Pretty sure Nvidia Reflex is causing input lag..

[TTT]

I've been testing my setting for Cod in the Cold War bot mode, I'm pretty sure turning Reflex on (Its on by default) in any scenario i.e GPU bound or not it still causes more latency when it is supposed to reduce it.

I've tried upping the render rate to pin my GPU at max usage and tried it on/off, better latency with it off. The same running uncapped but my GPU isn't maxing out its better with it off. Same with Gsync on or off.

That was with NULL completely off also, I tried turning NULL to 'ON' and pinned my GPU at max usage. The latency was still better with Reflex off.

This is with a 1070ti and i5 8400.

[jorimt]

I have an LDAT now, and I own this game, so I could do some quick tests, but they'd have to be very specific to apply to your build, as mine differs quite a bit (see sig), so please list out your target in-game settings, framerate, and refresh rate (highest Hz monitor I currently have is 165) and I'll try to scale. Also let me know if you're using HAGS (I'm not).

In my very brief tests tonight, Reflex was definitely lowering average input lag when my GPU was maxed.

As for LLM, according to official information, Reflex is suppose to override it when in enabled, regardless of what LLM is set to, so it shouldn't have an impact when Reflex is engaged.

Also to you and anyone else reading this, please note that any of my LDAT tests now or in the future will be with G-SYNC enabled only, as photodiode setups are not reliable for testing tearing scenarios, since they only measure one point on the screen at a time, and thus will typically miss the tearline (aka the input lag advantage over synced scenarios) more often than not.

[deama]

I have an LDAT now, and I own this game, so I could do some quick tests, but they'd have to be very specific to apply to your build, as mine differs quite a bit (see sig), so please list out your target in-game settings, framerate, and refresh rate (highest Hz monitor I currently have is 165) and I'll try to scale. Also let me know if you're using HAGS (I'm not).

In my very brief tests tonight, Reflex was definitely lowering average input lag when my GPU was maxed.

As for LLM, according to official information, Reflex is suppose to override it when in enabled, regardless of what LLM is set to, so it shouldn't have an impact when Reflex is engaged.

Also to you and anyone else reading this, please note that any of my LDAT tests now or in the future will be with G-SYNC enabled only, as photodiode setups are not reliable for testing tearing scenarios, since they only measure one point on the screen at a time, and thus will typically miss the tearline (aka the input lag advantage over synced scenarios) more often than not.

Do you have a linux system by chance? A linux vs windows comparison would be neat, especially with an optimised windows vs optimised linux.

[jorimt]

Do you have a linux system by chance? A linux vs windows comparison would be neat, especially with an optimised windows vs optimised linux.

No, just Windows 10.

On my system with G-SYNC (non-GPU-bound), at crosshair level ~150-160 FPS/165Hz offline multiplayer in Cold War (graphics maxed w/ray tracing off), it's as low as 26ms average input lag already (which means top of display it's averaging about ~23ms), and that's the whole button-to-pixel chain (LDAT "click" to USB to PC, system/game engine processing, render queue, render time of frame, monitor processing/GtG, etc) so I'm not sure how much lower it can go at that FPS/Hz anyway.

We'd probably have to start getting into potential frametime performance differences for such OS/optimization comparisons instead, which an LDAT or high speed setup can't capture.

[TTT]

Hi, sorry for late reply I've been messing with other things trying to get to the bottom of it.

I'm opening MSI Afterburner at the same time to use the fan curve and I think its that what is causing it and also causing it to stutter, even though having the frametime graph open when it stutters it isn't showing a frametime spike.

It seems to be just having Afterburner open is messing something up, even if I don't use the fan curve and if I turn off the monitoring its still not right, the screen moves much more fluid as well with it closed. It could be running at 140+ fps and moving it around it looked like it was about 50.

I'll do some more back and forth but I'm guessing my first thought was wrong and Reflex is fine.


Also are you still using Vsync on with Gsync?

Now playing Reflex enabled games forces your framerate to cap way more than -3 below refresh if you have Gsync+Vsync and there doesn't seem to be any way around it even if you wanted Vsync on and cap at 237 instead of the 225 autocap.

[jorimt]

Also are you still using Vsync on with Gsync?

Now playing Reflex enabled games forces your framerate to cap way more than -3 below refresh if you have Gsync+Vsync and there doesn't seem to be any way around it even if you wanted Vsync on and cap at 237 instead of the 225 autocap.

