Blur Busters Forums

Hi,

First time posting, go gentle!

I have been tinkering with various settings for quite some time with mixed results. One of the key things I wanted to understand and focus on was how to reduce Nvidia’s render latency, which I view when using the (Alt+R).

I was originally using an Acer Predator 34” curved and always had the nagging feeling that the monitor was limiting what I was able to see and would lose a number of 1on1 FPS encounters. My machine spec was certainly not a problem, so I thought it has to be the screen. I was averaging anywhere between 7-12ms according to the analyser

I then purchased an Asus 32" PG329Q WQHD Fast IPS G-SYNC Compatible 175Hz Monitor, however, I still wasn’t seeing great results with Nvidias render latency, averaging around the 5-7ms time now. I viewed an Nvidia settings video a little earlier and noticed the monitor on the video was sitting around the 1ms time, however this was for the game Valorant which I do not play. The game I do play is Hell Let Loose and some would say that the optimisation for the game is poor, I wouldn’t disagree.

In trying to find the best performance, I have tried G-Sync on/Off, ELMB, V-Sync on off and all manner of settings, it really is a cycle of going around in circles with the many different combination of settings available. I saw a few comments here about Pro players not using any of the G-Sync functionality etc.

I am not bothered about visual artifacts etc, I just want to focus purely on performance for FPS and getting the absolute quickest performance out of my new monitor and my machine. I am tech savvy but certainly no expert when it comes to the level of discussion found here henc emy reaching out for a bit of guidance. I can and will provide any additional information as requested and any help would be genuinely appreciated.

I have a second 32” screen which is connected directly to the MOBO and the main gaming screen direct into GPU.

Machine spec:
Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz (5.00Ghz)
ASUS ROG Strix GeForce RTX 2080 Ti OC Edition 11 GB
CorsairVengeance RGB PRO 32GB- DDR4 3600
Sabrent 1TB Rocket Nvme PCIe 4.0 M.2 2280

Nvidia Drivers 471.68

Monitor Settings:
Overlocking on @ 175Hz
Adaptive Sync On
ELMB Sync On
Shadow Boost Off

Nvidia Settings

Since render latency is the time the GPU needed to render the frame, that means the only way to lower it is to get a faster GPU and/or lower the game's graphics settings (lower resolution, lower shadow quality, lower anti-aliasing, lower ambient occlusion, etc, etc.)

I run the game on relatively lower settings anyway, i was hoping that this can be improved from software settings, as i run a Rog Strix 2080ti and i believe there arent massive gains to be had from the new cards over the card i have already. I was kinda hoping a pointer towards perhaps turning off the likes of G-Sync off and going with forced refresh rate may have been advisable as an alternative way forward.

Few additional images:

Fopa wrote: ↑

20 Aug 2021, 16:35

I was kinda hoping a pointer towards perhaps turning off the likes of G-Sync off and going with forced refresh rate may have been advisable as an alternative way forward.

Going by your info and screens, there's nothing to worry about regarding the render latency you're seeing. Those are very low levels. I also wouldn't rely solely on such readouts. They only show you a small part of the overall latency chain, and can be subject to some interpretation.

To give you an idea, anything between 15-30ms of total latency from mouse/keyboard, through the system, out to the display in your general AAA PC title can be considered very low, and is, by contrast, a far cry from your average (non-retro) console game of the past, which could be anywhere from 80-200ms of total latency.

Anyway, even if you were able to remove all other forms of latency (which is virtually impossible), you'd still be limited by the scanout time of your monitor...

At 175Hz, for instance, each refresh cycle takes 5.7ms to complete. In other words, barring all other forms of input lag, if you're using G-SYNC (which displays one complete frame at a time), your minimum "latency" would be 5.7ms, and this would merely be from the next frame having to complete render and the current frame having to complete scan-in before the next can be displayed.

If, however, you were running with no sync at 175Hz, assuming you could achieve, say, 1000 FPS in the given game (1ms render time per frame), the minimum "latency" (or more specifically, how long it takes your monitor to display new information per refresh cycle), barring all other forms of latency, would still be 5.7ms due to your refresh rate...

The only difference in that case, is you would get micro updates in each of those cycles (nearly six per, instead of the 1 per when synced) due to the amount of tear slices that could fit in each, but the amount of latency reduction (and tear slices) would obviously depend on how high of an average framerate your system could sustain in the given game; 1000 FPS is rarely achievable.

See here for more on that:
https://blurbusters.com/gsync/gsync101- ... ettings/9/

As for pointers, the lowest achievable input lag from a set-and-forget standpoint would always be the lowest possible in-game graphics settings + no sync + an uncapped framerate sustained as far above your display's maximum refresh rate as possible without maxing the GPU (as to prevent a sustained increase of pre-rendered frames in the render queue).

The rest is down to the inherit delay in your particular input devices and display, the achievable framerates and latency levels of your system, the performance of the given engine (many are simply unoptimized and can become internally bottlenecked, regardless of system specs), and, for online games, your ISP/ping/location, and the netcode of the given title.

Thank you for your informative reply, i genuinely appreciate the time and information provided. I guess i need to go and do more testing, thanks again!

Fopa wrote: ↑

21 Aug 2021, 05:49

Thank you for your informative reply, i genuinely appreciate the time and information provided. I guess i need to go and do more testing, thanks again!

No problem.