I'm using 4K with 8MSAA and x2 Sparse Grid SS in NI and can confirm that a few 360 pans recovers almost all of my lost frames due to starting P3DV4.1 with DL on.
So at default YSSY in the Carenado P31 parked at a gate frames jump back from 25/25 to 55 in the VC and 65 in spot view. Frame rates with DL off are 55/67 here.

Also it does not matter whether I pan in VC or Spot View the results are the same.

After cursing DL for the last few months this is really amazing and might be the find of the year for a lot of people. Additionally, as mentioned above, it suggests that DL frame rate loss may not be a major hardware issue at all but a fixable sim software one?

gb.

 

I tried the 360 last night and was amazed by the results, with the PMDG 777 at flightbeam kiad I went from ~18fps up to the low 40's.  Hopefully LM can figure out why this works and use it to tweak dynamic lighting.

36 minutes ago, regis9 said:

I tried the 360 last night and was amazed by the results, with the PMDG 777 at flightbeam kiad I went from ~18fps up to the low 40's.  Hopefully LM can figure out why this works and use it to tweak dynamic lighting.

When there's such an easy fix available that makes me wonder how much time LM invested so far in trying to address this issue affecting a rather huge amount of people.

7 hours ago, OzWhitey said:

To be fare to OP, I've been simming since the subLOGIC days, and this is literally the first time I've heard of the spot-view-and-spin theory. I don't think it's SOP for standard P3D use!

I'm interested that it seems to work well for some people here with the DL, there's clearly a group of simmers with substantial performance issues in this regard.

I'm currently very performance orientated (VR simming), so I was actually testing the frames with dynamic lighting last night with the PMDG QOTS II. I can see no - zero - frame rate impact with all the lights on, and landing lights illuminating ground selected (achieving 80 FPS with the PMDG bird in this setting, obviously though with some sacrifices in the slider department!)

My suggestions are:

- leave nVIDIA settings in default, don't try to apply AA other other processing outside of P3D.

- use latest graphics drivers.

- buy a ninja graphics card if you can. It was painful, but I bit the bullet and bought a 1080Ti. Whilst I've rarely gone top-end with this rapidly-changing technology over the years, our new sims can push even this cutting-edge hardware to its limits.

- be aware of what your dynamic lights are shining on. I'm not using payware airports, I believe some can cause performance drop when DL is used. I fly with home-made airports that allow me to balance visuals with performance.

- further testing of the spot view-and-spin methodology is recommended. I'd be interested in LM's take on this - does it consistently improve early performance, and if so, how can the program be optimised so that this rather cumbersome approach is not necessary? As noted though, I've never done this and am generally very happy with my sim performance, including during my current flying with VR+dynamic lighting+PMDG aircraft.

Cheers,

Rob

 

Thanks Rob.  I was starting to think I was nuts and missing out on some crucial step of gaining performance.  I do use payware airports but I leave NVidia settings alone and I'm on a 1080 (not Ti).  For the most part non VR flying is between 30-60fps and in the rift with FI I'm almost always at 90.  

5 minutes ago, pegruder said:

Thanks Rob.  I was starting to think I was nuts and missing out on some crucial step of gaining performance.  I do use payware airports but I leave NVidia settings alone and I'm on a 1080 (not Ti).  For the most part non VR flying is between 30-60fps and in the rift with FI I'm almost always at 90.  

Here's another one being in this hobby for a fair amount of years and never started my flying sessions with panning around in different views so you're certainly not alone.

2 hours ago, regis9 said:

I tried the 360 last night and was amazed by the results, with the PMDG 777 at flightbeam kiad I went from ~18fps up to the low 40's.  Hopefully LM can figure out why this works and use it to tweak dynamic lighting.

May be worth posting this feedback in the LM forums - certainly may expedite any development of a "fix" if this can be replicated (which it seems to be).

8 hours ago, DylanM said:

May be worth posting this feedback in the LM forums - certainly may expedite any development of a "fix" if this can be replicated (which it seems to be).

Just posted over there.

(https://www.prepar3d.com/forum/viewtopic.php?f=6312&t=127125)

gb.

This is one thing that I always did in FSX on the start

it was one thing in the NickN bible and I have used it ever since

This has been a must do for anything using the ESP engine. There was some documentation about it from Phil Taylor of Aces back then as a releif from certain stutters and fps loss at startup. NickN's bible and prior posts stressed the benefit of making a quick spot view spin.

As to why LM hasn't done anything to fix it - I'd speculate it's on a LONG list of fixes to be done down the road. Considering there is an easy workaround, I'd doubt if it were deemed a "critical" move.

Vic

10 hours ago, DylanM said:

certainly may expedite any development of a "fix" if this can be replicated (which it seems to be).

