1 hour ago, GSalden said:

Friday evening I will make a short vid to show how fps change when the FFTF value changes. This can be seen in both Locked as Unlimited mode...

Well there we have it - FFTF is not restricted to certain modes of operation or certain settings, because the sim has to collect the data and has to assemble the data for display whatever the setup, the proportion is of those two things - they always happen.

 

1 hour ago, GSalden said:

I have both FFTF Dynamic as Booster Live 2018. 

Friday evening I will make a short vid to show how fps change when the FFTF value changes. This can be seen in both Locked as Unlimited mode. 

As I was one of the beta tester I did many tests in both modes.

@Mike : thank you for posting 

 

 

 

 

Bring it on, GSalden!

Why you still need FFTF Live 2018 just baffles me though...

Unfortunately SteveW, being my P3D neurosurgery Gawd, is way way up there beyond my depth, even when he's saying something mundane. It's users' interpretation that helps me understand his verses most of the time 

Thanks,

Shouldn't FFTF be adjusted also based on the speed of the aircraft instead of just altitude? 600 knots? Mach 3? 

If you read the old Phil Taylor explanation, it would seem that the faster you fly the more time fibers would need and the faster the scenery will try to load. 

Curious to see the results on a Concorde, SR-71 or even the F-22 on full burner at Mach 2.2

First analysis of FSPS’s FFTF Dynamic P3Dv4

 

PC 3960x, GTX680, Windows 10, 2x1920x1200@59Hz.

Simulator: Prepar3D v4.2 Developer

FFTF App: FSPS FFTF Dynamic, P3Dv4, v1.0

Test Harness: IFPro 10.027ed

 

Setup

P3Dv4.2 is set to 24 fps Locked. This is around the point where the settings and the screen sizes can just cope with 24fps. The entire setup is such that the sim is handling the scenario nicely as it approaches EHAM. It is set to just keep out of difficulty handling the locked fps.

We want to see if the fps can be maintained as the scenery complexity increases. We want to compare the result with and without FFTF Dynamic.

FFTF Dynamic is set to AGL Mode 0 to 4000ft and 0.05 to 0.33 FFTF, 3 times per second is checked. FFTF Dynamic wants to run with admin privileges (set in the application manifest). Since it is a .NET App it carries with it the SimConnect module from the SDK Managed folder as it is not part of the P3D installation. The app uses approximately 80MB and between 0.2 and 0.4 % CPU with these settings.

The monitor app/test harness is a 10MB exe (x86) uses approximately 35MB memory with a load typically between 0 and 0.2% CPU during the run of the sim. The harness uses any x86 SimConnect installed with P3D (\redist). The 64bit function is handled by a codelegend 64bit bridge, which is private to IFPro. This configuration is leaner than an equivalent fully 64bit implementation.

The harness starts up with admin privileges and then in turn handles the start-up of FFTF Dynamic with affinity correctly applied. FFTF inherits admin privileges and starts up alongside the harness.

A saved flight file (.fxml) is used to get the sim straight into the cockpit, the aircraft continues un-paused on descent from the English Channel at 4000ft, over four minutes into EHAM. P3D is set with medium graphics values. Weather is injected by IFPro, no AI Traffic was used.

The harness repeats the exact same situation throughout every flight as close as can be maintained on the system, with and without FFTF Dynamic automatically.

The monitor analyses the output framerate from P3Dv4 and keeps a record in a file.  The file is a simple .csv that can be converted into a graph by Excel or something like that.

 

Core Affinity

FFTF Dynamic:

10,10,00,00,00,00=2560

IFPro:

01,01,00,00,00,00=1280

P3Dv4.2:

00,00,11,11,11,01=253

 

The graphs below show the result of two pairs of identical flights, one pair with FFTF Dynamic and the monitor, the other pair with only the monitor app:

 

 

Visually the results on screen were similar in appearance in both sets. No additional disturbances of note were added to the flow with the FFTF app. Weather is cloudy around 60 Nm vis.

Conclusion: The FFTF app appears to enable the simulation to maintain fps to keep a straighter line on the graph when the scenery complexity increases.

 

Notes:

The green boxes show the run with no FFTF Dynamic included. The red boxes show flatter lines with the exact same flight repeated.

The flight starts at 4000ft and concludes after four minutes at around 1200ft just on top of EHAM. This is the right hand end of the graphs. The blue line shows the actual fps.

EHAM scenery installed addon from a top brand set with medium textures.

IFPro was not applying Expanded Affinity as it is not appropriate with the setup. No tricks or clever tactics have been employed. The test was straight out of the box. FFTF Dynamic is a very good app technically, it installs well and handles nicely.

