1 hour ago, GSalden said:

TrueEart Netherlands is very hard on the system as also FT Amsterdam ( 6 runways and therefore more taxing than an airport with 2-3 runways )

As i own TrueEarth NL and FT EHAM now i will try it next time when i check the AI traffic next time. Will let it run w/o ASP4/SF3 for 60-90 mins. and look for fps drops.

Regarding EHAM: there are only 2 or 3 RWYs open depending on the used AFCAD afaik, so that shouldnt make a difference to an other bigger addon airport.

Edited August 30, 20187 yr by JoeFackel

16 minutes ago, JoeFackel said:

As i own TrueEarth NL and FT EHAM now i will try it next time when i check the AI traffic next time. Will let it run w/o ASP4/SF3 for 60-90 mins. and look for fps drops.

Regarding EHAM: there are only 2 or 3 RWYs open depending on the used AFCAD afaik, so that shouldnt make a difference to an other bigger addon airport.

Very kind of you.  

About EHAM. It is not the Afcad file but the rw plus it’s textures. Mathijs Kok once explained this.

5 minutes ago, GSalden said:

About EHAM. It is not the Afcad file but the rw plus it’s textures. Mathijs Kok once explained this.

Sorry, i meant of course the ADE which is swapped out when choosing the right one in the FT configurator regarding the actual wind situation.

31 minutes ago, Cruachan said:

Hi Gerard,

Perhaps something is being reset in the graphic drivers and switching AA is triggering this behaviour. I would imagine that you have tried this already, but do you see this effect while using a freshly generated Prepar3D.cfg with no or minimal changes made by the user? 

Did this reduction of frame rates over time happen before you SLI’d your 1080Ti?

I’m speculating about the possibility that you are experiencing some form of saturation and this is developing over time to the point where it affects frame rates. Changing AA may be forcing a ‘fresh start’, a graphic reset if you like.

Regards,

Mike

Hi Mike,

Also I found another thing : driver 399.07 seems an improvement for me.

At this moment I am testing an approach in Slew mode ( staying at one spot ) and after 50 min no framerate drop yet !

With all 3 before this one I experienced the framerate drop. And all have been clean installed.

My setup is also different from other flightsimmers : 2x 32” in Surround view + 1x 40” view —> connected to 1 card that is in SLI with a second card...

 

Also found out that 8x SSAA shows less cpu load than 4X SSAA or any other AA setting. Even lower than with AA off.

 

41 minutes ago, GSalden said:

Also found out that 8x SSAA shows less cpu load than 4X SSAA or any other AA setting. Even lower than with AA off.

Hi Gerard,

I’d love to read the explanation for that! It’s as if the CPU is more involved in sharing AA computations with the GPU at the lower SSAA setting than at the higher setting when the latter takes on most of the burden. When was the ability to select SSAA introduced to the Prepar3D GUI? Perhaps this feature is still in need of further refinement. I use 8xMSAA and have not been aware of this frame rate drop over time. However, I do prefer to lock at 31 (currently) so would not necessarily notice a gradual fall in FPS unless it occurred in a situation where it would not be expected.

Regards,

Mike

Edited August 30, 20187 yr by Cruachan

25 minutes ago, GSalden said:

Also found out that 8x SSAA shows less cpu load than 4X SSAA or any other AA setting. Even lower than with AA off.

 

Classic example of the GPU starving the cpu.

15 minutes ago, FunknNasty said:

Classic example of the GPU starving the cpu.

Hi Ken,

Can you expand on that statement please, as I’m not sure I understand. Gerard has a potent hardware setup by any standards so I’m wondering whether this would be true in his case.

Thanks.

Regards,

Mike

19 minutes ago, FunknNasty said:

Classic example of the GPU starving the cpu.

? 

Normally the GPU has to work harder with higher and more taxing AA settings.

What that has to do with less CPU load is a big question for me.

Only 8x SSAA shows a lower CPU load ; all other AA settings incl no AA show the higher CPU load...

1 hour ago, Cruachan said:

Hi Ken,

Can you expand on that statement please, as I’m not sure I understand. Gerard has a potent hardware setup by any standards so I’m wondering whether this would be true in his case.

Mike

As is often the case, those with potent hardware will ask the hardware to do more ....

54 minutes ago, GSalden said:

Normally the GPU has to work harder with higher and more taxing AA settings.

What that has to do with less CPU load is a big question for me.

Only 8x SSAA shows a lower CPU load ; all other AA settings incl no AA show the higher CPU load...

in simple terms ....your cpu is taking a nap while the GPUs are busting their arses trying to get that 6000x3000 picture out the door in under 5 minutes. The extra time to render that super sampled frame at the resolution you're running is holding up the assembly line to the point that everything downstream (handled by the cpu) is left standing around.

To me it looks like you describe the other way round...

You are saying that 8x SSAA is less taxing for the CPU than 4X SSAA.

However , the 1080Ti SLI setup has no problems calculating it. 

Franerate does not drop at all so the CPU is doing the same to me...

Edited August 30, 20187 yr by GSalden

3 hours ago, GSalden said:

To me it looks like you describe the other way round...

You are saying that 8x SSAA is less taxing for the CPU than 4X SSAA.

However , the 1080Ti SLI setup has no problems calculating it. 

Franerate does not drop at all so the CPU is doing the same to me...

 

No ....that's not what I'm saying.

I used the term ''''assembly line'''' as an analogy for the process of building a single frame. P3d uses clock cycles. Those cycles are fixed in time. P3d sends out instructions to your hardware that in an ideal world should be carried out within the cycle (might be 32ms). The fact that you are reporting lower cpu loads with higher higher AA settings (and higher cpu load with lower aa) within the same scenery tells me that not all instructions asked for by P3d are being carried out within that cycle (when using the higher AA settings). 

So, what I am saying is that the processing demand for 8xSSAA (which is done by the gpu) is sucking cycle time and leaving no time for other things that P3D is asking for and that otherwise would be handled by the cpu

Edited August 30, 20187 yr by FunknNasty

2 hours ago, GSalden said:

However , the 1080Ti SLI setup has no problems calculating it. 

Go into NCP and turn OFF "Shader Cache" (Global setting should have it set to On) and run the test again.

Your information is kinda all over the place but the best I can get from it is that the trigger is moving AA from 8XSSAA to 4XSSAA or from 4XSSAA to 8XSSAA will increase FPS when you are sitting stationary?

What I'd recommend is you need to find a "baseline" and always work from that rather than just "trying stuff" ... a good baseline would NOT be from a stationary position.  That baseline also requires that you have no AI, no weather, nothing that would be considered "dynamic" or skew results.  You would need a flight pattern ... preferably an AP viable pattern as it would reduce variances.  It's important that you run a pattern as that will ensure caching happens and all tiles/objects are loaded (even if not visible) within a 360 radius.

After you get a repeatable test, then go back to NCP and turn ON "Shader Cache".  You may also want to repeat the test with SLI On/Off.

Cheers, Rob.

I disagree, a baseline recording need only be from a stationary location within scenery as for example parked at the gate. Moving any controls or changing/panning the view will destroy the baseline figure. A second test further on with setup might be to see the result of a curved path of flight from a saved file with guaranteed behaviour set up in the sim with no flat-lines but it's only optional

Setting up two cards is what changes AA since it's mostly all that more cards can do for the sim. Be sure that the Profiles are all Restore/Apply'd even if you have to make a setting just to reset it to default, because those profiles contain problems if the system changes or a driver changes.