FYI for all, GPUs have texture units and shader units, texture units have their own L1 cache and as such need to be "populated" ... the 360 pan is a great way to populate, next time a shader needs to read a texture their is a good chance it'll be available on the GPU and no need to do a much slower process and fetch it from the CPU and down the chain to last stop being I/O (texture from file).  The GPU's texture caches are considerably faster and the shaders will be able to process DL calcs that much faster (shaders have a cache also for instructions).

Of course, at some point the texture cache will be flushed with another new texture for another new scene rendered - texture caching has been around for a long long time in GPUs.

LM have not given up on DL performance, but even with changes to the render pipeline don't expect miracles on the software side, it's going to require hardware also.

Cheers,  Rob.

46 minutes ago, vgbaron said:

As to why LM hasn't done anything to fix it - I'd speculate it's on a LONG list of fixes to be done down the road. Considering there is an easy workaround, I'd doubt if it were deemed a "critical" move.

Vic

And based on the things that were fixed this go around I'd venture a guess that add-on developers would complain if they did eventually fix it. I'm fine with it knowing a quick fix.

2 hours ago, vgbaron said:

This has been a must do for anything using the ESP engine. There was some documentation about it from Phil Taylor of Aces back then as a releif from certain stutters and fps loss at startup. NickN's bible and prior posts stressed the benefit of making a quick spot view spin.

"Must do" is a a great overstatement, Vic. I'm not the only one around here who has never done this in the last 30 years of simming. 

As for NickN's so-called bible, it contained a lot of non-evidence based statements of questionable veracity, presented as 'must do' items. I always took it with a very large pinch of salt.

So, I don't think that we should broaden this to suggest that everyone should/must be doing a routine pan-around prior to flying in P3D. Even if it did help some back in the FSX days, Prepar3D is now a very different engine.

As mentioned above, the issue here is a little different, as some people are getting MARKED frame reductions with dynamic lighting, followed by clear-cut improvements when panning. Conversely, a lot of people - such as myself - have no significant problems with the P3D lighting tech, and no improvement at all when doing this. If we can work out why this difference is observed, we may be closer to fixing the problem.

7 minutes ago, OzWhitey said:

As mentioned above, the issue here is a little different, as some people are getting MARKED frame reductions with dynamic lighting, followed by clear-cut improvements when panning. Conversely, a lot of people - such as myself - have no significant problems with the P3D lighting tech, and no improvement at all when doing this. If we can work out why this difference is observed, we may be closer to fixing the problem.

I bet if you were in the pmdg 747QOTS2 and a decent payware airport your frame rates will well and truly suffer using DL. I say this as I read earlier on that you don't use payware airports.

When I use default airports with an advanced add on like the pmdg737 and DL my performance is much better.

2 hours ago, Rob Ainscough said:

FYI for all, GPUs have texture units and shader units, texture units have their own L1 cache and as such need to be "populated" ... the 360 pan is a great way to populate, next time a shader needs to read a texture their is a good chance it'll be available on the GPU and no need to do a much slower process and fetch it from the CPU and down the chain to last stop being I/O (texture from file).

Now, I'm not a tech guy, but this doesn't make sense to me based on what I'm hearing people report. Aren't simmers getting low frame rates sitting stationary at the airport? The necessary textures for the render for that position would be loaded into the L1 cache after a certain - very brief - period of time. Going to a different view (outside), and panning around behind the desired view, should only fill the L1 cache with textures that were unnecessary for the initial view, which shouldn't help the frame rate at all when you return to the initial (cockpit) view that you started with.

Panning around would make sense if you're trying to fill the cache with textures that you're going to need when you start moving, I can believe that it would help stutters in that situation in there's a bandwidth problem - but that's a different situation altogether.

Based on what I've read, there's three things that seem potentially problematic:

- using NVidia inspector and/or high levels of AA in-sim

- using scenery that uses older objects

- hardware

Firstly, I note that the poster above mentions NVIDIA inspector settings. I don't use this program any more, I feel it's more reliable to just perform AA processing in the sim. Have the people who are having problems with DL tried disabling this? Have you tried turning your in-sim AA well down?

Secondly, is the problem consistent across a range of airports? I seem to remember Mir making a comment straight after the release of 4.0 about the impact of DL varying depending on what objects the lighting was shining on, and the need to optimise this. 

Thirdly, what hardware is associated with this problem? It's tempting to blame it on low-end video cards, but I've read about people with 1080Ti's experiencing low frames with DL, so it's not so simple.

I flew the QOTS II for about five hours yesterday, I was watching carefully for any drop in frames when I turned the dynamic lighting on, and I still didn't see any at all. So it's not a case that the tech is fundamentally broken, just that it doesn't work well for some people with a given combination of scenery, hardware and settings. The obvious question, then, is just how wide-spread is this problem, and what are the common factors.

 

 

 

 

1 minute ago, branks737 said:

I bet if you were in the pmdg 747QOTS2

Lol. That is EXACTLY what I'm flying! Whilst I'm pretty new to this great aircraft, I've done very little over the weekend except fly PMDG's masterpiece.