Jump to content
Sign in to follow this  
Mejjo

P3D 5.3 Hotfix 2 released

Recommended Posts

When I'm running the sim I am not seeing core zero being hit by system processes, another core is though.


Steve Waite: Engineer at codelegend.com

Share this post


Link to post
2 hours ago, SteveW said:

When I'm running the sim I am not seeing core zero being hit by system processes, another core is though.

Hi Steve,

I see the same. Although as we discussed on my i9-9900ks with HT enabled I don’t mask off the first 3 HT logical processors. I do assign the 3 primary threads to 0, 2 and 4 respectively but as we discussed I use Process Lasso to move everything else other than Windows functions to the last two cores logical processors 12-15. Additionally I leave the last core free of P3D thinking the op system would take that one but I find those last two LPs very under utilized and the Frame Worker thread core ( LP 4 ) and it’s open LP 5 the most utilized. I do run a native 30 HZ TV so I run unlimited and Vsync on in the sim and use NCP to cap frames at 30 which reduces my Main Thread load on LP 0 quite well. Since I don’t see near 100% loading I am not currently concerned about the core sharing going on with the first three cores and the three primary P3D threads being assigned there. Although the Frame Worker thread and that 3rd core are being worked the most consistently.

I will experiment today with not leaving the last core open for the operating system and leave everything else the same including my confining the addons to the last 2 cores and see if any loading particularly the third core changes.

Again thanks for all your support!
 

Joe

  • Like 1

Joe (Southern California)

SystemI9-9900KS @5.1Ghz/ Corsair H115i / Gigabyte A-390 Master / EVGA RTX 2080 Ti FTW3 Hybrid w 11Gb / Trident 32Gb DDR4-3200 C14 / Evo 970 2Tb M.2 / Samsung 40inch TV 40ku6300 4K w/ Native 30 hz capability  / Corsair AX850 PS / VKB Gunfighter Pro / Virpil MongoosT-50 Throttle / MFG Crosswind Pedals /   LINDA, VoiceAttack, ChasePlane, AIG AI, MCE, FFTF, Pilot2ATC, HP Reverb G2

Share this post


Link to post
44 minutes ago, joepoway said:

with HT enabled I don’t mask off the first 3 HT logical processors. I do assign the 3 primary threads to 0, 2 and 4 respectively

Hi Joe,

So long as the P3DCoreAffinityMask only allows one LP per core for the main tasks it will avoid scenery gathering tasks to run on them which is one main cause of stutter. It is logical todo that with AffinityMask as well otherwise you might still get unwanted activity from sim related tasks, there's no reason not to.

What JobScheduler entries are you using, I shall test it here...

Steve


Steve Waite: Engineer at codelegend.com

Share this post


Link to post
16 hours ago, SteveW said:

Did you say it happened before? if so do you remember anything about that?

Not really like this, the only few problems I had were at first V5 version with EA, but got fixed by LM long time ago... The last V5.2 worked for me perfectly and that's what it seems to be strange.
Yesterday I tried to disable the variable refresh rate (VRR) and HAGS on W10 settings and it seems to make things a little bit better, but it could be "placebo effect", further testing will be needed.

Edited by Alexx Pilot
  • Like 1

Share this post


Link to post
11 minutes ago, Alexx Pilot said:

Not really like this, the only few problems I had were at first V5 version with EA, but got fixed by LM long time ago... The last V5.2 worked for me perfectly and that's what it seems to be strange.
Yesterday I tried to disable the variable refresh rate (VRR) and HAGS on W10 settings and it seems to make things a little bit better, but it could be "placebo effect", further testing will be needed.

Yes maybe that's something. The refresh rates differing to the simulator frame output timing can make it appear a little jittery, but in your video it seems more than a little. It's certainly an odd problem to understand without hands on the PC.

 

Guys, any ideas that may help Alexx Pilot with this?

  • Like 1

Steve Waite: Engineer at codelegend.com

Share this post


Link to post

@SteveW I really appreciate your help, thank you so much for that 🙂 
I just need some time to test in every possible way and I will update if I find the problem, so maybe could help other people too having the same issue.

  • Like 1

Share this post


Link to post
11 minutes ago, Alexx Pilot said:

@SteveW I really appreciate your help, thank you so much for that 🙂 
I just need some time to test in every possible way and I will update if I find the problem, so maybe could help other people too having the same issue.

You're welcome! Sorry I could not think of something to help cure it so far. I don't think you can do better on the CPU side, maybe something to do with the monitor/refresh/etc.it is possible that something changed in v5.3 HF2. I will have time for another session trying to recreate the problem later.

