Still Stutters and poor FPS after the 1.14.6.0 Update!

The US, UK and Japan Points of Interest. There are other US, UK and Japan selections. Didn’t mess with those. ONLY the Points of Interest selections.

Are you a programmer? Do you have any idea what you’re talking about with multi-core processors? People act as if they are some magic bullet for everything and yet they really only shine when some task can be dumped off to another core that has nothing to do with what the main core is doing.

Anything that has to stay in blistering synch to the main task of rendering would take so much extra code to synch it back to the main processes that whatever advantage you gained would be lost. Everything vital has to come through the main core. Code like downloading graphics, building the 3d world from it, grabbing weather, computing ATC is sequential. Task A happens, then task B, then task C because each task relies on the successful completion of the one before it. If you send task B off to another core, task A still has to wait for it to complete before task C can begin. Does that make sense? The only time you can get away with allowing another core to do something for you is if that task is completely separate from the sequential stuff or interacts rarely with the primary tasks. Flight sim code is almost all sequential stuff. Now if you’re multitasking, browsing the internet while your flight sim is running (and ignore the hit on network latency) you can run your other program in another core although it will still be in contention with the FS task since it has to compete for instructions sent to the GPU and other inter-related things.

You have some sequences that have to happen to render a scene, all before the GPU even is utilized.

  1. Transfer of data to and from Asobo servers if you’re using photogrammetry.
  2. Filling in the parts you don’t get fully defined that are stored on your computer hard drives and hopefully in memory. Things like autogen, water features, animations, trees, etc. Composing all of those things together to build the 3d scene complete with lighting positions, fields of view…
  3. Weather info whether real world (another data transfer which is positional dependent) or generated if you have it turned on. I truly doubt most of us fly on cloudless days.
  4. Traffic info (similar to weather)
  5. Aircraft info especially interior with all the guages and real time displays.
  6. Continual interaction with all the possible controls. Many of those code events are highly dependent upon sequence.
  7. ATC (that one you might be able to have some of it such as the audio speech generation happen on another core)

These are just off the top of my head. I’m sure Asobo programmers could add a bunch more steps that have to complete to render even a SINGLE frame to send to the GPU to display and nearly all of those steps are sequential as I described above. You can’t fully display the shadows on the ground until you know where the clouds are above it as well as the dynamics of the terrain and what’s shadowed by the ambient light so you can’t send the cloud calculation off to another core because the next step on the main core has to wait for it to return full info before going on to the next. This is an extreme oversimplification of course and my steps are likely out of sequence.

Great FPS is primarily a function of internet speeds and the overall clock speed of the processor not the number of cores unless you are running multiple programs. Even then most other programs are designed the same. They default to the primary core. I’m not leaving the GPU out of this. Obviously the better cards can take that CPU info and render it instantly but most lag is not there with the faster cards. It’s at the CPU.

3 Likes

That actually makes a lot of sense because so many things run in the background on Windows 10 even if the only thing you launch is MSFS and nearly every program defaults to using the primary core for the bulk of its processing. Having MSFS start in an unutilized core then would give it more processing room. Might be worth a test.

But the stutters in my video are being caused by the main thread, which is something similar to the stutters we saw before the patch but in more acceptable way

Thank you. I created a support request and sent them my video.

2 Likes

that‘s what this thread is all about since weeks

I agree with you 100% on this but it’s crazy that today in 2021 an i9-9900K OC to 5.ghz can’t handle MSFS 2020. I feel like they ported a lot of unoptimized code from FSX and that they rushed MSFS 2020 out the door. I’m sure there’s a LOT of optimization still to be done. But unfortunately, they have to mainly focus on features and content as this is what will keep the game selling. In my opinion, I would prefer for them to take 6 months and focus on optimizing the game then to have new world updates and liveries…

3 Likes

Still getting major hick ups in some areas, for example north of Seattle when flying from Tacoma Narrows to Concrete. I have 48 GB of RAM and I didn’t see more than 60% overall RAM usage, so I doubt it had to do with swapping to the page file.

This is what I also suspect…

I am not making a criticism of the work of Asobo Studio. I have no doubts about all the processes at work to make the software work. I just note, from long experience, that games are not the strong point for modern multicores processors (except high frequency). Especially in a complex system like the flight simulator. I hope that we will eventually overcome this difficulty (we can clearly see that it is not an easy thing) because the frequency of our processors evolve little while the number of cores increases (except for the 11th generation of Intel).

1 Like

I tested stuttering again in the same places caused by incorrect processing of TLOD’s.
I turned off all online functions (Bing maps, Traffic, etc.). I also turned off all caches.
I set the lowest graphics quality. I just set TLOD at 200 to highlight the problem, because with a small TLOD it doesn’t get so angry.
I also tried two different airplanes (Daher 930, Cesna 152).

