I saw little or no performance gain in my test flight from KORD to KSTL in the PMDG777 w/ASN, MSE Photoscenery, FSDT KORD, Taxi2GateKSTL, and ProATC-X. 

 

---------------------------------------------------

 

Me too, actually.  I now have it set to '1', but didn't see the 10 plus FPS bump up that others are reporting (with v2.4). Nothing has gone down, and I suspect a few more FPS are here and there...so that is a good thing. Just nothing dramatic, at least with my system in use.

Even though the theory you are telling sounds perfectly reasonable, in practice these settings just don't work on my system. If I set a target of 30 fps in P3D together with VSync and Triple Buffering and set the Frame Limiter to 30 fps in NI, then the sim is just a stuttery mess. It perfectly holds the 30 fps in the default scenario with the F22 in Langley (frame variance as displayed by P3D is 0.1%). However, it stutters like hell. If I unlock the frames (unlimited in P3D with VSync and Triple Buffering + no limit in NI) then even when the sim is not able to hold 60 fps (which is obviously smoother than 30 fps), I see much less stutters. Same thing in aircraft that are heavy on frames like the PMDG 777. Much smoother with limited frames. So even if the theory makes perfect sense, it just doesn't work for me...

You've got to set the system up to work with Fixed if you want to use fixed. If you found 30 too much, perhaps you need less, or something else is wrong, maybe try a new NVidia CP profile, or reduced settings. So it makes no sense to say it doesn't work, if you want fixed you need to make it work.

Also you are ganging up two service threads per core elsewhere on the CPU, this can lead to contention and stutter. I would try 84=(01,01,01,00), 168=(10,10,10,00), 21=(00,01,01,01), etc. Masks like 244=(11,11,01,00), do the job of 1) keeping the main thread to a core, 2) leaving a core free, and 3) *intensify* the use of the cores by the service threads to maximize those other two cores. Too many threads will destabilise the main thread in the end, and it won't maintain smoothness overall.

 

yesterday evening i tested the values you wrote and they do not work well in my system: leaving cores unused cause blurries with FFTF=0.01 (and i see no point increasing it to change AM). Also with HT=off i noticed higher FPS but slower texture loading causing more blurries in heavy autogen situations.

 

Maybe it is also for my not-high-end GPU (ATI R9 270x)

 

So better value for me seems to be AM=254 (slightly better than no AM settings, 2-3FPS)

 

tks anyway!

yesterday evening i tested the values you wrote and they do not work well in my system: leaving cores unused cause blurries with FFTF=0.01 (and i see no point increasing it to change AM). Also with HT=off i noticed higher FPS but slower texture loading causing more blurries in heavy autogen situations.

 

Maybe it is also for my not-high-end GPU (ATI R9 270x)

 

So better value for me seems to be AM=254 (slightly better than no AM settings, 2-3FPS)

 

tks anyway!

 

Affinity Mask folks - please stop mixing in results (and debate) from that tweak with feedback on this thread.  AM has enough of it's own threads to not confuse this one with performance results exclusively from a different tweak.  

 

I understand they are all inter-related, etc, etc, but it is not helping this particular discussion.

yesterday evening i tested the values you wrote and they do not work well in my system: leaving cores unused cause blurries with FFTF=0.01 (and i see no point increasing it to change AM). Also with HT=off i noticed higher FPS but slower texture loading causing more blurries in heavy autogen situations.

 

Maybe it is also for my not-high-end GPU (ATI R9 270x)

 

So better value for me seems to be AM=254 (slightly better than no AM settings, 2-3FPS)

 

tks anyway!

The values I presented were for educational purposes, to explain the AM, not as an example of what will work on your setup. No point debating AM without a thorough understanding of the principles.

Affinity Mask folks - please stop mixing in results (and debate) from that tweak with feedback on this thread.  AM has enough of it's own threads to not confuse this one with performance results exclusively from a different tweak.  

 

I understand they are all inter-related, etc, etc, but it is not helping this particular discussion.

+1

So better value for me seems to be AM=254 (slightly better than no AM settings, 2-3FPS)

254=(11,11,11,10). So with 254 you have one core for the main thread (pair of LPs on the right), followed by all LPs loaded with service threads. This means that the OS will probably not target core 0 for new processes, because your main P3D thread is maxed out on it (usually). But no matter where any other process starts, elsewhere you have a service thread, and so there's others to take up the slack. Might work well in many systems. I hope that helps clear things up for you there a little.

+1

 

 

+2

+2

Post something useful guys.

254=(11,11,11,10). So with 254 you have one core for the main thread (pair of LPs on the right), followed by all LPs loaded with service threads. This means that the OS will probably not target core 0 for new processes, because your main P3D thread is maxed out on it (usually). But no matter where any other process starts, elsewhere you have a service thread, and so there's others to take up the slack. Might work well in many systems. I hope that helps clear things up for you there a little.

tks again SteveW for your help,  I did not mean that your reasoning was wrong, only that in my system (perhaps for a GPU not at the top) needs a larger number of LP compared to values such as 84, 244, 243 for proper texture loading (beeing FFTF=0.01). Your last explanation has sense imho.

 

...And sorry to the others for the OT 

....anyway....what about the Optimize Dense Clouds setting? Has anyone seen further thoughts / insights from LM?

....anyway....what about the Optimize Dense Clouds setting? Has anyone seen further thoughts / insights from LM?

Yeah...what does it do? Does it make clouds more visually appealing? Or reduce the hit when flying through them?

Optimize_Parts=1 worked for me aswell.

 

One thing to try for those who did not see a benefit

 

When I first loaded the NGX after the tweak, I saw no benefit and my frames were actually worse. But after I cycled through and loaded all my views I saw the same benefits as most posters to this thread.

 

The improvement in smoothness and fluidity is very apparent as well. 

 

Awesome tweak.

Well of course it's true. There's been a commercial airing on television for the past month or so proudly declaiming to the teeming masses of couch potatoes that "Six is greater than One!"...

 

...what an amazing discovery! I hope Steven Hawking has been watching.

 

Yeah...what does it do? Does it make clouds more visually appealing? Or reduce the hit when flying through them?

It's all performance related in how the engine render and output the stuff to your hardware ;-)

After a few flights  with this tweak I am definitely satisfied that has made a huge improvement to my 777 flights.