Reading the thread 5.2 seems to be hitting CPU1 harder than 5.1 and the discussion around affinity mask. I am not looking for a solution but more of information, education answer if there is one. I never had an issue until now and I was under the impression P3d no longer required affinity mask tweaks.

My P3d V5.2 the CPU 0 is running at 100% at all times, and I am getting stutters. If I change the Affinity mask and remove CPU0 from the equasion the thread moves to CPU 7 100%

Now here is the thing, If I use the old Affinity mask trick , task manager right click change affinity click any box then click it back the threads seem to run equally over all CPU’s and it also removes the stutters, a win for me.

The question is what is the difference from changing affinity mask in the p3d cfg file, process lasso and using the old task manager trick? Why different results?

Thanks

Edited June 27, 20214 yr by Ray Proudfoot
Standard text used

@SteveW is more qualified than me to speak about this. My own experience suggests something changed in v5.2 that pushed more work onto core 0 reducing the load on the GPU.

You don’t mention your CPU or whether Hyperthreading is enabled in your BIOS. Without that info it’s hard to comment on your experience.

In short, with Hyperthreading enabled an AM is essential. If it isn’t then it may not be but since I have no experience of running without Hyperthreading I can’t really help.

This discussion is more relevant to the topic where all discussion has taken place so I’ll merge this into that one.

1 hour ago, Dirk98 said:

I noticed SteveW used 28 so I decided to try (and I recall there was a discussion about 28 to 31 fps some time ago), no jitters so far, but maybe a snake oil. I should have better suggested 30 fps here perhaps.

The idea with 31, I am using 33, is monitor at 30 hertz Vsync Uimited and NVCP limiter set to 31 to avoid framerate spikes…

So limit just 5-10% above the monitor refreshrate.

 

1 hour ago, dabba said:

Now here is the thing, If I use the old Affinity mask trick , task manager right click change affinity click any box then click it back the threads seem to run equally over all CPU’s and it also removes the stutters, a win for me.

The question is what is the difference from changing affinity mask in the p3d cfg file, process lasso and using the old task manager trick? Why different results?

Thanks

Be careful with that trick. It may initially reduce stutters, but I had instances of very slow autogen loading during approaches at airports. This is something that you can't see right away.

5 hours ago, GSalden said:

The idea with 31, I am using 33, is monitor at 30 hertz Vsync Uimited and NVCP limiter set to 31 to avoid framerate spikes…

So limit just 5-10% above the monitor refreshrate.

 

I'll try that!! 

Thanks

7 hours ago, dabba said:

The question is what is the difference from changing affinity mask in the p3d cfg file, process lasso and using the old task manager trick? Why different results?

As best I know, the "old trick" which is applied after P3D has loaded, removes whatever CPU priorities that LM intended, which you do at your peril. 😉

If you set the affinity mask up front, you still allow LM to assign whatever priorities they want, but to a reduced number pf possible cores.

Maybe I'm the odd man out here, but did anyone ever think that core utilization, however that parses out, is the way that LM designed it and it's normal?

Just sayin'.

Edited June 27, 20214 yr by newtie
sp

9 hours ago, dabba said:

I never had an issue until now and I was under the impression P3d no longer required affinity mask tweaks.

It depends on whether Hyperthreading is enabled in your BIOS. It is on my computer and this test demonstrates why using all cores is a very bad idea.

Scenario 1. No AM set in prepar3d.cfg so all 12 cores of my CPU are used by P3D v5.2. Result? A horrible stuttering mess with the two virtual processors on core 1 being bombarded with instructions. cpu0 stuck on 100% and the overflow going onto cpu1. Unflyable and fps often lower than 30.

Scenario 2. AF of 1365 set in prepar3d.cfg. This assigns only VPs 0 2 4 6 8 and 10 for use by P3D. Result? A smooth flying experience with the sim transformed. 30fps can be maintained at all but the most demanding airports and even at those the drop from 30 is brief. Even when cpu0 hits 100% the sim remains fluid.

So the AM is definitely required for any system with Hyperthreading enabled. Rather than a tweak I would describe it as a mandatory requirement.

43 minutes ago, newtie said:

Maybe I'm the odd man out here, but did anyone ever think that core utilization, however that parses out, is the way that LM designed it and it's normal?

Just sayin'.

With my AM setting I see sim activity on cores 0 and 1 with scenery loading on cores 2 6 and 10. Not sure what 4 and 8 are used for as activity is quite low.

