http://www.gatwick-fsg.org.uk/affinitymask.aspx?SubMenuItem=hardware

 

The problem with Gatwick site is that it has been around awhile and is really concerned with FSX and P3DV1.

It assumes HT is of no value and the first two cores should always be disabled.

Despite the P3D link above these two assumptions are still very debatable when it comes to P3DV2.

 

gb.

 

Hi GB,

 

I have updated the affinity mask calculator to provide an expert mode, so that you can customise the affinity mask to use any logical cores and threads that you desire.

 

http://www.gatwick-fsg.org.uk/affinitymask.aspx?SubMenuItem=hardware

 

Regards,

 

 

Ted.

Thank you Ted . . . your new expert mod is most helpful!

Thank you!  Just starting with P3D.  Keeping FSX, but so far, really happy with P3D.

 

Chuck Nance

Hi GB,

 

I have updated the affinity mask calculator to provide an expert mode, so that you can customise the affinity mask to use any logical cores and threads that you desire.

 

http://www.gatwick-fsg.org.uk/affinitymask.aspx?SubMenuItem=hardware

 

Regards,

 

 

Ted.

 

Thanks Ted.

I can see the calculator page getting a lot more hits from now on.

 

gb.

This is very interesting....

 

I removed HT and set Am to 15 and CPU load went way up and GPU down. Amazingly Temps were also lower.

 

If I understand what LM are saying, CORE 0 is not necessarily "reserved" for Windows use. P3D comes with a programmed way of using AM and when I look at it, it is indeed AM84.

 

So the key here may be to enable CORE 0 for P3D but leave its thread "open" so we don't overload it. Again, it will be CPU-type-dependant.

 

No magic bullet. At the moment I am impressed with AM15 and HT Off, but a little concerned regarding the load on the CPU...anybody else got some more ideas?

There is no magical Affinity Mask that works for everyone.  Everything written online about Win 7/8 x64 is that the OS does an excellent job of scheduling tasks. By manually setting AMs which prevent cores from being used by specific apps, you are interfering with the built-in OS optimization of CPU resources. My advice is to not block P3d from using any cores (HT off in your case, AM=15) unless experimentation shows that a different AM removes stutters.

My advice is to not block P3d from using any cores (HT off in your case, AM=15) unless experimentation shows that a different AM removes stutters.

 

As we know, The Processor Affinity Mask sets the cores on which a users process may be enabled or disabled.

 

In addition, there is a Thread Affinity Mask that sets the core on which an individual thread within application may run, that is this is set by the processor itself, and the user has no choice.‌ If the user disabled a core required by a Thead Affinity Mask then Windows finds another core in order to keep the process running.

 

A consequence is that the developer may decide which core a thread should run on, only for the user to force it to use another . I suggest that's not what the devloper wants, especially as I seem to remember LM implying the Processor Affinity Mask won't be needed. (Any anyone remember that?)

 

EDIT

 

Beau Hollis (Rendering System Lead - Prepar3D® Team) on on: April 16, 2014 said:

 

 

Our default AM masked off the first core because some OS tasks happen on the main core and in the past we have seen more stable results masking it off provided you have a 4 or more cores to play with. As I said a little earlier in this thread, we are changing that for 2.3 to use all cores and will continue to look into paging performance as a whole.

 

At one point in time, Windows XP was the norm and the venerable OS did pile both system processes and drivers onto core 0. But that was also when CPUs were exciting, innovative technology if they had two physical cores.  

 

When P3d 2.x first came out, I did a lot of experimentation with the AM, as A) I noticed that core 0 was blocked by default and B ) my PC has 12 physical cores. I pretty much tried every permutation of AM, including forcing TrackIR and weather injectors (at the time OPUSFSX and FSGRW, and now ASN) to their own cores with no overlap with P3d. I also experimented a bit with HT on, although I generally run my system with HT off.

 

Other than getting a slight boost in FPS and a little reduction in the frequency of stutters when allowing the OS to let P3d use all cores, I saw no discernible difference with any AM settings. But that's just me.

 

But blocking P3d from using core 0 in a quad core system has got to be a bad thing and why LM thought that it wasn't is beyond my comprehension. That's why this forum is littered with posts about blurries with the default AM setting. At least LM has promised to change the default AM. That isn't going to stop some people from tinkering, though.

When I installed ORBX OpenLC Europe, I got very mixed results having 5 cores active out of 8 (4 physical cores, HT enabled). France and the UK were great, but Italy sometimes showed nothing but blobs of paint - no autogen at all, let alone the land-class stuff.

 

Needless to say, I have been experimenting with my AM setting, taking off from the Rome airport and heading down the Tiber River at 250 kt and at 3000 FT. I have tried 4 settings: 255, 254, 253, and 252. 255 gave me the sharpest land-class, but sometimes with unacceptable stuttering. 252 had the least stuttering, but with slight blurring. However, 254 and 253 gave the best overall results. FPS was about 35 for all of these cases. I monitored the core usage, which was as expected: 255 used all 8 cores, 254 and 253 used 7 cores, and 252 used 6 cores.

 

I will be doing a lot of flying in Europe over the next few weeks to see if my current settings (AM=254) continue to be correct. If it is, I will be buying the ORBX NA land class the day it comes out.

 

Frans