On 1/2/2025 at 11:56 AM, Cpt_Piett said:

I've been giving myself a massive headache trying to troubleshoot VRAM overflow (with a 4090!) and terrible stutters overflying certain PG-areas.

I was also having massive stutter issues (out of control), and you know how I fixed the problem?  I  reversed back to the old driver I had. Sure enough, almost instant solution, no stutters. You can’t underestimate the power a faulty driver can have on MSFS. It can induce massive stutters. I would try different drivers and see if you’re able to regain your sanity. You will need patience but it might solve the issue.

Solving stuttering  could potentially take days/weeks, and the frustrating part is that it is difficult to diagnose the issue. In most cases, it’s the result of a faulty driver. 

Edited January 3, 20251 yr by Huascar

On 1/2/2025 at 11:56 AM, Cpt_Piett said:

AutoFPS cannot save the day. Even with a TLOD of 100, there are severe stutters.

Yes, it is not a frame issue.  My FPS were in the low 30s with AutoFPS and it was a stutter fest.  I tried several drivers until I nailed it with a Nvidia driver released in early November.  With MSFS you will always have some stutters, it’s the nature of the beast.  

2 hours ago, Richard Sennett said:

would I notice a huge difference with that AMD chip ?

https://cpu.userbenchmark.com/Compare/Intel-Core-i9-10900K-vs-AMD-Ryzen-7-9800X3D/4071vsm2356328

 

 

 

Edited January 4, 20251 yr by turbomax

Cool thanks for link 🙂 

The performance impact of photogrammetry

We know that PG hits the CPU hard. I just wanted to find out how hard. And also check the impact on RAM and VRAM. 

Test: Fenix A319 KLAX ILS runway 25R from 3000ft to full stop on the runway

Settings: Time noon, clear skies, 4K, FG on, DLSS+DLAA, TLOD 100, OLOD 200, most settings ultra (a few settings e.g. traffic airport quality are off)

I've disabled Asobo's handcrafted KLAX as it causes bad stutters on approach. 

PG off: 

PG on:

Well, I didn't expect this - performance is significantly better with PG on. Some important numbers:

                                     PG off          PG on

Average FPS:                 91.1            92.7      

1% low average FPS:     37.1            47.3

0.1% low average FPS:  11.5             15.8

FTV <2ms:                      86.3%        89.6%

RAM:                               34.1            35.4

VRAM:                             19.3            18.4

Looking at these numbers, the difference does not seem huge. However, when doing these two flights, I noticed that the sim was a lot smoother with PG on. Fewer stutters overall, and the stutters weren't as severe. 

Also note that the range on the y-axis is greater with PG on. The reason for this is that there was one severe frame time spike at more than 300ms. But overall, the frame time graph looks a lot better with PG on vs off.

However, with a TLOD of 100 and PG on, all the buildings look terrible at 3000ft. There's constant popping in of buildings and objects. For both PG on and off: the runway is blurry even at 500ft. 

I could probably crank the TLOD up to 200 to improve the PG rendering. But in doing that, I expect to see more frequent and severe stutters. 

Other PG-areas might not be as bad as LA - I deliberately chose it as it's one of the largest PG-areas in the sim. 

I'll do one more test with TLOD 200 and report back. 

23 minutes ago, Cpt_Piett said:

I'll do one more test with TLOD 200 and report back. 

It never even occurred to me to fly with 100 TLOD. I do 200 minimum, and occasionally switch to 400 with not that much difference in FPS, at least not that I particularly noticed.

Just now, HiFlyer said:

It never even occurred to me to fly with 100 TLOD. I do 200 minimum, and occasionally switch to 400 with not that much difference in FPS, at least not that I particularly noticed.

True. It's just that landing at LAX with a high TLOD doesn't work very well. For me at least. Usually I use AutoFPS with a TLOD max set to 600. 

I'm currently testing TLOD 200. I don't think my hardware can handle much more than that doing an ILS approach into LAX. 

Photogrammetry off vs on at TLOD 200

Exactly the same as above, Fenix A319, KLAX ILS 25R. I added a third session with FSHud and PG on. 

PG off:

PG on: 

PG on + FSHud w/ AIG models (max 20 active and 30 parked):

                                     PG off          PG on       PG on + FSHud

Average FPS:                 91.0            84.5         81.8    

1% low average FPS:     54.2            29.1         36.1

0.1% low average FPS:  25.6            10.0         17.3

FTV <2ms:                      89.3%        76.9%     70.5%

RAM:                               35.9            33.0         42.7

VRAM:                             19.7            19.9          21.1

At a TLOD of 200, the average FPS is still okay. The 1% and 0.1% low average values as well as FTV <2ms are however significantly worse with PG on, indicating that the sim is not as smooth. 

