Limiting Framerate kills framerate - why?

[SteveW 1,792]

Let me try and get some theory behind this:

FFTF set in the cfg alters the ratio of processing frames to loading data - left out of the cfg FFTF assumes around 0.3. We can see this in the graphs below.

Remember that data requirements to actually build a frame mean FFTF won't drop below or rise above whatever is achievable at the time irrespective of setting because at least some of each side of the ratio must be completed at all times in order to render the scene.

The sim tries to maintain a "constant fps" irrespective of available data, hence we see blurriness when the data is not fully loaded in time because it gets on with drawing the frame regardless. So let's look at that a bit closer:

When I refer to a 'constant fps' I mean this is the rate maintainable for the current settings on the hardware if you will. For example; we can add shadows on that setting and fps goes down accordingly, we can reduce resolution on that kit and fps rises accordingly. In other words fps is constant.

So there is always a theoretical rate available and there's always some of each side of the ratio completed whether we set unlimited or fixed.

In fact fixed or locked fps similar to Unlimited VSync=Off gets on with the next frame right away (up till the buffer is filled) and uses up all the processing it can usually unless the system is very little taxed. Setting fixed or locked is not a simple limit since it builds look ahead frames which have to be replenished after any wobble in the flow. Setting Unlimited + VSync=On instructs the GPU to target frame output for the monitor refresh. Only this way can the system be reliably 'held back'.

So I can set FFTF in the cfg and by utilising the Unlimited + VSync=Off setting I can make the following graphs which show the background task taking longer with FFTF=0.1 and completing more quickly at FFTF=0.9 - fps remains almost constant. The graphs of no cfg setting and FFTF=0.3 show similar results proving the manual to be correct. Following the graphs I have included an excerpt from my site that explains FFTF in a little more detail:

 

 

 

"fps provided by the sim is the result of Performance Available. Changing FFTF does not change fps, it only alters the background loading performance. fps changes are the result of unavailable throughput. When the system has available throughput we notice that the fps doesn't change (fps is always subject to some alteration with any change including a 1 pixel wider view).

Smoothness within the sim changes if we alter the background process ratios with FFTF depending on where we are within scenery - fps changes readily only when the system is already at the limit.

When near to the ground we have less of a view to the horizon, FFTF can be set slightly lower to reduce the proportion of loading new scenery to other background tasks. It does not alter the ratio to rendering time. The simulator always tries to maintain the same fps.

Setting FFTF lower can ignore scenery out of view as it is further away, nearby objects are rendered earlier so reducing the loading proportion results in scenery further away rendered later. If in this case fps rises it's because the system is already at its peak before changing the value."

[iFlySimX 751]

Interesting thread....so could increasing the FFTF help get rid of blurries? I’m I understanding it correctly?

[Michael Moe 495]

24 minutes ago, Le Français said:

Interesting thread....so could increasing the FFTF help get rid of blurries? I’m I understanding it correctly?

IT does. I bought Dynamic FFTF and set it to 4500 feet 0,4 and 0feet at 0,01 and the blurry testing KLAX area are gone. I run UNLIMITED@25HZ VSync on

Not sure about autogen load though but the app does work. 

Thanks Michael Moe 

Edited August 14, 2018 by Michael Moe

[GSalden 4,656]

5 hours ago, Michael Moe said:

IT does. I bought Dynamic FFTF and set it to 4500 feet 0,4 and 0feet at 0,01 and the blurry testing KLAX area are gone. I run UNLIMITED@25HZ VSync on

Not sure about autogen load though but the app does work. 

Thanks Michael Moe 

Michael is correct.

When making flights I always use it . Max fps on the ground with heavy airports and more time for terrain loading up in the air...

[David Roch 1,982]

A quick question.

Why 0,4 and not 0,33?
What is the benefit to increase this value and what is the max?

[GSalden 4,656]

1 hour ago, David Roch said:

A quick question.

Why 0,4 and not 0,33?
What is the benefit to increase this value and what is the max?

The higher the value the more time for terrain loading . However, less time for achieving high fps...

Max ? Probably 0.99 , but then your framerate is at the lowest point.

[David Roch 1,982]

Crystal clear, thanks Gerard!

[iFlySimX 751]

So that program is really worth it?

