Upside-down SU9 Graphics Performance

[PacificSet90456]

In 2020 when MSFS was made available with its mind-blowing graphics, I upgraded my system for the best possible graphics according to the recommended MSFS specifications at that time. My configuration is 9 3900x, 2080Ti, 32 GB, 4k monitor. Graphic performance has decreased over the past 22 months even after trying hundreds of recommendations found here and elsewhere. The performance of SU9 was like watching a slideshow crippling MSFS for me. Graphic performance can mean many different things. Here only FPS is evaluated for the “best” graphics settings.

This is not a “how-to”, instructions, or specific recommendations for users. Users have tried the many of the same changes I tested with no change at all. I am not a graphics engineer and I do not understand what all these graphics parameters do or how they interact. What works for me probably won’t work the same for anyone else.

These FPS results are relative for comparison only. Actual FPS during flight should be better. To test differences in graphics parameters, I chose for worst-case performance and to establish a baseline KJFK Parking Spot 207 and the stock Asobo A320 cold and dark. The aircraft is not moving in order to use the same position for each test. In order to limit the scope of the testing, only 26 graphic parameters would be changed keeping everything else the same.

My baseline SU9 performance of FPS 5.9 is unacceptable.

My graphics are set to ULTRA. My system is a “high-end” system and performance should be much better. Memory was filling up and Windows started moving parts of memory to virtual storage. Basically, MSFS trashed my system!

I changed my graphics to HIGH which gave me a FPS 27.8.

This was a significant improvement compared to ULTRA. The render times decreased significantly. Memory usage decreased as well.

I wasn’t satisfied with this performance. For the next test the graphic parameters were set to LOW which gave me a FPS 45.1.

This was a huge performance increase! Render times improved significantly and memory usage decreased as well.

In order to find the best balance between FPS and graphics quality, the graphics settings were changed to a mix of HIGH, MEDIUM, and LOW. The goal was to achieve an acceptable FPS, no stutters (to be tested later), and good graphics. The updated settings provided about 35 FPS for this test.

I am very puzzled. It seems like graphics performance is upside down. Shouldn’t ULTRA be 45.1 FPS? Shouldn’t LOW be 5.9 FPS? I always thought ULTRA provides the highest FPS and graphics quality and LOW provides the lowest FPS and graphics quality. Also, shouldn’t the bigger the system have a higher FPS? And the smaller older system have the lowest FPS?

Looking only at FPS, SU9 performance (FPS) seems to be “upside-down”. Of course there are other factors to consider such as stutters, pauses, scenery and buildings popping in and out that were not part of this evaluation.

Note: These are the only 26 graphics parameters that were changed: Light Shafts, Object Level of Detail, Reflections, Render Scaling, Resolution, Shadow Maps, Terrain Level of Detail, Terrain Shadows, Terrain Vector Data, Texture Resolution, Texture Supersampling, Texture Synthesis, Trees, Volumetric Clouds, Water Waves, Windshield Effects, Ambient Occlusion, Anisotropic Filtering, Anti-aliasing, Bloom, Buildings, Contact Shadows, Depth of Field, Display Mode, Global Rendering, Grass and Bushes

[JetWash1023]

Not sure quite how you’ve got this so wrong but you have.

Graphical fidelity is inversely proportional to performance. ULTRA taxes your machine the most and therefore will give you the lowest FPS, LOW taxes your machine the least so should give you the highest FPS.

At the moment the game is so limited by MainThread CPU usage that there can be very little between LOW and ULTRA. Everything rests on LODs, which can murder performance on most machines when set much past 200.

[snowrider349]

5.9 FPS seems to be reasonable for rendering 7680x4320 on a 2080TI

[arniex78]

@PacificSet90456

I am very puzzled. It seems like graphics performance is upside down. Shouldn’t ULTRA be 45.1 FPS? Shouldn’t LOW be 5.9 FPS? I always thought ULTRA provides the highest FPS and graphics quality and LOW provides the lowest FPS and graphics quality. Also, shouldn’t the bigger the system have a higher FPS? And the smaller older system have the lowest FPS?

Is this a rhetorical question - or would it be logical according to your perception ?
Quite simply explained - ULTRA SETTING (highest possible settings) best graphics - lowest FPS !
LOW SETTINGS (lowest setting) worst graphics - highest FPS !
and I wouldn’t call a 2080 Ti “high end” anymore - maybe the 3090Ti - not that your graphics card is bad - but even my 3090 sometimes “stumbles” at the highest settings (although I absolutely can’t complain in terms of performance on the MSFS for the graphics offered).
The problem is - I see here also in the forum - of course the performance can be optimised - BUT people want, realistic weather worldwide optically and technically, AI traffic worldwide, realistic photogrametry worldwide + NEXT Gen graphics etc etc and want to have at the same time with a 3 or 4 years old hardware and this with 60 FPS (constant). That won’t work - I’m not saying the engine runs optimally no question, but you have to cut corners somewhere. Most of it, for me anyway - so the settings that cause the most stutters are the LOD settings - from the rest the PC remains pretty unimpressed (64GB RAM, 3090RTX, 10940X CPU, all SSD). So the best thing to do, as you do it - turn on developer mode - go through the settings - look at the latency and find the optimal settings. There are also strange errors with the MSFS, no question - but when I look at what is offered, I find that it runs well on balance.
The only thing I can possibly “blame” the Asobo team for is unfortunately the lack of information about what has been changed - for example, as Ben Supnik (XP11) often does “ok we changed this and you need to be warned that the SIM now runs slower so you need to change this setting a bit” (as an example).
Maybe something is moved from the graphics card to the CPU or vice versa, let’s say they improved the weather compared to SU7 (thank god ) - and it requires more power (on the graphics card), then I have to lower the quality of the clouds. The problem with these constant complaints (not related to your topic here) is that when people complain, the features are unfortunately “cut” and often without the information (LOD of the trees etc., weather - please don’t etc.). But when I look at the other SIMS - XP11 with xEnviro fully switched on - even my PC barely manages 25 FPS - you’d rather need a NASA supercomputer.
And that has always been the case (since FS98), that the SIMS always needed 3 hardware levels until they worked well enough (ok FSX about 5 later hardware generations ) - so in general when the MSFS runs - it doesn’t do that bad at all !