  • Like 1

Steve Waite: Engineer at codelegend.com

Share this post


Link to post
48 minutes ago, Alexx Pilot said:

I just need some time to test in every possible way and I will update if I find the problem

Can you try setting your monitor refresh rate to its lowest Hz? Then in P3D set VSYNC=ON and Unlocked on the slider in P3D Display Settings. That should always be be very smooth. If not maybe there is something else going on other than monitor issues.

  • Like 1

Steve Waite: Engineer at codelegend.com

Share this post


Link to post
1 hour ago, SteveW said:

Hi Joe,

So long as the P3DCoreAffinityMask only allows one LP per core for the main tasks it will avoid scenery gathering tasks to run on them which is one main cause of stutter. It is logical todo that with AffinityMask as well otherwise you might still get unwanted activity from sim related tasks, there's no reason not to.

What JobScheduler entries are you using, I shall test it here...

Steve

Hi Steve,

I'm currently running:

[JobScheduler]
AffinityMask=65535
P3DCoreAffinityMask=16383
MainThreadScheduler=0
RenderThreadScheduler=2
FrameWorkerThreadScheduler=4

Basically leaving the last core i.e. LP's 0-15 for the Operating System. Remember I also confine virtually all other apps to the last 2 cores i.e. LP's 12-15 with Process Lasso so basically only P3D is using the first 6 cores and sharing the 7th core with addons that didn't grab the last core. I could be wrong but by leaving the HT LP's on the first 3 cores open and since there are virtually no other apps using them doesn't that let P3D take advantage of LP 1, 3 and 5  for scenery gathering tasks. I think I understand the additional issue of the sharing of internal microarchitecture components called execution units from the link you posted but wouldn't that also only come into play if loads were approaching 80%+ ?

Also since I take advantage of my 30Hz TV with Unlimited/Vsync on in sim and 30 fps cap with NCP, LP0 (and LP 1 ) are no where near 100% neither are LP2 or LP4 however my LP4 the FrameWorker seems most tasked. 

If I were to mask off LP1, 3 and 5 using 65493 are you saying the prevention of any scenery gathering tasks on those 3 cores sharing the 3 main threads would be better even if the sharing isn't coming close to maxing out the cores? I saw you were experimenting running the Render and Frame threads on LP2 and LP3 on the same core wouldn't that even be worse than sharing a core with background scenery gathering?

My next experiment will be to not leave the last core free for the Operating system by using:

[JobScheduler]
AffinityMask=65535
P3DCoreAffinityMask=65535
MainThreadScheduler=0
RenderThreadScheduler=2
FrameWorkerThreadScheduler=4

and of course keeping all the same confinements to the last two cores LP's 12-15 aggressively using Process Lasso. 

Again I'm thinking, by so aggressively using Process Lasso to keep everything off the first 6 cores LP's 0-11 it may negate the need to mask off LP 1, 3 and 5 but I defer to your wisdom in these matters for sure.

Thanks again,

Joe

 


Joe (Southern California)

SystemI9-9900KS @5.1Ghz/ Corsair H115i / Gigabyte A-390 Master / EVGA RTX 2080 Ti FTW3 Hybrid w 11Gb / Trident 32Gb DDR4-3200 C14 / Evo 970 2Tb M.2 / Samsung 40inch TV 40ku6300 4K w/ Native 30 hz capability  / Corsair AX850 PS / VKB Gunfighter Pro / Virpil MongoosT-50 Throttle / MFG Crosswind Pedals /   LINDA, VoiceAttack, ChasePlane, AIG AI, MCE, FFTF, Pilot2ATC, HP Reverb G2

Share this post


Link to post
3 minutes ago, joepoway said:

Hi Steve,

I'm currently running:

[JobScheduler]
AffinityMask=65535
P3DCoreAffinityMask=16383
MainThreadScheduler=0
RenderThreadScheduler=2
FrameWorkerThreadScheduler=4

 

That's not good, during the flight I have main task on LP0 and a scenery gatherer on LP1 (take a look in Task Manager graphs, LP view, and not activity on both LPs of core zero whci is what you are trying to avoid) and stutter. The AMs you are using completely misses the major point of avoiding sharing the main cores with scenery gathering. Corrallng other apps away from those main cores is perfect.

Assuming 8 core HT enabled 16LPs I would go to:

[JobScheduler]
AffinityMask=65493
P3DCoreAffinityMask=65493
MainThreadScheduler=0
RenderThreadScheduler=1
FrameWorkerThreadScheduler=2
 

07,06,05,04,03,02,01,00=core number
11,11,11,11,11,01,01,01=AffinityMask = 65493
11,11,11,11,11,01,01,01=P3DCoreAffinityMask = 65493
00,00,00,00,00,00,00,01=MainThreadScheduler = 0
00,00,00,00,00,00,01,00=RenderThreadScheduler = 2
00,00,00,00,00,01,00,00=FrameWorkerThreadScheduler = 4
 

See the bold zeros, that prevents unwanted activity on the sister LPs of the main tasks - this is critical to good performance.


