For those of you showing little activity on your other cores ... is this while you flying thru somewhat populated scenery or just sitting on the runway at some airport?

 

With HT OFF all 8 physical cores are showing 70-75% with core 0 showing a little more at 80-85% and GPU usage is around 60-80% pending weather/scenery.

 

So I'm a little puzzled why some of you are showing so much activity on one core and so little on the others?

 

Cheers, Rob.

I'm a little puzzled why some of you are showing so much activity on one core and so little on the others?

With an Intel HT pair, one showing 90% and the other showing 1% is a core running at 91%.

 

There's one main thread tries to run as fast as possible, unless on a powerful enough system perhaps does not have need to run at maximum throughput for the settings chosen. The other threads start up on subsequent logical processors, and can be thought of as data gatherers supplying the main thread. A third way to encumber the main thread, is to provide too many data gathering threads. Often for example with an outside view panning the aircraft at the ramp might only show one thread at or near max and the others around 25% or lower. It's because they don't have so much to do at that time, but probably they will creep up at some other stage of the flight. With your setup Rob if all the threads are quite busy yet unmaxed, seems fine. But would make a poor example when demonstrating the pitfall of HT with P3D/FSX.

Ran a quick test just now at LIRF and it's not as dense as some scenery areas so probably why I wasn't seeing 70% range on the other cores, but still relatively even distribution with a spike when I activate mouse look.  But as you can see, need to get some motion going so the terrain system gets working ... my concern/question was are people testing core usage when just sitting on the runway and not moving?  

 

Not really trying to demonstrate the pitfall of HT ... just with HT OFF one gets a better idea of thread spread (not suggesting it is the better option for all).  HT ON will vary from CPU types as Intel continually change how HT works ... which I think is why we see so much variance in some of the testing.

 

 

Graphics settings and add-on used are in video description.

 

With HT OFF Not really sure why core 1 (base ref 0 to 7) wasn't working as hard?

 

Cheers, Rob.

With HT OFF Not really sure why core 1 (base ref 0 to 7) wasn't working as hard?

core 1 is hosting the second main thread which is more or less a thread pool doing the least work in wait states for the others. Although on less cores/LPs that thread might be combined with another, demonstrating what P3D does with respect to handling threads. But it does not handle the HT enabled pitfall of putting that thread on the second LP to share the core with the main thread.

I tried AF setting for my cpu and posted a pic of the cores running a page or two back. I was not getting the results I had hoped for so in the config I removed the number but left the  [JOBSCHEDULER]  line in place and then things  picked up.

 

Before with the number 1364 (for an i7 5930k) in the config my graphic was showing one core doing 90 percent of the work and all the rest doing very little but after removing that number and leaving nothing after the [JOBSCHEDULAR]  insert it now shows one core doing 97, one core doing 70 and the rest all doing 40 to 50. 

 

So for me no AF tweak works much better. 

No, if I recall your graphic clearly showed the first logical processor doing all the work which is correct. If you've got LP 0 and LP 1 showing a lot of activity, that means you've shared core zero unwisely.

Unwise or not I am now getting 5 fps more without the tweak so I am leaving it at that. 

you changed nothing else?

No just the AF so basically I got back the fps that I had before trying the tweak, now all my cores are active and I can tell my cpu is working harder as well because it is running a bit hotter from around 45 to 50c with tweak and now at around 55 to 62c without the tweak. 

Your graphic showed 6 pairs of LPs in 6 columns with core 1 second in from left at 94% sharing its other LP (LP3) ideally running basically nothing. That's just one bit right. You could perhaps utilise the CPU more effectively with an AM=4084=(11,11,11,11,01,00).

I might try that but almost afraid to upset this setting. But I do love to experiment. 

Thanks

you can't hurt anything by trying. Key to HT is so long as you got HT enabled you want a core to look like your 94% 0% pair in column 2 from the left on your graphic, was perfect. Maybe the problem is with the rest of your AM. So to allocate the whole CPU with maximum threads is AM=4094=(11,11,11,11,11,10) or 4093=(11,11,11,11,11,01). No AM=(11,11,11,11,11,11) so you can see the theme from your CPU monitor.

A couple of things that get repeatedly overlooked in these threads. A frame rate limit of 30 does absolutely nil if you are not able to maintain that under most circumstances. You should start with that but if you run at 25 FPS consistently then you have zero overhead left and you are subject to spikes and precieved stuttering. You should find your mean average FPS then set it to a couple of FPS lower so that the P3D is not overloaded. 

 

The way I look at this is similar to a PID loop. You do not want to be chasing your tail on an unattainable value. Things will be much more stable if the program is constantly able to attain a consistent FPS. 

 

The other thing that is not often considered is the refresh rate of the VC gauges. In FSX gauges were rendered at 18hz with XML . Of course someone could program C++ or whatever and not use XML... I'm not sure how these have evolved in P3D but their refresh rate isn't always directly tied to framerate. This is why people who use a simpit with NO vc get much better results in performace as they only render the scene

Might be totally misplaced but related to long frames or stutter/pause .

 

I have  onboard audio on a Z87-Pro motherboard related to RealtekHD audio , and everytime i update the drivers i manually have to set the quality to CD 2 channel quality , instead of default DVD quality . Both for headset,microphone and speakers. I believe it helps a bit actually. I use AC97 on front for headset and microphone

 

Back to your core HT ON /OFF  :wink:

 

I am happy with AM=15 and Vsync off for tubes like the 777 anyway . Not 100% perfect but much better than FSX etc etc.

 

Michael Moe

The way I look at this is similar to a PID loop. You do not want to be chasing your tail on an unattainable value. Things will be much more stable if the program is constantly able to attain a consistent FPS. 

 

Have to disagree

 

I get better performance with unlimited

 

I get 30-40 fps worst case, say landing at busy Fly Tampa YSSY, PMDG 777 in VC, ASN overcast enhance 90mile cloud, dense autogen, 4xsgss & msaa, 16xAF, MFAA on too - UT2 AI, Vatsim etc etc with all the eye candy on. LOD 6.5, high water/shadows orbx/rex etc

 

I can get 55-65 fps easily in the cruise typically

 

What is the point of locking frames to 30fps when for 90% of the flight i can easily average 60.

 

I can get 120fps in the flyby or spot views in cruise and 70-90 on ground

 

120+fps, looks excellent on my 1440p 144hz panel with G-Sync.

 

When the sim is loaded down with work, its loaded down for a duration, not a single stutter frame.

 

for me 90% of my flight at 60fps is better than 99.99% of my flight at 30fps

 

The Gsync works well down to about 40fps, 30-40fps and the G-Sync starts to break down and stutters can be seen, horizontal panning etc.

 

I absolutely prefer temporal resolution of 1080p or 1440p @ 60hz over spatial resolution of  4K@30hz, maybe its just my eyes, but its clearly better experience visually read immersion

 

With G-sync and modern mainstream computers i really think locking frames in this way is not making best use of your equipment

 

mileage of course will vary with what you are running and the rig under your desk, just don't think this mantra of lock at minimum that you can always achieve is relevant today.

 

-rcladin