I have no evidence for it, but I get the feeling that setting the PhysX processor to the GPU instead of the CPU, as it was for me, might have had a good effect (though does MFS use PhysX?)

Also setting the "Perform scaling" set to the GPU and ticking the "Override the scaling mode set by games and programs" since I use 90% Render scale in game to help out running at 1440p.

Probably easy enough to test.

I've turned sharpening off and it does look better. But overall, the game looks much better to me (possibly subjective) compared to what it's been from the start.

I personally wouldn't be messing with these nV control panel items.   The sim itself has a full range of settings.  Back in the days of DirectX9 games where the game itself offered just a resolution setting and maybe a single low/medium/high slider, this nV control panel stuff was useful.  Not so much anymore.  There are combinations that will do really bad stuff between the options in the sim and the options here.

36 minutes ago, Bottle said:

I have no evidence for it, but I get the feeling that setting the PhysX processor to the GPU instead of the CPU, as it was for me, might have had a good effect (though does MFS use PhysX?)

It does not use GPU-accelerated PhysX, and I don't think it uses the PhysX engine at all actually. The amount of GPU-accelerated PhysX games is extremely limited nowadays.

Yeah, I didn't think it used PhysX.

Then my next best guess is changing Texture filtering quality from Quality to Performance.

I tried the NVIDIA sharpening in one of my games and found the default values were over the top.  I had to tone it down considerably and then it provided a nice touch of sharpness without too much edge artifacting or noise.  I can't recall the exact sharpness setting I used, but it was much less than the default and it only cost 1-2fps.

But with a game like MSFS I would be reluctant to apply sharpening here, although I suppose I should try it.  I'm thinking that when flying, atmospheric haze is a big factor in perceived sharpness of objects on the ground and that any added sharpening is going to look unnatural.

40 minutes ago, Bottle said:

Yeah, I didn't think it used PhysX.

Then my next best guess is changing Texture filtering quality from Quality to Performance.

I'm assuming his suggested settings that make the most impact are:

- ambient occlusion -> performance

- FXAA / MFAA -> off (this seems to conflict with MSFS' TAA implementation)

- Low latency mode -> on (which changes pre-rendered frames)

- Power management -> prefer max performance

- Texture filtering aniso sample opt -> on

- Texture filtering quality -> high performance (obviously impacts texture quality)

- Texture filtering trilinear opt -> on

Good video, but on the colour settings, really try and use 10 bit with full dynamic range if you can...

 

One simple thing to try (though I'm aware many in the community have an aversion) is to allow Nvidia Geforce experience to apply its "optimized" settings designed for MSFS.

They provide a good basis for enhanced FPS within the sim, and a reasonable place to begin before moving on to more personalized tweaks.

EDIT: For those who would still rather be eaten alive by sharks rather than use GeForce, here are the settings.

 

Edited November 13, 20205 yr by HiFlyer

4 hours ago, bobcat999 said:

😄  I am wondering if he accelerated it after recording it actually, so that it didn't go on for too long.  If that is the case, more should do it I think! 👍

I use Firefox with the  universal "Playback Speed" add-on which allows once mouse click for any speed from .2 normal to 2.0 normal. And 10x, 40x fast forward. 2 other great Firefox add-ons for image screen captures: Nimbus Capture and Awesome Screenshot.

19 minutes ago, HiFlyer said:

One simple thing to try (though I'm aware many in the community have an aversion) is to allow Nvidia Geforce experience to apply its "optimized" settings designed for MSFS.

In my opinion, the main issue with the "optimized" settings is that these are from a render performance point of view, not for a simmer point of view. This is so because it is optimization meant for getting the most fps out of the video card as it is tuned by the video card manufacturer.

However as simmers I believe we do prioritize some visuals and settings which are more important, like distance and details, to the point we might be fine tuning down other settings which are less important for the simming experience in order to maintain our fps in exchange.

For a practical example, your example is using DLAA which is less sharp and detailed than TAA (a natural effect of how DLAA is working by the way). I wouldn't sacrifice this. Texture Resolution is set to MEDIUM, which I wouldn't sacrifice either nor Terrain Vector Data.

This is the same problem when using the video guides showing up the difference with each setting because these are always taken apart of the others. However when combining the settings you'll find you're hitting some walls you might not expect performance wise. The most impactful problem is when the combination of settings you're using is making the renderring engine over budgeting the frame time and resources. In this case you'll have stutters because of the bus traffic between the RAM and VRAM let alone the CPU/GPU waiting for each other.

This is why I've taken a different path and so far there are many reporting a significant increase in fps and rendering quality with less stutters in a wider range of hardware than what it was initially designed for:

https://forums.flightsimulator.com/t/my-2070-super-4k-settings-and-suggestions-episode-2/272854

You might want to review the rationale explaining the choices made both in the NVidia CPL and in FS2020.

I documented a 12-15% reduction in GPU load w/ no change in image quality nor smoothness when I set Low Latency Mode to ULTRA, that is compared to being OFF.

Also, LOD Bias to Clamped improved some aspects of cloud edges w/ their grid-like patterning from some angles.  I read DX-12 will fix cloud appearances substantially.

 

2 hours ago, RXP said:

However as simmers I believe we do prioritize some visuals and settings which are more important, like distance and details, to the point we might be fine tuning down other settings which are less important for the simming experience in order to maintain our fps in exchange.

Which is why I said it was a good base as far as FPS, on which to then make ones own preferred optimizations.....

10 hours ago, Bunchy said:

Unfortunately, the guy didn't speak fast enough for me.

Apart from the Firefox playback speed add-on (mentioned earlier) I found that the [excellent] freeware Potplayer also has a replay speed function. I had to slow it down quite a few notches to make it intelligible.

3 hours ago, Noel said:

I documented a 12-15% reduction in GPU load w/ no change in image quality nor smoothness when I set Low Latency Mode to ULTRA, that is compared to being OFF.

That makes sense as low latency mode to on or ultra reduces overhead since fewer frames need to be prerendered. Definitely notice an improvement in performance, although ultra could cause jitters under some conditions on my rig. 

3 minutes ago, DylanM said:

That makes sense as low latency mode to on or ultra reduces overhead since fewer frames need to be prerendered. Definitely notice an improvement in performance, although ultra could cause jitters under some conditions on my rig. 

Haven't noticed any jitters yet hopefully you didn't jinx me