On 6/17/2017 at 2:45 AM, Anders Bermann said:

Terribly sorry but I'm a little confused with this topic...

I don't have an Affinity Mask in my config. Should I add one? And what should I add? (I have a Core i7 6700HQ)

Hi Anders:

If you go to the Windows Start Menu, under the Lockheed Martin group click "Learning Center."  Go to Prepar3D + Getting Help + Advanced Configuration, and various additional tuning options are described.  What we are discussing is under the [JobScheduler] heading (scroll down in the Help file to see the description); it does not normally exist in the Prepar3D.cfg file, and thus the default value when it is not present is 14 (use all available cores).  If you wish to set specific cores for P3D to run on in a multi-threaded system, you can add a [JobScheduler] section to the cfg file and add the AffinityMask entry with the value you desire; Steve described various options depending on which cores you want to run on, whether you have HT enabled, etc.

On my 6700k OC system, I added:

 

[JOBSCHEDULER]

AffinityMask=85

 

to the end of my Prepar3D.cfg file for best performance.  This limits P3D to run threads on the 4 physical cores of the 6700k only, and not to run any threads on the virtual cores.  I found this worked best for my system rather than letting the Windows job scheduler free reign to use all cores for P3D.  I could go further to specify affinity masks also for my add-ons so they do not conflict with the main P3D thread, but I haven't found the need at this point on my system.

If you are happy with the way P3D is running on your system, this is not necessary.  But it may be helpful depending on what add-ons you have and what microprocessor you are running. YMMV.

I find that an AM=245 is what works the best for me and my 7700k

I installed P3Dv4 onto a drive that at one time had P3Dv3 on it. With v3 the sim was very smooth and I had no issues at all. I formatted the drive using default format setting.  When I installed v4, it had terrible stutters all the time, every 2-3 seconds would get a micro-stutter.  I tried everything I could think of, searched the forums, and tried every trick in the book, but nothing fixed it.  I remember reading the P3D installation guide about changing the "allocation unit size" when formatting the hard drive, so I thought I'd try it.  When I selected the default setting, the allocation unit size was set to 1024 bytes.  I changed it to 64k bytes, formatted with this setting, installed P3Dv4, and all of my stutters are gone.  Now it flies smooth as butter.

Basic maybe, but you might consider defragging the drive with a professional defrag program if you are not using a SSD and haven't done it in a while. Did wonders for me after all the installation of new P3Dv4 stuff. Forgot about defrag and drive had gotten to 20% fragmented.

In my case the solution was to enable HT, Vsync (no triple buffering). No tweaks whatsoever in Prepar3dv4.cfg. One may get micro stutters if Autosave is enabled. 

On 17.6.2017 at 1:57 AM, SteveW said:

When we have CPU containing a great many cores we still might not want the sim to break into more than four or six parts and so we will start to use AMs rather more knowledgably by th

Thanks Steve for stepping in and explaining it all (again).

I get good results now with NCP refresh rate 59 and AM 85.

For extra smoothness I dropped the shadow quality from ultra to high. 

 

Cheers
bs

Check out my post on P3D VSync and performance settings:

 

Is this the kind of microstutters we're talking about?

 

Stutters are "long frames" ... in a perfect world with 60Hz refresh, you'll want all frames to be rendered exactly at 16ms, for 30Hz you want all frames to rendered at exactly 33ms.  Stutters become apparent when you see significant variance in how long it takes to render a frame.

Unfortunately YouTube videos, one's ISP latency/bandwidth, and even the recording process itself can also introduce frame skipping, this is a different source of what might look like "stutters".

The best way to analyze stutters and discuss stutters is with a tool that logs timeframe data, one such tool I use is FRAPS (not for recording video but for recording timeframe data) ... I'm sure there are other tools also, but FRAPS has worked well for me.

These results are going way back to P3D V3 (not V4) when I was testing timeframes:

Should be rather obvious the stutter mark at 88ms to produce one frame, then 45ms next frame and 37ms next frame ... this slow then fast variance in rendering frames makes a "stutter" (eye sees a significant difference is how long a frame took to render).

With VSynch OFF and Unlimited you can see there is considerable variance.  

Now with VSynch ON 30Hz Unlimited, you can see the overall collection of timeframe points are more consistently grouped.

The objective should be to come up with graphics settings that minimize the number of "long frames" ... I can guarantee you this is never going to be Max, Max, Max, Max, all add-ons, kitchen sink, and large bottle of vodka.  Not now, not in the future, never ... the key to less long frames will be:

1.  Graphics settings

2.  Add-ons used and where you fly (rural or major cities)

3.  AffinityMask has some influence and will vary by CPU and other hardware in one's PC

If you want to use Dynamic Lights, then  you'll need to drop your AA settings.  I use 2XSSAA at night and 4XSSAA during the day on an overclocked TitanXP.  Building Shadows, SimObject Shadows, Terrain Shadows combined with AutoGen settings will also have a significant roll in "long frames". 

I have the following graphics options based on the type of flying I'm doing and my location and time of day/night:

CommercialCityDay
CommercialCityNight
CommercialRuralDay
CommercialRuralNight
GACityDay
GACityNight
GARuralDay
GARuralNight

Commercial means complex aircraft (PMDG, etc.)

GA means lightweight aircraft (Carenado, Alabeo, etc.)

In all cases I use TEXTURE_SIZE_EXP=10 (should be used ONLY with 1080Ti or TitanXP or anything with 11-12GB VRAM)

Since I like Dynamic Lights, all my night settings will have DL enabled and 2XSSAA and most shadows turned off with the exception of Internal/External vehicle and Clouds.  I can share my settings if anyone is interested, they work well for what I do but keep in mind my goal is 30 FPS since my monitor refresh is 30Hz at 4K res.

Cheers, Rob.

@Rob Ainscough

Thank you for explaining this issue in-depth. I'll do some more testing and see which configuration works best for me.

19 minutes ago, Rob Ainscough said:

In all cases I use TEXTURE_SIZE_EXP=10 (should be used ONLY with 1080Ti or TitanXP or anything with 11-12GB VRAM)

Cheers, Rob.

Have you done any testing on this setting specifically, Rob? In terms of overall performance hit and whether it effects the CPU as well? I have an 1080 Ti and when I used the setting, it seemed like the CPU worked a lot harder and caused other terrain rendering tasks to become backed up slightly. Though that's anecdotal at best. Would love to see some actual testing done on it -- Or at least, what's your general experience with the various steps?

40 minutes ago, Rob Ainscough said:

In all cases I use TEXTURE_SIZE_EXP=10 (should be used ONLY with 1080Ti or TitanXP or anything with 11-12GB VRAM)

Hey Rob-

Does that texture line apply to photo scenery? I can't find anything on it.

Thanks,

Mark

 

40 minutes ago, Rob Ainscough said:

 

 

1 hour ago, Sethos1988 said:

Would love to see some actual testing done on it -- Or at least, what's your general experience with the various steps?

I'm working on my new web site (http://www.p3dwiki.com) specifically for P3D V4 ... it's taking me some time to organize my data/info, but I hope to have something going by July "ish".

1 hour ago, newtie said:

Does that texture line apply to photo scenery?

It's supposed to yes, I had some issues with it and PR ... LM suggested the PR source needs to be reworked for higher resolution.  I don't have the source PR so I can't confirm or deny.  If using TEXTURE_SIZE_EXP=10 and you fly fast and low, then you'll likely hit issues with terrain and AG not loading fast enough but the actual setting doesn't seem to cause any FPS degradation (just might cause blurries at higher speeds).

Cheers, Rob.

Thanks, I'll give it a try.

 

Cheers,

Mark