[HethrMasn]

@PacificSet90456

In order to limit the scope of the testing, only 26 graphic parameters would be changed keeping everything else the same.

Note: These are the only 26 graphics parameters that were changed: Light Shafts, Object Level of Detail, Reflections, Render Scaling, Resolution, Shadow Maps, Terrain Level of Detail, Terrain Shadows, Terrain Vector Data, Texture Resolution, Texture Supersampling, Texture Synthesis, Trees, Volumetric Clouds, Water Waves, Windshield Effects, Ambient Occlusion, Anisotropic Filtering, Anti-aliasing, Bloom, Buildings, Contact Shadows, Depth of Field, Display Mode, Global Rendering, Grass and Bushes

---

[trgz]

Looks like you want to drop your Render Scaling to 100% for starters

[bucsgolf142293]

Trying to render in 8K? And blaming the sim for 6fps on ultra. Ultra setting means the best graphics, not ultra fast fps.

There’s a lot to blame the sim for with bugs and CTD but at your settings/system trying to upscale render to 8K is asking a bit much. Watch some of the MSFS settings guides to dial your settings in appropriately.

[PacificSet90456]

@JetWash1023

Graphical fidelity is inversely proportional to performance.

You are 100% correct! I was deliberately only measuring raw FPS, not graphics fidelity. ULTRA does not necessarily mean MAXIMUM FPS. And LOW does not mean MINIMUM FPS.

[JetWash1023]

No, ULTRA absolutely does NOT mean maximum FPS. It means maximum graphical fidelity which, by definition, means lowest FPS!

[ProvenSilver]

You guys are out of my league here, but this is what I have discovered. I’m not at my flightsim computer now so I may have some nomenclature wrong, but it is a stock HP OMEN with Ryzen 5 5600G, RTX3060, 16 GB 3200 RAM, and I have two copies of MSFS 2020 - one Store and one Steam (long story). I have tried all types of internal graphics settings and find that my best performance is to select (Set to defaults or something like that) at the bottom of the Graphics Settings page. Then I restart and everything is on ULTRA and FRs are better than I get when I go through everything trying to pick and choose. The only thing I do with my GPU is use MSI Afterburner with nothing changed except to drop the voltage to about 8.5 and the fan curve modified to keep the temperatures down. No overclocking. Without that, my GPU runs at 78 to 81 C and that is too hot for my liking. With the above settings, I run at about 58 to 61 C. By the way, the Steam copy is on SSD and the Store copy is on SDHD and, once loaded, I see no difference in the flying experience. Another by the way, I fly no add-ons except the FBW A320. There are plenty planes in the Deluxe version to suit this old man.

[PacificSet90456]

@arniex78

The only thing I can possibly “blame” the Asobo team for is unfortunately the lack of information about what has been changed - for example, as Ben Supnik (XP11) often does “ok we changed this and you need to be warned that the SIM now runs slower so you need to change this setting a bit” (as an example).

Yes! The lack of information is appalling! Release notes are published but don’t contain any information as to possible impact of changes or fixes.

For example, each World Update contains a lot more objects and object details. But nothing about the impacts of the additional objects. Before and after performance information in the Release Notes might help.

I don’t think Asobo has made the connection between FPS and graphics quality very clear. They initially published recommended system specs but never connected them directly to the graphics settings. I thought that a top-end system would be able to deliver high FPS with high graphics. And do so without a lot of tweaking and tuning! Very bad assumption on my part!

Maybe the definitions of ULTRA, HIGH, MEDIUM, and LOW need redefining somehow to differentiate between ULTRA “graphics” and ULTRA “FPS”. I honestly thought I could have higher (maximum?) FPS somehow by figuring out how to configure the ULTRA settings while not using LOW or MEDIUM.

Thank you for your feedback!

[PacificSet90456]

@trgz

Looks like you want to drop your Render Scaling to 100% for starters

Thank you for the suggestion! Actually, I was trying to keep everything the same outside of the parameters I changed to look at only the FPS differences. I could run the same test only changing the render scaling but I assume the results would be the same but not the same FPS. That is higher graphics = lower FPS and lower graphics = higher FPS. My FPS would be much higher at 2k or 1080p but I wouldn’t like the graphics at all.