Steve Waite: Engineer at codelegend.com

Share this post


Link to post

...you have to be in flight so that the sim is gathering scenery. When the sim is at rest the scenery gatherers are also at rest.


Steve Waite: Engineer at codelegend.com

Share this post


Link to post
20 minutes ago, SteveW said:

See the bold zeros, that prevents unwanted activity on the sister LPs of the main tasks - this is critical to good performance.

Hi Steve,

I do run test flights and I have and do fully understand the point you are making and how LP's are masked or not and all the order sequencing from your previous posts and our discussions.

Perhaps because I haven't monitored the graphs, which I do know how to do I am missing what you are seeing. I have been monitoring the load numbers real-time and logging highs, lows and averages with HWiNFO64. I will monitor the Task Manager Graphs as you suggest. Ironically even doing it the way I have been without masking those LP's the sim is running better and smoother then I have ever seen it.

Again thanks and know that I do understand and appreciate your point and perhaps my data gathering source during experiments has been flawed and/or limited 😒.

Joe 

Edited by joepoway

Joe (Southern California)

SystemI9-9900KS @5.1Ghz/ Corsair H115i / Gigabyte A-390 Master / EVGA RTX 2080 Ti FTW3 Hybrid w 11Gb / Trident 32Gb DDR4-3200 C14 / Evo 970 2Tb M.2 / Samsung 40inch TV 40ku6300 4K w/ Native 30 hz capability  / Corsair AX850 PS / VKB Gunfighter Pro / Virpil MongoosT-50 Throttle / MFG Crosswind Pedals /   LINDA, VoiceAttack, ChasePlane, AIG AI, MCE, FFTF, Pilot2ATC, HP Reverb G2

Share this post


Link to post
13 minutes ago, SteveW said:

Assuming 8 core HT enabled 16LPs I would go to:

[JobScheduler]
AffinityMask=65493
P3DCoreAffinityMask=65493
MainThreadScheduler=0
RenderThreadScheduler=1
FrameWorkerThreadScheduler=2
 

07,06,05,04,03,02,01,00=core number
11,11,11,11,11,01,01,01=AffinityMask = 65493
11,11,11,11,11,01,01,01=P3DCoreAffinityMask = 65493
00,00,00,00,00,00,00,01=MainThreadScheduler = 0
00,00,00,00,00,00,01,00=RenderThreadScheduler = 2
00,00,00,00,00,01,00,00=FrameWorkerThreadScheduler = 4
 

I think you meant:

07,06,05,04,03,02,01,00=core number
11,11,11,11,11,01,01,01=AffinityMask = 65493
11,11,11,11,11,01,01,01=P3DCoreAffinityMask = 65493
00,00,00,00,00,00,00,01=MainThreadScheduler = 0
00,00,00,00,00,00,01,00=RenderThreadScheduler = 1
00,00,00,00,00,01,00,00=FrameWorkerThreadScheduler = 2
 

since LP1, 3 and 5 are masked off obviously a cut and paste typo I presume

Joe

  • Upvote 1

Joe (Southern California)

SystemI9-9900KS @5.1Ghz/ Corsair H115i / Gigabyte A-390 Master / EVGA RTX 2080 Ti FTW3 Hybrid w 11Gb / Trident 32Gb DDR4-3200 C14 / Evo 970 2Tb M.2 / Samsung 40inch TV 40ku6300 4K w/ Native 30 hz capability  / Corsair AX850 PS / VKB Gunfighter Pro / Virpil MongoosT-50 Throttle / MFG Crosswind Pedals /   LINDA, VoiceAttack, ChasePlane, AIG AI, MCE, FFTF, Pilot2ATC, HP Reverb G2

Share this post


Link to post
58 minutes ago, joepoway said:

Also since I take advantage of my 30Hz TV with Unlimited/Vsync on in sim and 30 fps cap with NCP, LP0 (and LP 1 ) are no where near 100% neither are LP2 or LP4 however my LP4 the FrameWorker seems most tasked. 

Why do you cap the fps to 30 with NCP in addition to running unlimited/Vsync on a 30 Hz TV?  The latter already does the trick of delivering smooth 30 fps with a core0 load that adjusts to the complexity of the situation. Further limiting with NCP is not needed (or counterproductive in the worst case).

  • Like 1
  • Upvote 1

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  
  • Tom Allensworth,
    Founder of AVSIM Online


  • Flight Simulation's Premier Resource!

    AVSIM is a free service to the flight simulation community. AVSIM is staffed completely by volunteers and all funds donated to AVSIM go directly back to supporting the community. Your donation here helps to pay our bandwidth costs, emergency funding, and other general costs that crop up from time to time. Thank you for your support!

    Click here for more information and to see all donations year to date.
×
×
  • Create New...