NVIDIA GeForce 441.87 WHQL: Download & Discussion

How does this frame limiter interact with the various forms of V-sync?

For example, I'm running GTA V with nVidia's Adaptive V-sync because I saw posts saying that was the way to go, and it seems nice and smooth that way, so should I even be using the frame limiter? I am using a 60Hz monitor.

I've tried other forms of frame limiters long ago and recall lots of arguments about the benefits of setting the frame limit to one frame below the native Hz of the monitor but my knowledge is completely outdated now.

 

Guys how can I "import and export Freestyle/Ansel filters" ?

 

Kinda curious myself, most info about NULL, framelimiters and things always talk about freesync/gsync first and high refresh rate monitors.
Meanwhile I'm just here with a plain old 60hz display and CPU starved in most games thinking "do I need NULL? What do I set it to in case of CPU bottleneck and how will it affect performance in this case? What number to use for an FPS limit? Do I need Vsync on top of that limit?" etc etc

 

What, using RTSS was too complicated for you?

 

Well no, but it's obviously better to have it part of the driver package rather than needing external software.

 

Unless the external software works better. Which it still might.

 

Max framerate setting does not work in CSGO no matter on what you set it to, anyone else has this problem? For me only RTSS settings work, on 144hz g-sync compatible.

 

@windrunnerxj From what I've read, if you are CPU limited, you should not enable the Low-Latency modes because the MPRF/Flipqueue is there among other things to smooth out CPU bound stutter and help the CPU consistently supply the GPU with stuff to do in a way that it can "catch up" when it's a bit behind.

Meanwhile, if one is playing at 4K and are almost always GPU bound, then lowering the size of the flip queue might be a fine way to go.

@CaptaPraelium did a very in depth write-up on this on the BFV subreddit sometime ago (Future Frame rendering) so it's a good read I thought.

On the other hand, supposing that you're only ever playing games with V-Sync engaged and you balance your settings so that you never dip below that target, you may be able to then get away with a lower flip queue size provided you still have enough overhead.

So, lowest input lag V-Sync in achieved as I understand it by doing the following:

1) Properly Limiting FPS inline with the BlurBusters guide. An in-engine limiter at 60 can actually be lower than RTSS at 59.98 or some such as I understand it, but some in-engine limiters are not better than RTSS so it just depends (BFV and OW have proper in-engine limiters though it seems based on BattleNonsense tests).

2) Use double, not triple buffer V-Sync -- less buffers = less lag -- this was also tested by BattleNonsense at some point iirc and the expected one frame of additional lag was present with triple buffering. Even moreso at 30 fps, triple buffering is just not the way to go and it is crazy to me that so many devs on Switch triple buffer at 30 fps (DOOM / Alien Isolation port have soooooo much input lag)

3) If targeting 30 fps, use half-refresh v-sync from inspector -- adaptive v-sync doesn't work correctly in conjunction with fps limits it seems

4) If it doesn't cause problems, lower the flip queue -- displaylag did some tests forevor ago that showed in SFV for example, dropping the flip queue size with traditional vsync did reduce input lag.

As I understand it, this is how one can get the "optimal"/"lowest" input lag with traditional V-Sync and with a controller, it feels pretty close to console (or better in some cases) in my experience.

 

@Warkratos This is what greatly confuses me.

In the past, from the tests I've read recently at least, "Optimal" vs "Prefer Max" power modes showed little to no discernible difference in fps/game performance.

So what on earth does this stuff mean? Does it mean that optimal and prefer max perf modes will work differently with the FPS limit engaged than without it?

Does it mean those power modes have been altered in general?

It would be very helpful to see more testing done specifically with regards to this power mode setting in conjunction with and without the new fps limiter and looking at input lag me thinks.

Or, has this already been done and I'm not aware of it?

 

The only thing I can think of is that in some scenarios where you limit the framerate, the GPU may not be taxed heavily enough and thus downclock more (or rather too-) aggressively when prefer maximum performance isn't set. This could lead to larger clock fluctuations or lower clocks in general and thus potentially lower performance and framerates/frametimes, stutter, higher (perceived) latency and other things.

For example, I can see this behavior on my 2080 Ti in some less demanding/older games with FPS limited to ~60 where the GPU/memory isn't taxed enough and drops clocks (sometimes even to 2D clocks), at which point the GPU is clocked to low and performance tanks long enough (or permanently) to cause low FPS, horrible stutter, G-Sync artifacts, etc.. Prefer Maximum performance seems to help with that in conjunction with cranking up Anti-Aliasing in Nvidia Inspector (if possible), as well as other options, use DSR resolutions etc. and/or unlocking the framerate to keep the GPU from being starved, but that doesn't always work either. It depends on the game.

It's one of the annoying side effects of dynamic clocks. I wish we could just set fixed clocks like in the old days so the card will always run at maximum clocks (that we know can run stable all day in modern, taxing games) when in 3D Mode no matter the load and be done with it or at least have a "Force Maximum Clocks" Option that we could set for certain software in Nvidia Inspector or the Control Panel to override dynamic clocks while said games are active.

 

the early AC games and starcraft 2 are examples, the formers will suffer fps dips when panning the camera across city roofscapes, and in starcraft 2 the framebuffer utilisation will put the game into an inescapable low power 3d.

 

I noticed with recent Drivers that Windows 10 is popping up a message about the Control Panel not being found, and sends you to the MS Store to download it.

Restarting the Container Service fixes it permanently though.

But is this normal behaviour now?

 

Which it still is.

 

is your dch control panel installed?

 

I used DDU to remove older drivers, then I installed 441.87 and for me Nvidia Control Panel app was installed automatically from Windows Store with proper popup message in system tray.

 

I just got the option to install it, windows just said it wasn't installed.

But another way to fix it is to restart the "NVIDIA Display Container LS" service.

I used DDU as well to remove the old driver.

This warning only seems to happen to recent drivers. Older ones don't show the Control Panel popup.

 

Possibly. Will be interesting to see what the review sites reveal when these features are tested.

 

That's how I noticed it, then I seemed to have a hybrid install of DCH and Standard drivers. A real pain to get the standard ones installed again.

 

Well I guess if you say it then it must be true.

 

I only just noticed the minimum FPS is 40, i wonder why.

Not that i use 30 or anything, but some really old games (like Interstate '76) prefer 30.

EDIT:

It doesn't even seem to work in old games... *facepalm*

Only V2 - Allow All seems to work in really old games.

Unless that's because the new V3 Limiter requires 3D Acceleration, and i was running it in Software Mode (Interstate '76 that is).