Things are a bit worse with FSHud, but not that worse. For some reason, 1% and 0.1% low average is better (which doesn't make much sense). The FTV <2ms (which correlates well with stutters - the higher the better), is worse though. The frame time graph with FSHud looks a lot worse. When flying the approach with PG and FSHud, I certainly noticed more stutters - although it was not severe.

Overall, I've noticed that performance is significantly better in 2024 vs 2020 when using FSHud. Not sure if it's the same with BATC (I own it but don't use it). 

One very noticeable visual difference, is that even at a TLOD of 200, there's constant popping in of PG-objects in the near distance. Because of this, I much prefer to have PG off. Drawback: LA CBD does not look nearly as good with autogen buildings. 

Edited January 4, 20251 yr by Cpt_Piett

2 hours ago, Cpt_Piett said:

Overall, I've noticed that performance is significantly better in 2024 vs 2020 when using FSHud. Not sure if it's the same with BATC (I own it but don't use it). 

FS2024 has better performance than MSFS with any addon that weighs on the CPU. Also FSLTL injector tanks the CPU frametime in MSFS but it's almost irrelevant in FS2024.

Vanilla VS Vanilla, MSFS is lighter overall and much lighter on the GPU.

Edited January 4, 20251 yr by MrFuzzy

11 minutes ago, MrFuzzy said:

FS2024 has better performance than MSFS with any addon that weighs on the CPU. Also FSLTL injector tanks the CPU frametime in MSFS but it's almost irrelevant in FS2024.

Vanilla VS Vanilla, MSFS is lighter overall and much lighter on the GPU.

Agree. 

I think it's time to stop thinking that your hardware and settings are the cause of stuttering and other existing problems. For me the major problem is their cloud server, no more no less. After FS2020 it's the same with FS2024. It depends on the day and time of the week you use FS. It's completely random for a good or bad experience. Yesterday for me it was catastrophic at London-Heathrow or Madrid-Barajas. In addition to the problems I already mentioned about the TLOD. See also the air traffic planes landing with micro jerks without their landing gear .... It's just for me a frustrating and completely immersion-killing experience. All FS is based only on the representation of the entire planet which makes it beautiful but everything else is catastrophic for immersion.

6 hours ago, filou said:

 

I think it's time to stop thinking that your hardware and settings are the cause of stuttering and other existing problems. For me the major problem is their cloud server, no more no less. After FS2020 it's the same with FS2024. It depends on the day and time of the week you use FS. It's completely random for a good or bad experience. Yesterday for me it was catastrophic at London-Heathrow or Madrid-Barajas. In addition to the problems I already mentioned about the TLOD. See also the air traffic planes landing with micro jerks without their landing gear .... It's just for me a frustrating and completely immersion-killing experience. All FS is based only on the representation of the entire planet which makes it beautiful but everything else is catastrophic for immersion.

 

Well said. I was also contemplating this morning the issue of trees. Their sizes are not in proportion to houses, buildings and other structures. They’re also ubiquitous. For instance, the outside of Washington DC looks like a forest, or more like the Amazon jungle. The trees are everywhere and in unrealistic quantities. While both 2020 and 2024 have made some significant technological advances, there are many features that completely destroy immersion. I tried 2024 for a day, and I do remember seeing planes landing without the gears deployed, and moving in illogical directions. How long will it take them to fix that issue is anyone’s guess.

Edited January 4, 20251 yr by Huascar

I wonder when they will fix that popping of objects in front of me. Really hate this the most. Can live,  sometimes,  that trees and buildings looks like from Minecraft, but this popping of airport parts a few meters in front of my plane is making me nuts...

On 1/4/2025 at 12:23 PM, filou said:

I think it's time to stop thinking that your hardware and settings are the cause of stuttering and other existing problems. For me the major problem is their cloud server, no more no less. After FS2020 it's the same with FS2024. It depends on the day and time of the week you use FS. It's completely random for a good or bad experience. Yesterday for me it was catastrophic at London-Heathrow or Madrid-Barajas. In addition to the problems I already mentioned about the TLOD. See also the air traffic planes landing with micro jerks without their landing gear .... It's just for me a frustrating and completely immersion-killing experience. All FS is based only on the representation of the entire planet which makes it beautiful but everything else is catastrophic for immersion.

Yep..
Cloud gaming is the issue.