 

17 hours ago, vortex681 said:

There's a lot of discussion about the potential value of changing FFTF but not so much about how successful (or otherwise) the FFTF Dynamic app is at doing so.

If you have good framerates with changing the default FFTF then you don't need a utility which changes it either.

I have now got some practical experience with the FFTF utility, and it suits my system very well. The reason? With the default setting I am lucky to reach 15 fps with all the scenery and AI traffic I like at denser airports like EGLL and EHAM. This is due to using three windows for scenery for my three projectors.

Up till now I've always set FFTF low, in the CFG file. Something like .10 or even as low as .01. This can give me better frame rates -- up to 20 instead! ;-)

BUT with that as my 'permanent' setting in the CFG file, I do get noticeable blurries when flying my 738.  For that reason I came to far prefer hazy dull weather for flying rather than sharp clear sunshine!

The FFTF utility progressively raises the FFTF to my default max of 0.33 when away from the airport.  Currently I'm using the frame rate limit options in FFTF for this, but I can see that the Altitude option could work just as well.

Result: I can stop worrying about the visibility. No blurries.

Pete

 

 

4 hours ago, Slayer said:

Curious to see the results on a Concorde, SR-71 or even the F-22 on full burner at Mach 2.2

I think at the altitude they fly there's not a lot of ground detail to go blurry in any case! ;-)

Pete

 

and their flapping cheeks get in the way of vision...

Now, that’s the way to present a test analysis!

Excellent and, from where I stand, these results do seem to be pretty conclusive.

Much kudos, Steve!

Would there be anything to gain by repeating the tests with fame rate unlocked?

Mike

1 minute ago, Cruachan said:

Now, that’s the way to present a test analysis!

Excellent and, from where I stand, these results do seem to be pretty conclusive.

Much kudos, Steve!

Would there be anything to gain by repeating the tests with fame rate unlocked?

Mike

Thanks Mike.

Part of the way it works is that there is overhead in the system. The same tests without locking, using Unlimited VSync=Off don't prove much other than what I showed in the earlier set of graphs. There's little point posting the graphs of those at all.

The same results are evident when departing, there's more free power in the system to keep the fps up during the departure and climb out.

1 hour ago, SteveW said:

Part of the way it works is that there is overhead in the system. The same tests without locking, using Unlimited VSync=Off don't prove much other than what I showed in the earlier set of graphs. There's little point posting the graphs of those at all.

Fair enough. I use Locked 30 these days with no VSync or TB so, once I get my AM sorted out (still testing!) I can try reintroducing the FSPS FFTF Dynamic utility and then, hopefully, I should be good to go 😀

Regards,

Mike

Hi Steve,

It’s interesting to note how much smoother the FPS graph line displays while the FFTF Dynamic utility is active. This implies improved performance during your flight scenario. Are those persistent dips being caused by the now familiar inherent long frame instances irrespective of whether the utility is or is not in use? Certainly the utility’s 3 times a second setting doesn’t seem to be having any obvious negative impact.

Regards,

Mike

1 hour ago, SteveW said:

Notes:

The green boxes show the run with no FFTF Dynamic included. The red boxes show flatter lines with the exact same flight repeated.

The flight starts at 4000ft and concludes after four minutes at around 1200ft just on top of EHAM. This is the right hand end of the graphs. The blue line shows the actual fps.

EHAM scenery installed addon from a top brand set with medium textures.

IFPro was not applying Expanded Affinity as it is not appropriate with the setup. No tricks or clever tactics have been employed. The test was straight out of the box. FFTF Dynamic is a very good app technically, it installs well and handles nicely.

 

I wish you would run the same test will just FPS locked at 24 (as you have done and I also use locked at 24) and a FFTF put in your config of 0.01

Lets see how that works out. v the app.

Because unless you have texture loading problems I see no reason to use it.

FFTF set at 0.01 FPS locked at 24 job done for me.

To see that in the same test would show if the app is really needed. Because the best looking one of them charts is the one with 0.05, well 0.01 should be even better.

Edited March 29, 20188 yr by Nyxx

23 minutes ago, Nyxx said:

I wish you would run the same test will just FPS locked at 24 (as you have done and I also use locked at 24) and a FFTF put in your config of 0.01

Lets see how that works out. v the app.

Because unless you have texture loading problems I see no reason to use it.

FFTF set at 0.01 FPS locked at 24 job done for me.

To see that in the same test would show if the app is really needed. Because the best looking one of them charts is the one with 0.05, well 0.01 should be even better.

What happens if you set the AG or LOD slider one notch higher ?

Then you might need a higher FFTF value to avoid blurries...