Tested DLSS E and F Presets

User Support Hub Virtual Reality (VR)
,
1

I decided to finally try comparing DLSS E and F presets to see if I could tell any difference.

Test conditions:

  • RTX4090
  • 7950X3D
  • 64 GB Ram on XMP profile
  • Pimax Crystal OG at 1.0 render scale (Quality) in Pimax settings
  • No Smoothing
  • No resolution override in OXRTK
  • Cessna 172 G1000
  • DLSS v 3.8.10.0
  • DLSS Performance
  • VR Camera set to default position

While it was hard to tell I think preset F did have a slight advantage on clarity. I was specifically looking at the PA button and left side (engine) of the MFD to make this call. For the PFD, it was hard to tell much difference between the two and I observed no significant difference in performance.

Other Observations:

  • The distance to the glass screen plays a big factor. I think in F may work a little further away. I think the airplane and seating position may factor in as to why some people don’t get a benefit.
  • I did some experimenting with sharpness. Everything from 50 to 200 basically looked good. At a sharpness of 50 the battery and alternator buttons had slightly less shimmer. At 0 sharpness I could tell the image was blurry and the buttons still had their shimmer.
  • I also tried Balanced and Ultra Performance. The difference in performance seemed to be around 8 to 10 FPS between the modes. In Performance I was seeing between 43-55 fps. Balanced looked a touch better but it would be tough to not dip into the 30s.
  • Under any combination of settings, the world details looked more or less the same. It’s only the instrument cluster and screens I noticed big differences.

I wish I could do this experiment blind to the settings! These are all subjective observations too. I’m curious to see what others may find.

1 Like
2

Comparing:

  1. Performance on F + oversampling in OXRT (4500 x xxxx)
  2. Quality on E without oversampling (4312 x 5102)

(both giving me 35-40 FPS), I believe distant objects towards the horizon looks a tad sharper and more like the real world with the 1st config, with less shimmer. But this is very subjective.

I prefer clarity over FPS, 35+ seems acceptable to me in MSFS2024 (smoother than MSFS2020 at the same FPS).

7800x3D, 4090, 32 GB RAM
Pimax Crystal OG
A2A Comanche, regional airport add-on (ESMS), weather preset with few clouds, clouds on High, TLOD 100

1 Like
3

That was going to be my next test.

Seems like DLSS Tweaks could be fined tuned a bit more to get things dialed in.

It may be we could fine tune the quality factor for better results over the final resolution rendering? :thinking:

4

That’s a very specific TLOD.

I played around with TLOD but I didn’t see a major difference from 100 to 200 and I seemed to stay GPU locked. But I did not do any scientific type tuning.

5

TLOD100, post corrected.
I tried TLOD50, terrible blur of terrain textures.

6

to the heroes testing DLSS and presets, if you ever come across a version and preset that really fixes the smearing/ghosting on the airliner(example: B738M) glass screens that would be amazing.

3.7 still has all the presets right? 3.8 only has E and F from what I understand, hence the smaller filesize.

They say F is for Ultra Quality and DLAA, so basically no upscaling. And E is the other DLSS presets. I think from all the comparisons videos I watched on youtube the E preset seemed to be better on average.

7

It’s better to see by yourself in the headset instead of watching the videos…

2 Likes
8

Not a great solution, but I’ve noticed distance is such a huge factor. Clarity is like an inverse squared or cubed relationship to the instrument cluster.

I’d see if you could bind camera translations to something like a hat switch it would help. I do it even with smaller aircraft and I don’t even think about is so much now.

MSFS would benefit so greatly from a simple zoom feature if we had that.

1 Like