[AnkH 2,261]

1 hour ago, David Roch said:

Crystal clear, thanks Gerard!

Not at all, it is still an open question if the sim can recover from the autogen loss when using the FFTF Dynamic Tool. Rob did his test with a blank P3D, no addons installed. The open question is: if you use FFTF with a value below 0.33 while taking off, then 0.33 or more when above 4000ft, will this still result in proper autogen drawing after some miles and minutes or is the autogen drawing initially hampered by the low FFTF value and never recovers? Can you shed some light, Gerard?

[Chris Bell 165]

2 hours ago, GSalden said:

Max ? Probably 0.99 , 

4.0

[GSalden 4,656]

14 minutes ago, AnkH said:

Not at all, it is still an open question if the sim can recover from the autogen loss when using the FFTF Dynamic Tool. Rob did his test with a blank P3D, no addons installed. The open question is: if you use FFTF with a value below 0.33 while taking off, then 0.33 or more when above 4000ft, will this still result in proper autogen drawing after some miles and minutes or is the autogen drawing initially hampered by the low FFTF value and never recovers? Can you shed some light, Gerard?

You can set the FFTF value to your own system. Eg 0.01 on the ground and 0.4 at 2000 ft and above. Then your will have a longer terrain loading time whn flying low, but at the cost of max fps.

However, you probably would be away from that heavy airport with AI traffic anyway, so the framerate would op up.    

 

[jabloomf1230 2,445]

41 minutes ago, Chris Bell said:

2 hours ago, GSalden said:

Max ? Probably 0.99 , 

4.0

Correct. As stated a few times before, FFTF doesn't range from 0 to 1, but rather from some small number (0.01) to integer values like 1, 2 or even 4. Since FFTF only a limit, there are diminishing returns of setting FFTF at an integer value, as there may never be such a high demand in the sim for scenery texture loading. Hence, such a high value will have no effect.  As Gerard has stated, it's best to set the high FFTF value at a number that is just above the default (0.33) and then see whether that improves matters. I originally ran FFTF Dynamic  in a range of 0.01 to 0.4, but after experimenting, I have been using 0.2 to 0.65 and 7,000 AGL. Each setup will need its own experimentation to see whether autogen and texture loading starts noticeably lagging.

All this "goodness" (AA, mipmapped textures, model multiple LODs, "the blurries", autogen not loading, etc.) is left over from the ACES team trying to keep FSX usable even when the hardware is stressed. Without these shortcuts, portions of the screen would end up rendering as black. With these tricks, at least the sim continues to look vaguely unobjectionable. In contrast, LR (although they use some of the same time-honored tricks) just set a lower FPS limit of 20. When that is breached, XP11 goes into "slow motion mode":

https://www.x-plane.com/kb/the-simulators-measurement-of-time-is-slow/

 

Edited August 14, 2018 by jabloomf1230

[Dirk98 354]

27 minutes ago, jabloomf1230 said:

I originally ran FFTF Dynamic  in a range of 0.01 to 0.4, but after experimenting, I have been using 0.2 to 0.65 and 7,000 AGL. Each setup will need its own experimentation to see whether autogen and texture loading starts noticeably lagging.

And what are your TFR and VSYNC settings in P3Dv4 with the above?

Thanks.

[Dirk98 354]

15 hours ago, Nyxx said:

I run locked at 26, plus FIBER_FRAME_TIME_FRACTION=0.01 for FSL and locked at 31, plus FIBER_FRAME_TIME_FRACTION=0.01 and I get perfect texture loading and no stutters. VS  and TB ON, my refeash is 120hz. So it runs at 24 for FSL and 30 for PMDG.

What does the P3Dv4 science say about the above aproach? Doesn't FIBER_FRAME_TIME_FRACTION=0.01 negate TFR=26 locked fps benefit?

Thanks.

Edited August 14, 2018 by Dirk98

[jabloomf1230 2,445]

28 minutes ago, Dirk98 said:

And what are your TFR and VSYNC settings in P3Dv4 with the above?

Thanks.

FFTF Dynamic also has a frame rate range target, but in Prepar3d.cfg, the frame rate limit is set to 32. Vsync is on, but with a 60Hz monitor, I've never seen any difference, either on or off.