[DX11] Help regarding Nvidia Profile Inspector (NPI)

[black-coffee]

I didn't wanna give up on it, so I tried 32xS [Combined: 2x2 SS + 8x MS] and have better performance than with 4x MS 4xSGS...?!? Both options have the same apparent quality of image, the 32xS has better performace however for whatever reason. And yes, I was talking 60FPS as I personally find 30FPS unbearable to play with, for a driving game you need buttery smooth video, I also use G-Sync, once you go G-Sync or FreeSync you just can't go back to just V-Sync... With 4xMS4xSGS FPS 30-60 but very susceptible to frame drops. With 32xS the FPS more stable between 50-60.
Also word of advice, anyone using the g_traffic command beyond 2 or many traffic mods may experience dramatic drops in FPS.

[roextended2020]

Hi BenganJ, thanks for the tutorial. It is productive
I have 3 questions:
1. Set the scale to 100% in the tutorial.
Why can't I use 400%, what problems does 400% bring?

2. Can I open MFAA in NPI at the same time?

3. The AI traffic packet occupies CPU and 40 frames, causing my GPU to fail to output at full capacity. Is there a solution to this problem?

[BenganJ]

@black-coffee

Good to hear that you have found some parameters that work better.
The gaming environment is so different for all of us that it can pay off
to test the alternative settings rather than using the default ones that
is shown in first post.

@roextended2020

1. This scheme ONLY works when you set Scaling to 100%, it was the same on DX9.
What happens if you use any other setting is that you LOSE AA!
2. That's a good question because if I can remember correctly MFAA ONLY works
if there's already MSAA in place, then it can make so the a lower level of MSAA
together with MFAA will act as almost the same AA, but with better performance!
In DX9 I tested that, but to be honest, I haven't done that in DX11 yet. The question
is if FORCED MSAA, the one we use is enough OR if it MUST be implemented in the
game. So please, be my guest and test that you too and report your result back
here!!
3. Yes, as black-coffee says in the post above yours, density of AI traffic is a BIG FPS
killer, so set g_traffic at a proper value AND try some sort of FPS cap to ease the
work on the GPU, such as G-Sync, FreeSync, V-sync .....! There are several Frame
Rate Limiters in NPI e-g-!

[JHTrucker]

@black-coffee - Good to see you're still persisting!
I, like you need 60fps for racing sims, project cars2 for example, but i found having to use 30 fps for ETS2 (forcing DX9 AA) many years ago was the only real option for me or use a smaller screen res. At first 30 is horrible but i found after an hour or so i'd forgot about it and i never had a stutter or frame drop, it was smooth, not 60 fps smooth but more than good enough for slow truck driving.

[black-coffee]

G-Sync and FreeSync DON"T cap the frame rate, they are just a tech that is great for low latency and low lag. ONLY V-Sync can limit the FPS.

[JHTrucker]

@LarryA - GPUs x 2 Evga GTX 1080 ICX SLI "No"

Do you purposely disable SLI for ATS or can't you get it to work?
Just wondering as i'd noticed some time ago that the ATS profile doesn't include an SLI code but the ETS2 profile does.
Perhaps copy the ETS2 SLI code and paste into ATS to check?
I've never used SLI so this is just a guess!
Feel free to try or ignore.

[BenganJ]

@black-coffee

I know that fact, but I just mentioned them in the same context as Frame Rate limiters,
as they are often companions!

[JHTrucker]

@black-coffee - I don't know if you tried 32xS after my comment regarding your 2060 @ 4x and LarryA's 1080 @ 32xS but your further comment stating 32xS has better performance than 4x inspired me to test this out.

3840x1024 - GTX760
I have "start in cab 1" - set AA and start game, check FPS, no driving or looking around, just parked.
4xMS + 4xSGSS = 24 fps
8xS (combined 1x2 SS + 4xMS) + 4xSGSS = 26 fps
16xS (combined 2x2 SS + 4xMS) + 4xSGSS = 26 fps
32xS (combined 2x2 SS + 8xMS) + 8xSGSS = 26 fps
Repeating the above kept the same results.
My GTX760 is obviously bottle-necked at this AA, running at 100% load.
Whilst the above are not game play fps for me they do show that using the 8/16/32xS options are less demanding than 4xMS + 4xSGSS and probably 8xQMS + 8xSGSS, i forgot to test.
Does the combined super sampling part make the difference? Who knows.

As an extra, i reduced my res to 3072x768 and repeated the above.
4xMS + 4xSGSS = 40 fps
8xS (combined 1x2 SS + 4xMS) + 4xSGSS = 42 fps
16xS (combined 2x2 SS + 4xMS) + 4xSGSS = 42 fps
32xS (combined 2x2 SS + 8xMS) + 8xSGSS = 42 fps
I then did a few deliveries around Washington and the heavily wooded road areas (@16xS) and kept a steady 30 fps (vsync2) which in the past using 4xMS + 4xSGSS was choppy in those areas.

[BenganJ]

@JHTrucker & @black-coffee

I have used the 32xS (combined 2x2 SS + 8xMS) + 8xSGSS almost since the beginning and
I have felt it to be very smooth. It's just a feeling I have and is more or less a comparison
with how I remember the DX9 solution.

[JHTrucker]

@BenganJ
What would be handy is if someone with say a GTX1660ti could do a quick fps comparison between 8/16/32xS to see if the fps are all the same, vsync off & limiter off. No driving, just as i did for a quick test.