Result: Stuttering in the same areas is still the same!
My SIM also dropped several times while stuttering (CTD).

I think the problem is somewhere in basic map data or in the SIM.
I no longer think that stuttering is caused by the performance of CPU, GPU, RAM etc. (in my case)
I reported this problem to SIM support, so I wonder what will answer me.

I recommend others, who have similar problems, try to turn on Debug LOD (Options - Debug LOD - Display exterior models LOD) in Develop mode. To see if stuttering is causing the same problem.

2 Likes

New addition to the show…
Turned off “windows game mode” and “xbox game bar”
No more CTD but…there is still no way to land on an airport and also try recording it.
FREEZES just before touchdown for 5 secs and sometimes a little more earlier even for 10 secs
GREAT JOB ASOBO !!! :-1:


And check the pic in Miami.Where are the roads between the land?

My guess is there is not much more to optimize. Yes, they used FSX technology but FSX didn’t invent it. Many of the core routines for creating 3d graphics (which is the bulk of the processing) were built a decade or more ago and the work of absolute mathematical geniuses. You can tweak only so much in stuff like this. The fundamentals are exactly the same as the most highly optimized first person shooter. What is very different though is that the environment isn’t canned as it is in a game. It’s constantly being updated from distant servers with constant movement into areas and environments that are different for every user and where they fly. And all that info has to be assembled and crammed into animation frames fast enough that human eyes can handle and see smooth flow. Any slow down in any of the processes that create that frame, a missed packet from the internet, a process on your computer kicking off (and there are hundreds running in Windows at any time), etc, etc… causes your processor (no matter how fast it’s capable of going) of deciding either to wait or to go on with incomplete data. My guess is that any optimization available is just dealing with those hiccups which requires some real compromise on either speed or visual accuracy. In other words if you fix it one way you’ll get people complaining about drops in FPS but if you fix that you’ll get people screaming about something not rendering right.

I believe that the UK update kicked their asses with the complexity of scenery such as London and they made one of those compromises opting for graphical accuracy over speed. When complaints started coming in they did some more tweaking the other way. Of course much of the problem wasn’t only code in my opinion but the mere fact that as soon as the update hit everybody jumped on the sim again and tore up the server traffic which of course amplified everything.

When the stutters causes by the internet connection or from the Server connection why do i have still stutters when i turn off all online functionality? I dont understand that.

And we are still going

RdrThread, Manimulators and CoherentGTDraw are all over the place, just random 40min flight FL045 on the TBM, nothing really intresting around flying around ZGLG

I want so hard to love this sim but you are making it very hard, i’m really starting to belive Asobo is in over their heads after all and I defended this sim very hard since the alpha.

If you want my advice, stop World Updates, stop working on new eye candy, you have enough of that already, focus your WHOLE team on fixing the stuff you already have and is broken, then finish core systems of the aircraft like engine logics and avionics and such until you can at the very least insert airways on the garmings and change baro units, and make it so the plane reliably follows it till the end laterally and vertically, then you should start thinking of improving the eye candy with the world updates

just my 2 cents


1 Like

Please see my video: https://youtu.be/ekAtPqIvYFM

I agree - reading the posts. I have a “NVidia 1660 Super”. This is “mid class” in best case. My story: after 1.14.6.0 I have “degraded fps” with less then 20 fps. These problems I had not before. The “76T Bishop 1 frame per second” bug is gone.

I assume the stutter problem is in the vage “server request” - Asobo/Microsoft did not tell what the reason fpr the fixed “76T Bishop” bug was. To investigate I use the following Options, General, Data:

My reasons: To reduce server bandwidth I switched off the data I don’t need (much). As for rolling cache: I put 32 GByte rolling cache on my second SSD. But there was no read/write activity. Therefore I assume that rolling cache is not used in version 1.14.6.0.
Note: I put a fixed sized 16 GByte pagefile.sys on my second SSD. I have 16 GByte RAM.

After these changes I got no more “degraded fps”. But I will not say that I have found a “solution”. Specially if Asobo/Microsoft gives no hint whatsoever about the problem!

Update: The rolling cache has benefits. I just used the wrong measurement tool. Now I use 16 GByte rolling cache on my fastest SSD. I updated the graphics driver to NVidia 465.89. I see only one difference: If I record video the fps degradation in 465.89 is less then in 457.09, my old driver.

3 Likes

I saw the Video and its exactly my issue what i have with stuttering in the msfs 2020

Here’s a video I recorded of the micro stutters I’m having. During the recording, I was getting a constant 32FPS, but the issue makes it look like I’m getting <20. I hope they get it fixed soon… :frowning:

1 Like