I read you Vic, and I have indeed just spent the last 30 minutes reading through the post and am none the wiser, although my head cold may have something to do with it....... 

 

I thought it was more complex than just understanding how binary works... I thought the right setting related to the CPU, your graphics card etc and which cores you would want to enable and not enable, which is not immediately apparent, hence my question. Cheers, Mark

...Re vsync stutter; a 120Hz monitor rather than a 60Hz monitor would half the stutter.

 

...I am sure it is the fact that P3D doesn't run in true full screen so the whole locking fps at half refresh rate doesn't work...

Yes Richard, if you mean the settings available with NI, you are correct, forget 1/2 refresh etc. in P3D, only desktop vsync is supported. Faster refresh monitors reduce that little micro-stutter.

 

 

RE; back to AMs a second.

 

What might not be clear when we make a binary value to apply a mask, we say for example, 1110, that value is the reverse order of the LPs shown in Task Manager. Starting on the Right and going Left, The zero is LP0, the ones are LPs 1, 2, and 3.

 

In Windows Calculator, Programmer mode, we switch to Bin (on the right mode panel), type in 1110, then switch to Dec (on the right mode panel), to see the decimal equivalent required for the AM setting to be placed in the config file [JOBSCHEDULER] AffinityMask value.

I have two questions for you Steve.

 

I never noticed any real difference playing around with vsync and maybe that's because I'm using a Samsung LEDTV rather than a computer monitor. Don't know out of my head what refresh rate I've got with that TV but probably 100 Hz or more. What do you think?

 

Then my second question and asked about this in some other thread too but wanted to hear your opinion on this. Noticed yesterday that my Titan card is constantly maxed out looking at the GPU load when running P3D. Used the built-in monitor in NI to check this.

 

Since I learned you never want your GPU constantly maxed out I guess I should try addressing that issue but I'm not sure how. Any particular settings in NI or P3D that would max out the GPU?

I just don't get to play with enough different hardware to have authority on this side of the equation. Generally you will have vsync forced on and so making changes in NI will have no result. There are those reporting vsync off occasionally, they find messing with Aero restores it. Not sure what's going round to cause that. With your card maxed out, could be it's just doing a lot of stuff. On the one hand we want to max it all out, on the other we want to save thrashing our equipment and opening the door to let the heat out. DX11 does more on the GPU than FSX DX9 so when we used to have quiet PCs running FSX, we went to DX10 and the fans fussed a bit, then to P3D and the fans fuss more.

 

Any particular settings in NI or P3D that would max out the GPU?

AA modes. Maybe increasing the look ahead buffer (when using fixed internal). Autogen re: dense urban or city. Weather, thunderstorms.

OK, my worries about having my GPU at 100% isn't temps in the first place because I have a cool case and I've also checked the temps and so far no problem there.

 

What I'm a bit worried about is when the GPU load is nearly 100% all the time what will happen when even more is asked by the GPU meaning it will be overloaded? I guess that will result in bad performance inside P3D when my GPU can't keep up with what's expected.

 

Will try to play around a bit with all settings later on trying to figure out what is the major cause.

I love how this thread is 90% discussion about Affinity Mask between 2 or 3 people completely derailing the subject of the thread, which is OPTIMIZE_PARTS...there are plenty of other AffinityMask threads, why do the mods let this one get hijacked?

 

I have to spend more time filtering out the AM junk to read actual reports related to OPTIMIZE_PARTS.  Awesome job guys.

I'd say the impetus of the OPTIMIZE_PARTS discussion wavered and there are a few that stayed behind to discuss AM where other discussions failed to get across. Maybe you understand AM thoroughly, but it's hardly junk, and it's just a few guys enjoying the forum.

OK, my worries about having my GPU at 100% isn't temps in the first place because I have a cool case and I've also checked the temps and so far no problem there.

 

What I'm a bit worried about is when the GPU load is nearly 100% all the time what will happen when even more is asked by the GPU meaning it will be overloaded? I guess that will result in bad performance inside P3D when my GPU can't keep up with what's expected.

 

Will try to play around a bit with all settings later on trying to figure out what is the major cause.

What will happen, is that you will get the crash 'driver has disconnected from GPU'...the G card hangs, and times out, (yes..can happen even with the M.S. tweak to increase wait times...).  You usually get the B.S.O.D. and a crash to desktop.

Later tests shows bad perfomane of optimize parts with a2a aircrafts. Expecially not native ones. With fps unlimited max fps 40 in ftx norway wild areas, 50-60 with op=0. A2a p51d (it is fsx native). Also noticed a little less difference with a2a c172 (p3d native).

