When I first installed 5.3 on a clean Windows 11 install I had intermittent stuttering - it was worst in the PMDG 747 and not nearly so bad in the 777 and the FSLabs and it usually wouldn't start until some time into the session. Sometimes it was really bad and a reload of the scenario from autosave would help. I tried HT off/on and changing AffinityMasks, disabling HAGS, automatically clearing backup memory, EA on/off, setting road traffic to zero etc etc etc but none of the changes completely cured the issue. I think what finally fixed it for me was a combination of HF1 and using Process Lasso to move other apps off the P3D cores.

My recommendation is to use HT on and accept the default AffinityMask and then use Process Lasso to move your non Windows CPU intensive apps off the cores that P3D uses. In PL you can see which cores P3D is using and set a CPU Affinity (Always) to use other cores - for me 7 -11. I also did this for my browser (Edge) and a couple of other apps that run in the background and set P3D to 'Induce Performance Mode'. Process Lasso also showed that Windows Search Indexing was CPU intensive so in Windows Search settings make sure Search Indexing is not applied to your user and flightsim folders. I also used Orbx Central to sort the scenery index to their suggested default order as I have FTX Global and LC. I've done numerous flights since including 2 12 hour+ long hauls and the stuttering has completely gone. I have a lot of addons and scenery and I think my setup is fairly typical.

There is clearly still an issue with 5.3 that is causing stuttering for some people but as someone who used to have the issue and now doesn't I think it should be possible to eliminate it on most (non VR) systems.

5 hours ago, Rob_Ainscough said:

First you need to find out which are your best cores via the motherboard BIOS/EFI ... usually indicated with an * asterisk next to the core (on ASUS boards EFI).  If you aren't sure which is best, you can download Intel's PDT tool here where it will test your CPU and identify best cores.  I believe Intel only ID two best cores, AMD ID two best cores and two second best cores via AMD Ryzen Master software from AMD (not relevant in your use case).

Just tried this but in the BIOS nothing is marked with an asterisk and the Intel tool did not write down which cores are prefered :-/

This issue is really weird. I had it this morning after about 30min on a default airport with a default plane.

Restarted the Sim, same airport, same airplance, no issue. I hope that I can find out what is triggering the fps stutter stuff.

Just to be clear: This is not your usual "stuttering" or low FPS, which can have a myriad of reasons, it is a multiple hour long smooth stable 60+ fps flight which all of a sudden goes haywire, with single digit FPS and a CTD some time after that (for me it was 2 minutes). It happened to me once so far on the HF1, previously on 5.3 not. It happened in a non-heavy situation at cruising altitude over the desert with clear skies, so it's not a hardware related problem. It's a software bug, by all means, probably induced by some combination of scenery/weather/EA/whatever factors.
I think it has something to do with the KMCO stuff they fixed, but that is just a guess. I also will not rule out that it is caused by ActiveSky, because that's the only other thing that got updated with the Hotfix. Anyone without ActiveSky experiencing that?

7 hours ago, Rob_Ainscough said:

Looks like there could be multiple different issues some are having, not necessarily all the same issue.  Is everyone that experiences these "freezes" using HiFi ASP3D with EA on and ASP3D set to use EA options checked and running ASP3D on the same PC?

I went over this with Joe but might be worth repeating here as it relates to the new JobScheduler settings (I know Steve has provided input also):

First you need to find out which are your best cores via the motherboard BIOS/EFI ... usually indicated with an * asterisk next to the core (on ASUS boards EFI).  If you aren't sure which is best, you can download Intel's PDT tool here where it will test your CPU and identify best cores.  I believe Intel only ID two best cores, AMD ID two best cores and two second best cores via AMD Ryzen Master software from AMD (not relevant in your use case).

AffinityMask: value in all cases should be set to ALL the cores you have (real or HT).  

