P3Dv4.3 : my observations

[Guest]

10 minutes ago, SteveW said:

I disagree, a baseline recording need only be from a stationary location within scenery as for example parked at the gate.

Ok, I guess we disagree ... I don't see how being stationary would establish any relationship to what one actually does in a flight simulator?  Based on how rendering is culled and optimized and the on GPU cache process, stationary would not provide any meaningful data.

Cheers, Rob.

 

[SteveW 1,792]

Simply that by touching the controls or altering the view you can't repeat that each test. It doesn't matter about the view so long as it is IDENTICAL each time. Moving the controls or view won't help.

[SteveW 1,792]

Meaningful repeatable data is what's required when making tests. Repeatable tests and guaranteed repeatability counts 100%. Anything else is a mistake.

[SteveW 1,792]

I tried panning programmatically in tests hoping to get more solid repeatability but it doesn't work like that and couldn't be used confidently. Settling time within the sim can take several minutes and that just amplifies the problem.

[SteveW 1,792]

...Also worth noting if wanting to do repeatable testing is avoid messing with shaders because if they are building that also detracts from repeatability. Unless they are unset each time the same and then we would simply be measuring shader build time which is blurring the figures.

 

[SteveW 1,792]

1 hour ago, Rob Ainscough said:

 stationary would not provide any meaningful data.

 

 

Obviously you don't see my point Rob and frankly that's an over-burdening statement with no basis of reason. And I feel I have to tackle you on this point you make - sorry about that - because there's people spending hours watching and it doesn't need that. Look here:

Guys - It's OK to measure the response of the starting procedure until settled. That's repeatable and measures all things equally.

You don't need to move, or pan around - unless you think there's something else worth looking at in the sim - that infers you might want to measure every view available in the sim.

Instead you just need one repeatable view.  Moving or panning simply increases variance on your result.

Repeating that startup process gives you all the real data you need (not "none") and that is what the sim responds like throughout the flight.

[SteveW 1,792]

..Proof of that concept!

I've posted this shot before Rob.

Here you can see FFTF altering the behaviour of settling time. Not what I would call no data:

The simulator is measured during startup with the exact same procedure each time no faffing with view or controls - that would spoil it:

 

[SteveW 1,792]

...and here again using Prepar3D.cfg to set FFTF and seeing real data - no less!

 

 

[JoeFackel 760]

8 hours ago, GSalden said:

Very kind of you.  

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

Okay, i entered EHAM with stock Bonanza, switched to EZCA World view looking from the tower over the apron in direction around 020 and let it run with UTL for around 4 hours. It started with 18,8 fps and after 4 hours it was 17,9 fps. 

[Guest]

Steve,

We're at the point where the only way to determine what and if anything meaningful comes out of this is with application profiling using tools provided by nVidia (since Gerard is using nVidia hardware) and with the cooperation of LM to implement those tools into the source code and run a special compile/build with those tools in place ... then start testing.  Gerard has all kinds or variables in his FS setup including an airport that doesn't have a P3D V4.x installer ... the shear combinations of possibilities/variables is staggering.

At this point I'll bow out of this thread, I'm not trying to derail you folks (or get involved in a debate of testing procedures) from finding the source of Gerard's issue (if it is an issue) ... if you can help Gerard find the source of his troubles then I'm all in favor and a big thank you for helping out the community.

Cheers, Rob.

[SteveW 1,792]

2 minutes ago, Rob Ainscough said:

Steve,

We're at the point where the only way to determine what and if anything meaningful comes out of this is with application profiling using tools provided by nVidia (since Gerard is using nVidia hardware) and with the cooperation of LM to implement those tools into the source code and run a special compile/build with those tools in place ... then start testing.  Gerard has all kinds or variables in his FS setup including an airport that doesn't have a P3D V4.x installer ... the shear combinations of possibilities/variables is staggering.

At this point I'll bow out of this thread, I'm not trying to derail you folks (or get involved in a debate of testing procedures) from finding the source of Gerard's issue (if it is an issue) ... if you can help Gerard find the source of his troubles then I'm all in favor and a big thank you for helping out the community.

Cheers, Rob.

 

I don't think so, but bow out if you must.

What the top graph shows is a flight in from a saved point, no adjustments to anything, just the load up and sim setting off down the path to EHAM. The lower graph is at a standstill. Both graphs give you what you need Rob.

[GSalden 4,655]

Finally good news :

Now with the latest NVidia driver 399.07 + 4x SSAA I do not had any framerate drops in the last 3 tests...

All 3 drivers before I have tried with a clean install and experienced the framerate drop. Using 4x SSA, 8x SSAA or 8X MSAA made no difference.

The moment I installed 399.07 icw 8X SSAA I did not see a framerate drop at several stationary test spots. The only thing I remarked was a lower CPU load when using 8x SSA compared to using 4x SSAA.

I have done 2 Frankfurt-Amsterdam flights with 4x SSAA and no framerate drop so far...

I want to thank Steve, Rob, Mike and all others who anticipated in this thread trying to help with tests and suggestions. You guys are the best 😉

 

 

 

[AnkH 2,233]

Good to know, I am certainly going to install this driver now. Thanks for all your efforts!

[SteveW 1,792]

9 hours ago, GSalden said:

Finally good news :

Now with the latest NVidia driver 399.07 + 4x SSAA I do not had any framerate drops in the last 3 tests...

All 3 drivers before I have tried with a clean install and experienced the framerate drop. Using 4x SSA, 8x SSAA or 8X MSAA made no difference.

The moment I installed 399.07 icw 8X SSAA I did not see a framerate drop at several stationary test spots. The only thing I remarked was a lower CPU load when using 8x SSA compared to using 4x SSAA.

I have done 2 Frankfurt-Amsterdam flights with 4x SSAA and no framerate drop so far...

I want to thank Steve, Rob, Mike and all others who anticipated in this thread trying to help with tests and suggestions. You guys are the best 😉

 

 

 

Great news! As you know I was particularly interested in the driver after you reduced the cores. I didn't expect a fault in the driver though since it's usually the profile which can't be used reliably after hardware or driver changes that's why I go on about doing the restore/apply so often.

 

Edited August 31, 2018 by SteveW