Interesting observation if someone could please test.

Installed an AM into my p3d.config.

Whenever I add/install a new paint for FSLABS, I always get a box that says "restart P3D" after I try to open P3D.

I opened my p3d.config today to look at something and noticed a ";" in front of the AM line.

Does FSLABS turn-off/disable AM settings?

Thanks.

@Dirk98

 

Hallo Dirk,

thanks for your settings.

Frank from EDDM

 

2 hours ago, Dreamflight767 said:

Does FSLABS turn-off/disable AM settings?

You need to ask on their forum but I’d be surprised if they did.

8 hours ago, Ray Proudfoot said:

It depends on whether Hyperthreading is enabled in your BIOS. It is on my computer and this test demonstrates why using all cores is a very bad idea.

Scenario 1. No AM set in prepar3d.cfg so all 12 cores of my CPU are used by P3D v5.2. Result? A horrible stuttering mess with the two virtual processors on core 1 being bombarded with instructions. cpu0 stuck on 100% and the overflow going onto cpu1. Unflyable and fps often lower than 30.

Scenario 2. AF of 1365 set in prepar3d.cfg. This assigns only VPs 0 2 4 6 8 and 10 for use by P3D. Result? A smooth flying experience with the sim transformed. 30fps can be maintained at all but the most demanding airports and even at those the drop from 30 is brief. Even when cpu0 hits 100% the sim remains fluid.

So the AM is definitely required for any system with Hyperthreading enabled. Rather than a tweak I would describe it as a mandatory requirement.

Ray, do you have all 12 cores / 24 VPs running when you stand on the ground with your airplane with HT on?  In my case, I only have VP 0 loaded, while the others are more or less idle. Once I am airborne, the load on other VPs represents a zig-zag shape on the load-time, and my understanding is that these VPs are busy loading the scenery during my flight, which is a good thing. 

I am using v4.5, so maybe things are different in v5.2?

6 minutes ago, Afterburner said:

Ray, do you have all 12 cores / 24 VPs running when you stand on the ground with your airplane with HT on?  In my case, I only have VP 0 loaded, while the others are more or less idle. Once I am airborne, the load on other VPs represents a zig-zag shape on the load-time, and my understanding is that these VPs are busy loading the scenery during my flight, which is a good thing. 

I am using v4.5, so maybe things are different in v5.2?

That's a typo by me. I have a 6 core, 12 VP CPU. Just loaded P3Dv5.2 at EGLC and 0 remains at 100% with 2, 6 and 10 heavily used for loading scenery then drop to near idle once it's finished. That's at 60Hz.

Same load at 30Hz and 0 averaging in the 90s. Same readings for the other cores.

Once airborne (in slew mode) 2, 6 an 10 are used for scenery loading as you suggest. They're up and down. Soon as I pause the sim they drop to zero.

 

21 hours ago, Afterburner said:

@Ray Proudfoot Could you perform the test if you have time?  Set your screen refresh rate to 60 Hz, set the internal fps limit to 30 fps, and make sure that you add the following under the [MAIN] section of your p3d.cfg before starting the sim:

FIBER_FRAME_TIME_FRACTION=0.01

This ensures that you don't have an unnecessary loss in fps from limiting the frame rate internally. Also make sure that vertical sync in Nvidia Control Panel is set to "Use the 3D application setting". Do a test, and tell us how the smoothness feels compared to your standard configuration. That's all!

I've just run this test and compared it to my 30Hz setting. With 60Hz and an internal frame rate limiter set plus the FIBER entry 30fps was maintained throughout my circuit around EGCC using UK2000 Manchester in a Carenado PC12. But there were lots of micro-stutters and it was definitely not as fluid as 30Hz. The fact that core0 stayed on 100% throughout was probably the reason.

In contrast, flying the same circuit with 30Hz and unlimited and no FIBER entry the core0 averaged in the high 50s / low 60s. 30fps was maintained of course and it felt a lot smoother especially during turns but straight line flying was also as good as it gets. Smooth as a baby's bottom! 😁

That suggests a lot of work is being done internally to maintain 30fps rather than letting the hand brake off in P3D (unlimited) and controlling the fps externally with the monitor refresh rate.

I've turned TripleBuffering off as an experiment and haven't noticed any deterioration in performance. If you're in the market for a new monitor I really would recommend one with 30Hz.