P3DCoreAffinityMask: this mask tells P3D what cores it can use relative to the AffinityMask (above)

MainThreadScheduler= best real core (main simulation everything ties to this core and will be close to 100%)
RenderThreadScheduler= 2nd best real core (DX12 work)
FrameWorkerThreadScheduler=3rd best real core (Autogen work)

With my 9900K stock clocks (not overclocked) the values that worked the best for me are: 

HT On

AffinityMask=65535
P3DCoreAffinityMask=65535
MainThreadScheduler=7 (that was identified as my best "real" core, these values represent real cores)
RenderThreadScheduler=1 (this is the 2nd best core)
FrameWorkerThreadScheduler=2 (any core other than 0)

HT Off

AffinityMask=255
P3DCoreAffinityMask=255
MainThreadScheduler=7 
RenderThreadScheduler=1 
FrameWorkerThreadScheduler=2 

These are the straight forward cases.  Now, if you are getting "really involved" in core assignments and you want to exclude specific cores and use Process Lasso or something else to assign a process like ASP3D or ChasePlane or other related executables to their own cores, then you'll need to define P3DCoreAffinityMask (note: always leave your AffinityMask to it's max core count value).

MainThreadScheduler, RenderThreadScheduler, FrameWorkerThreadScheduler represent "real cores" base 0, so for a 9900K that's 0 thru 7 ... but they are "relative" to the masked values from P3DCoreAffinityMask.  So if you have an P3DCoreAffinityMask that excludes core 0 (i.e. 65532) so as to allow the OS to do tasks on core 0 and P3D to do tasks on the remaining cores (1-7), then setting MainThreadScheduler=0 would actually be telling P3D to use Core 1 (zero base 0 - 7). 

So lets do a scenario for a 9900KS where you have HT On and your best core is real core 3 and your second best core is real core 5 (zero based 0 - 7) and you're using ASP3D, LittleNavMap, and ChasePlane.  I would recommend the following:

HT On

AffinityMask=65535
P3DCoreAffinityMask=16320
MainThreadScheduler=0  (this is core 3 - zero based)
RenderThreadScheduler=2 (this is core 5 - zero based) 
FrameWorkerThreadScheduler=3 (this is core 6 - zero based) 

For Process Lasso (or whatever tool you use to assign EXE process affinity) you set ASP3D to core 1 ( zero based 0 -7), you set LittleNavMap to core 2 (zero based 0 -7), and ChasePlane to core 7 (zero based 0 -7).

Now, here is the kicker that complicates using P3DCoreAffinityMask and core exclusion.  It's possible P3D may operate better if you don't exclude 3 cores as the terrain engine will use as many cores as it can to page in/out terrain as you move thru the virtual world.  It might be a good to allow P3D to use 6 of the 8 cores, free up core 0 for OS, and free up core 7 for ASP3D and LittleNavMap, and ChasePlane ... those 3 apps MAY not need to have their own cores and one core is sufficient to process all 3 apps without running into resource contention.  In this scenario then you'd be looking at:

HT On

AffinityMask=65535
P3DCoreAffinityMask=16380
MainThreadScheduler=2  (this is core 3 - zero based)
RenderThreadScheduler=4 (this is core 5 - zero based) 
FrameWorkerThreadScheduler=5 (this is core 6 - zero based) 

For Process Lasso set ASP3D, LittleNavMap, ChasePlane to run on core 7.  

In addition, you might want to switch the values in RenderThreadScheduler and FrameWorkerThreadScheduler pending on your balance between GPU and CPU loads.  If your GPU isn't working hard then set RenderThreadScheduler=5 and FrameWorkerThreadScheduler=4 per example scenario above so it gets a best core.

I do NOT recommend enabling the OS Hardware-accelerated GPU scheduling.  Along the OS settings, I haven't noticed any difference between Game Mode On or Off.

AffinityMask calculator HERE and many thanks to Ted Porter.

If you are overclocking a 9900K and you can get 200Mhz or more (stable) with HT Off via EFI/BIOS, that will be a better option for 8 real core or higher processors.  If you are NOT overclocking, then HT On would be the better option with P3D V5.3.

Clear as mud?  

Cheers, Rob.

What was that you were saying about this not being complicated or requiring a software engineer?  😉  

These are not things an end user should have to be figuring out.  If I were a commercial customer and had to figure this out on my own I would be ballistic.  Is this how LM ships 5.3 to commercial customers who are paying the big bucks or are we getting the same experience as they are?  One thing that has become very evident to me is that LM is not a software company LOL...  Yeah yeah, obvious I know but you can REALLY tell with the point upgrade process and all of this AM/Core parking/thread scheduler, renderthreadscheduler, framworkerthreadscheduler, HT on or off, this is just NOT user friendly and moving more towards something a dedicated software engineer would support than your average end user. 

