higher the current physical refresh rate (assuming the display processing latency is the same or better on the higher refresh rate monitor), the lower latency lower-than-refresh-rate framerates will have with G-SYNC; 60 FPS 1000Hz > 60 FPS 500Hz > 60 FPS 360Hz > 60 FPS 240Hz > 60 FPS 120Hz > 60 FPS 60 Hz, and so on.

Also, different reviewers measure display lag differently. For instance, TFTCentral found only 0.6ms of actual display lag on the AQN not counting GtG transitions:
https://tftcentral.co.uk/reviews/asus-r ... g27aqn#Lag

Basically, the AQN is one of the lowest latency monitors currently available, and is reaching CRT-levels latency-wise (~0ms).

That all said, if you're a multi-monitor user and/or need to create custom resolutions, you may want to consider another monitor over the AQN due to the fact that it has to use DSC at 360Hz. See:
viewtopic.php?f=16&t=12076&start=90#p95307
viewtopic.php?f=16&t=12076&start=90#p95316

[Kaled]

Thanks Boop, Chief Blur Buster, and Jorimt that was very informative

I have a question about the multi-monitor setup limitations with DSC but I think it's better to ask on the other thread and keep this thread focused on this topic.

[Traveler]

Also, different reviewers measure display lag differently. For instance, TFTCentral found only 0.6ms of actual display lag on the AQN not counting GtG transitions

What does this mean?!.. we can never reliably know what monitor has the lowest input lag on the market right now unless we buy an external latency measuring tool and test it ourselves?

Basically, the AQN is one of the lowest latency monitors currently available, and is reaching CRT-levels latency-wise (~0ms).

According to whom?

[jorimt]

What does this mean?!.. we can never reliably know what monitor has the lowest input lag on the market right now unless we buy an external latency measuring tool and test it ourselves?

Monitor reviewers are attempting to isolate and measure display latency only:

InputLagChain_display.png (30.85 KiB) Viewed 18755 times

RTINGS includes GtG transitions (aka pixel response times) in their total measurements, whereas TFTCentral separates the GtG from the actual processing latency, the latter of which is usually much lower (0.6ms at max refresh rate for the AQN in this case).

As for testing yourself, any easily accessible/affordable measurement tool you could get would require 1) training yourself to understand what you're looking for and how to interpret the results (you'd be surprised how relative and subjective latency testing can end up being depending on how you go about it), and 2) said tool would probably only be capable of recording the whole button-to-pixel latency, and not just the display latency itself, which will always give you a higher number.

For instance, I have an LDAT from Nvidia (photodiode with a built-in simulated mouse click), and I get as high as 25ms latency results on my AQN in certain scenarios because it's counting more than display latency in its readings.

The important thing to understand when testing general latency with photodiodes or high speed cameras is that how high the total latency number is, is less important than the difference between scenarios and that baseline, the baseline of which is going to be different for virtually every test configuration (given test device/peripheral + monitor + system + game combo, etc).

Measuring a near-instantaneous white flashing test square transition at the top left of the screen is going to have much lower latency than measuring a swinging hammer animation in the middle of the screen in Overwatch, and so forth.

What's important with display latency/processing is how quickly said display allows the latest pixel information in the very top left of the screen to start scan-in after it's been sent by the GPU.

According to whom?

The reviews? Honestly, many of the higher-end VRR-capable gaming monitors over the past few years have already been closer to CRT display latency/processing levels (at their max refresh rates) than not. The AQN is simply one of them.

[Traveler]

Monitor reviewers are attempting to isolate and measure display latency only

That's useful for when we want to assemble the lowest latency gaming rig by choosing the lowest latency peripherals which includes monitors of course.
When we get our hands on lowest latency monitor display, mouse and keyboard, then all it's left is OS optimization and hardware overclock/tuning. After that full end-to-end latency measuring (us included) can be done by ourselves which will tell me what is our specific system latency like.

... whereas TFTCentral separates the GtG from the actual processing latency,...

So in other words, this is the monitor latency result we should be looking for if all we care about is to know the monitor's input lag?
If I can deal with ghosting, overshoot and crosstalk because I'm not motion blur sensitive too much, but I am bothered with desync/latency/delay/input lag ect., I shouldn't care much about monitor's response time (GtG)?

... and 2) said tool would probably only be capable of recording the whole button-to-pixel latency, and not just the display latency itself, which will always give you a higher number.

Like OSRTT/OSLTT?

For instance, I have an LDAT from Nvidia (photodiode with a built-in simulated mouse click), and I get as high as 25ms latency results on my AQN in certain scenarios because it's counting more than display latency in its readings.

Because that's an end-to-end latency measuring tool, yeah.
Also, if you don't mind sharing, how did you get an LDAT?

Measuring a near-instantaneous white flashing test square transition at the top left of the screen is going to have much lower latency than measuring a swinging hammer animation in the middle of the screen in Overwatch, and so forth.

Latency in the middle of the screen where crosshairs are located is what is most important then.

The reviews? Honestly, many of the higher-end VRR-capable gaming monitors over the past few years have already been closer to CRT display latency/processing levels (at their max refresh rates) than not. The AQN is simply one of them.

I understand that.
Though, does this mean that high-end monitor's input lag is basically a non-issue at this point and response time is the speed they actually differ the most between one another?