 

Do not use op for a2a planes that is my exp. Anyone having bad results with a2a or other ga?

Quote P3D developer Kevin Cartrette:

 

"This attempts to optimize multiple parts of a model by combining them into fewer meshes if possible. If you have a complex model that is not already optimized to reduce draw calls etc, this will attempt to do that for you. There is some overhead to this, as it happens at runtime when the model is being loaded and ideally the model would already be optimized before runtime. This only happens for FSX Models, and not older model types."

 

 

Quote P3D developer Adam Breed:

 

"The reason this is disable by default is we identified potential issues with crashing on Windows 8 machines."
 

 

Geesh guys, don't you ever check the P3D forum, you know....the guys who actually made this product....

@ SteveW

 

I tried AM 254 last night as I liked the idea of just leaving 1 core free and then I set TrackIR and ASN to only use the free core not used by P3D. Seems to work really well. Very smooth.

 

Thanks

 

 

I'd say the impetus of the OPTIMIZE_PARTS discussion wavered and there are a few that stayed behind to discuss AM where other discussions failed to get across. Maybe you understand AM thoroughly, but it's hardly junk, and it's just a few guys enjoying the forum.

+1

Geesh guys, don't you ever check the P3D forum, you know....the guys who actually made this product....

Do you see the "Best answer" on top of this thread?

What will happen, is that you will get the crash 'driver has disconnected from GPU'...the G card hangs, and times out, (yes..can happen even with the M.S. tweak to increase wait times...).  You usually get the B.S.O.D. and a crash to desktop.

 

Yeah, I know that is what will happen in a worst case scenario but I've very rarely seen that happen...not even sure I've seen it once since I got my Titan card. I was more thinking of the performance impact in P3D when the GPU constantly is at 100% leaving no headroom at all and how I guess that could cause severe stutters and/or FPS drops when you're entering some demanding cloud formations, an area with lots of other online traffic etc.

 

Anyway I think I found a solution to my high GPU load. First I tried lowering pretty much every setting in P3D but to no avail. The only setting that would make the GPU load drop with P3D running was to set TESSELLATION=0. However I do want P3D to make as good use of my GPU as possible so disabling Tessellation all together didn't seem like an option to me.

 

What I finally found out was that if I would limit my FPS externally and/or internally (or internally not limit but rather set a target) the GPU load finally dropped and after doing this I could actually see how different other settings had an impact on the GPU load meaning I was able to tweak settings until I had decreased the GPU load to a comfortable level which in my eyes is somewhere between 70-80% under normal conditions. At that level I know my GPU is well utilized but at the same time it will have some extra strength left when needed.

 

Thought I would share this since I always ran unlimited FPS before both in FSX and P3D since that would usually give me better performance in most situations looking at the FPS counter but after this discovery and seeing how swamped my GPU was running unlimited frames I'm now running my FPS locked at 30 both in NI and P3D and so far things are very smooth. Only did a very short test flight in the NGX yesterday but even after that short flight I was very pleased with the result and also checking the GPU load log afterward seeing the GPU load moved in a healthy way between 70-95% or so throughout the flight.

Quote P3D developer Kevin Cartrette:

 

"This attempts to optimize multiple parts of a model by combining them into fewer meshes if possible. If you have a complex model that is not already optimized to reduce draw calls etc, this will attempt to do that for you. There is some overhead to this, as it happens at runtime when the model is being loaded and ideally the model would already be optimized before runtime. This only happens for FSX Models, and not older model types."

 

 

Quote P3D developer Adam Breed:

 

"The reason this is disable by default is we identified potential issues with crashing on Windows 8 machines."

 

 

Geesh guys, don't you ever check the P3D forum, you know....the guys who actually made this product....

 

Yep, this answer explains it all very good I think and makes it easy to understand why you see different results using this parameter depending on what aircraft you're flying.

 

I'm doing all my flying in the NGX and I had great results using this parameter. This of course was a great discovery but at the same time I'm a bit surprised since that looking at the answer above from P3D suggests this would point to the fact that the P3D model of the NGX is not fully optimized out-of-the-box for P3D.

 

Did anyone already ask PMDG about this? Would be very interesting to hear they comment on this.

Here you go Richard

Guys,

We'll need further guidance from LM on this if there's something that needs fixing - our models most certainly are optimized to reduce draw calls and our P3D models are not just the FSX ones thrown into the P3D folder - they were adjusted and exported using the P3D specific tools provided in the P3D SDK. If there's something LM's realtime optimizer is doing that we don't know about then they will need to make that information public so that we (and other devs) know about it.