Going from 5.2 to 5.3 or 5.3 to 5.3HF1 should not be a 9 step process for a point release, not even a major release should require that many steps.  It is a recipe for SW disaster. I don't have another app or program that requires jumping through this many hoops for an upgrade and if I told my customers they had to do this to move from a .2 to a .3 release they would go ballistic.  

Honestly I appreciate all of this is very simple and second nature to some but I have been out of engineering for 20 years and when I moved to Sales they cut off my technicals.  Snip Snip and all that knowledge appears to be gone because I am indeed finding all of this quite complicated but that is just me.  

Edited December 27, 20214 yr by psolk

31 minutes ago, psolk said:

What was that you were saying about this not being complicated or requiring a software engineer?  😉  

These are not things an end user should have to be figuring out.  If I were a commercial customer and had to figure this out on my own I would be ballistic.  Is this how LM ships 5.3 to commercial customers who are paying the big bucks or are we getting the same experience as they are?  One thing that has become very evident to me is that LM is not a software company LOL...  Yeah yeah, obvious I know but you can REALLY tell with the point upgrade process and all of this AM/Core parking/thread scheduler, renderthreadscheduler, framworkerthreadscheduler, HT on or off, this is just NOT user friendly and moving more towards something a dedicated software engineer would support than your average end user. 

Going from 5.2 to 5.3 or 5.3 to 5.3HF1 should not be a 9 step process for a point release, not even a major release should require that many steps.  It is a recipe for SW disaster. I don't have another app or program that requires jumping through this many hoops for an upgrade and if I told my customers they had to do this to move from a .2 to a .3 release they would go ballistic.  

Honestly I appreciate all of this is very simple and second nature to some but I have been out of engineering for 20 years and when I moved to Sales they cut off my technicals.  Snip Snip and all that knowledge appears to be gone because I am indeed finding all of this quite complicated but that is just me.  

Using AM and the latest provided features are optional. No need to use and change it mandatory. LM provided excellent CFG entries which allow user, if they like, to squeeze the maximum performance out of their systems and run the sim extremely smooth IF, read the if, the system is correctly setup AND software is correctly I stalled AND compatible. 

LM allows also a install and just run.....without making ANY changes. 

Maximum flexible.

But for the moment there seem to be a nasty bug in the latest HF1 (second HF1 release) wich triggers fps decrease and stutters. 

LM Support Forum are full of user reports and I myself have seen it once. 

Marcus

1 hour ago, Fiorentoni said:

This is not your usual "stuttering" or low FPS, which can have a myriad of reasons, it is a multiple hour long smooth stable 60+ fps flight which all of a sudden goes haywire, with single digit FPS and a CTD some time after that (for me it was 2 minutes). It happened to me once so far on the HF1, previously on 5.3 not.

If you're getting 60fps that suggests you're hittting your graphics card pretty hard. You don't list any kit in your signature unfortunately.

I have a 30Hz monitor and with VSync enabled my fps is of course limited to 30. Flying west over Russia and close to the Belarus border enroute to Paris my GPU has a 5% load at FL380 in the PMDG737.

Try limiting your refresh rate to 30Hz via NCP and see if that helps. I've not experienced this stutter-fest.

1 minute ago, Ray Proudfoot said:

If you're getting 60fps that suggests you're hittting your graphics card pretty hard. You don't list any kit in your signature unfortunately.

I have a 30Hz monitor and with VSync enabled my fps is of course limited to 30. Flying west over Russia and close to the Belarus border enroute to Paris my GPU has a 5% load at FL380 in the PMDG737.

Try limiting your refresh rate to 30Hz via NCP and see if that helps. I've not experienced this stutter-fest.

That’s already been done (me) and still this issue has occurred. It should be stressed that this does not occur every flight and occurs when performance is butter smooth. Then snap it happens. Like said above. We’ve got something else going on.  

How Should a I7-4770K HT On be at best?

And what About FSL Policy? In 5.2 i used ProcessLasso 0,2,3,4,5,6,7 for P3D when using FSL. And 0-7(all) when using PMDG birds as it was smoother. No ideer why.