I just tested this in Cold War:

- G-SYNC on + NVCP V-SYNC on + in-game V-SYNC off + Reflex (on or boost) = auto FPS limit
- G-SYNC on + NVCP V-SYNC on + in-game V-SYNC off + Reflex off = no auto FPS limit
- G-SYNC on + NVCP V-SYNC off (Use 3D app) + in-game V-SYNC on + Reflex (on or boost) = auto FPS limit
- G-SYNC on + NVCP V-SYNC off (Use 3D app) + in-game V-SYNC on + Reflex off = no auto FPS limit

[TTT]

Also are you still using Vsync on with Gsync?

Now playing Reflex enabled games forces your framerate to cap way more than -3 below refresh if you have Gsync+Vsync and there doesn't seem to be any way around it even if you wanted Vsync on and cap at 237 instead of the 225 autocap.

I just tested this in Cold War:

- G-SYNC on + NVCP V-SYNC on + in-game V-SYNC off + Reflex (on or boost) = auto FPS limit
- G-SYNC on + NVCP V-SYNC on + in-game V-SYNC off + Reflex off = no auto FPS limit
- G-SYNC on + NVCP V-SYNC off (Use 3D app) + in-game V-SYNC on + Reflex (on or boost) = auto FPS limit
- G-SYNC on + NVCP V-SYNC off (Use 3D app) + in-game V-SYNC on + Reflex off = no auto FPS limit

So basically there isn't a way of fully utilizing the FPS you can get (If you can max your monitor) whilst also fully getting the benefits of Gsync (Having Vsync on with it) if you are playing a Reflex game now.

[jorimt]

So basically there isn't a way of fully utilizing the FPS you can get (If you can max your monitor) whilst also fully getting the benefits of Gsync (Having Vsync on with it) if you are playing a Reflex game now.

The way to look at it is this; if you need Reflex, it means your system is currently GPU-bound, and if your system is GPU-bound, it means whatever framerate your system is reaching is maxing your GPU.

In that case, reaching the -3 average FPS of your Hz is undesirable where render queue input lag is concerned, at which point, to prevent a GPU-bound situation without Reflex, you'd have to set an FPS limit well below your refresh rate anyway. On the flipside, if your system is able to maintain a -3 average FPS of your Hz, and it's not maxing your GPU, you don't need Reflex enabled.

The point is, if your system is GPU-bound, even though Reflex doesn't allow you to eek out every bit of your upper G-SYNC range, it's dynamic FPS limiter allows you to maintain a higher variable framerate while avoiding GPU-limitation, as opposed to a fixed framerate via the potentially much lower FPS limit you'd have to set to achieve the same thing with Reflex off.

[TTT]

So basically there isn't a way of fully utilizing the FPS you can get (If you can max your monitor) whilst also fully getting the benefits of Gsync (Having Vsync on with it) if you are playing a Reflex game now.

The way to look at it is this; if you need Reflex, it means your system is currently GPU-bound, and if your system is GPU-bound, it means whatever framerate your system is reaching is maxing your GPU.

In that case, reaching the -3 average FPS of your Hz is undesirable where render queue input lag is concerned, at which point, to prevent a GPU-bound situation without Reflex, you'd have to set an FPS limit well below your refresh rate anyway. On the flipside, if your system is able to maintain a -3 average FPS of your Hz, and it's not maxing your GPU, you don't need Reflex enabled.

The point is, if your system is GPU-bound, even though Reflex doesn't allow you to eek out every bit of your upper G-SYNC range, it's dynamic FPS limiter allows you to maintain a higher variable framerate while avoiding GPU-limitation, as opposed to a fixed framerate via the potentially much lower FPS limit you'd have to set to achieve the same thing with Reflex off.

True!

Thanks.

[Chief Blur Buster]

Also to you and anyone else reading this, please note that any of my LDAT tests now or in the future will be with G-SYNC enabled only, as photodiode setups are not reliable for testing tearing scenarios, since they only measure one point on the screen at a time, and thus will typically miss the tearline (aka the input lag advantage over synced scenarios) more often than not.

Still, photodiodes don't go completely useless in tearing tests, as long as you run enough samples with enough randomized tearline placements. Then you get your latency range data / latency volatility data / average latency.

What's interesting is average latency of top/bottom edges is identical in VSYNC OFF, something not achievable with other sync technologies (unless combining VSYNC ON + strobed, a global sync technology with a global strobe flash). Unless the panel is using a different scanout velocity on the panel than on the cable (like 60Hz signals on certain 240Hz panels)

So provided testing methodology is developed accordingly, photodiode tests can still have great use for VSYNC OFF tests.