Thanks Michael Moe

 

3 minutes ago, micstatic said:

That’s already been done (me) and still this issue has occurred. It should be stressed that this does not occur every flight and occurs when performance is butter smooth. Then snap it happens. Like said above. We’ve got something else going on.  

Have you checked the Performance tab on Task Manager and noted the GPU load? 100%?

22 minutes ago, Ray Proudfoot said:

If you're getting 60fps that suggests you're hittting your graphics card pretty hard. You don't list any kit in your signature unfortunately.

I have a 30Hz monitor and with VSync enabled my fps is of course limited to 30. Flying west over Russia and close to the Belarus border enroute to Paris my GPU has a 5% load at FL380 in the PMDG737.

Try limiting your refresh rate to 30Hz via NCP and see if that helps. I've not experienced this stutter-fest.

Hi Ray,

sorry I was not talking about myself specifically, since I do use 30 FPS, yes. My GPU is going easy during cruise.
As I said, nothing changed to prior to the hotfix except that this has happened to me now for the very first time; others seem to be effected more often.

Ah and my specs: CPU Ryzen 3600, GPU 1080ti, 32 GB Ram.

It happened to me on a clean v3 hf1 install and default prepar3d.cfg first run. I had EA on with volumetric clouds enabled. ASP3D was off. While it was stuttering like crazy I saw how my GPU would go from around 40% load to 0% load.

On my previous P3D v3 I experienced these stutters as well with the PMDG 747.

5 minutes ago, Samaritano said:

It happened to me on a clean v3 hf1 install and default prepar3d.cfg first run. I had EA on with volumetric clouds enabled. ASP3D was off. While it was stuttering like crazy I saw how my GPU would go from around 40% load to 0% load.

On my previous P3D v3 I experienced these stutters as well with the PMDG 747.

Thats the same what I saw reg. GPU load. Load was about 80-90% and FPS are good. FPS drops and stutters starts and the GPU load decreaased too.

11 minutes ago, Flyinggok said:

Thats the same what I saw reg. GPU load. Load was about 80-90% and FPS are good. FPS drops and stutters starts and the GPU load decreaased too.

Sounds like it still is an issue with 5.3. PMDG, FSL or default aircraft?

I  havent had it yet with HF1.1

Michael Moe 

1 hour ago, Michael Moe said:

How Should a I7-4770K HT On be at best?

And what About FSL Policy? In 5.2 i used ProcessLasso 0,2,3,4,5,6,7 for P3D when using FSL. And 0-7(all) when using PMDG birds as it was smoother. No ideer why.

Thanks Michael Moe

 

its like this in my cfg. (but still using ProcessLasso as above)

[JobScheduler]
AffinityMask=253
P3DCoreAffinityMask=253
MainThreadScheduler=0
RenderThreadScheduler=2
FrameWorkerThreadScheduler=4

 

Michael Moe

@Rob_Ainscough

I have a 5950X with SMT Enabled according to Ryzen Master CCD0 : Core 2 is best and core 7 second  and on CCD1 Core 10 is best

 

So I set:

AffinityMask=4294967295  = Use ALL Cores
P3DCoreAffinityMask=4294434807 = Use all but don't use SMT on the best cores
MainThreadScheduler=2 
RenderThreadScheduler=9
FrameWorkerThreadScheduler=6

 

I see in Ryzen Master that my Physical Core 2 is now used nearly 100 %  and physical Core 7 and Core 10 are "used" but I´m not sure if this is from other workload.

So it "looks" correct .

But when I understood your explanation starting always with 0 for the first physical core then the correct setting would be:

MainThreadScheduler=1 = First Physical Core based on starting with 0
RenderThreadScheduler=9 = Number 10 Physical Core based on starting with 0
FrameWorkerThreadScheduler=6 = Number 7 Physical Core Based on starting with 0

But when I use the above configuration the physical Core 1  is used the most instead of the physical Core 2? 

Thanks for your help.

 

Update 🙂 I think i have it: 

I have to "skip" the masked cores 🙂  And take the HT into account:

AffinityMask=4294967295
P3DCoreAffinityMask=4294434807
MainThreadScheduler=2
RenderThreadScheduler=16
FrameWorkerThreadScheduler=11

Now i see the load distribution I was expecting 🙂

 

Edited December 27, 20214 yr by 331BK