It took some time, but because I started this poll I wanted to come back on this topic now 🙂
First off all, thank you very much for voting and the great discussion! (until now 196 members took part!) 👍
I know that this poll was not as complete as many suggested, but I think it revealed the most basic settings we're using in our sim right now.
I'm aware this all depends on the P3D-version you are using and your monitor refresh rate and so forth...but I did not want to bloat up this poll.
I've seen a lot of different settings people are using and of course every system is different, with various addons and hardware, but I think I identified four major configurations which are used very often.
They can be split into two groups:
1) Monitor refresh rate < 60Hz (or monitors with any other sync-feature)
If you're lucky and your monitor supports it, you can set the refresh rate to 30Hz or even lower and then use VSYNC to match the fps to the monitor's refresh rate.
2) Monitor refresh rate >= 60Hz
I would guess the majority does not have the ability to set their refresh rate to 30Hz or below, so in that case there are three major configuration I've seen:
a) Use the P3D internal-frame-limiter
A lot of people are using the internal limiter (~31%).
Also a lot of other people said it's rubbish and did not understand why it is used at all.
So I must admit I've used the internal limiter for quite a long time too, until it did not satisfy me anymore.
b) Use the NVidia-frame-limiter
Another 25% are using the NVidia Limiter. Wether via NV Control panel or via NV Inspector.
I have not used it yet, but planned to try it, but more on that later.
c) Use RTSS with the scanline x/2 option
This tool seems to be a bit undervalued, because at the moment only <5% are using it to limit their frames.
I've been using this tool for some time now. For me the option "scanline sync x/2 = -1" was performing quite well.
I'm sorry, but this is going to be a bit longer 😂 🙈
With these options in mind I wanted to do a comparison to find out, which configuration works best.
For a real comparison however, you need a measured value to be able to quantify the differences.
For that purpose I used a tool called "CapFrameX", which is absolutely great for measuring frame times.
So this was my test setup:
Intel i7-9700K @4.9GHz
NVIDIA RTX 2080 8GB (Palit Super Jetstream)
32GB RAM @3200MHz
2x M.2 SSD Samsung EVO 970 (OS + Sim)
Three Full-HD monitors using Nvidia-Surround @60Hz (only 60Hz available)
OS and NVidia-driver
- NVIDIA driver 446.14
- Windows 10 - 64Bit - 1909 - Build 18363.836
P3DV5 - HF1 (with only selected addons)
- ORBX Global Base/OpenLC NA/Building/TerraFlora2
- FS Global Mesh 2010
I also tried with all addons disabled, but the results differ not that much.
- no AM in use (using all 8 cores - 8 full cores, no HT for the 9700K)
- my settings are very reasonable and not every slider right (maybe I add them later, but I recognized they did not play a major role regarding fps - as long the system is not overloaded of course)
- No cfg-tweaks in place
- Absolutely no traffic
- Weather: only Clear Skies
- Using a Standard Aircraft - Mooney Acclaim
I recorded a flight in P3D, so I could load this recording and use the exact same flight for all captures.
Taking off @KFUL, flying some circles and then landing @KLAX (around 13min flight time)
As I only have monitors with 60Hz, I had to focus on the second group 2) with config a), b) and c) mentioned above.
I tried a lot of different settings and variations, so I will pick the most representative for evaluation.
Here are some basic findings:
- There are always some little stutters, which I could not get rid of.
Even with disabled addons and all sliders left they are still there, no matter which configuration I tried.
It's not happening every second, but appr. 6-7 times during the 13min flight. (remember it was LA I was flying over)
- CPU (Core0) load varies depending on the configuration (RTSS is always @100%, where an external limit with NVCP was around 50-60%)
I've read about the behavior using RTSS (using up the idle cycles) and it does not seem to harm, so I accepted this behavior as is.
- I also tried the AM-trick (removing core 0 and then set it again using task-manager) -> it seems to spread the load better, but regarding frame times it made absolutely no difference at all (snake oil...ha ha)
- VSYNC and Triple Buffer on or off also makes not much of a difference for my setup, but I guess, that's because of the 60Hz refresh rate.
And here is the table for 5 different configurations - using the standard deviation for comparison:
So when looking at the standard deviation of the FPS, the configurations using RTSS are definitely recommended.
I tried different combinations with unlimited internal, using the P3D limiter@33 and the combined option with the RTSS limiter @24fps (no big difference).
Interesting is, that the NVCP-limiter is even worse compared to the internal limiter.
And here come the diagrams (showing FPS during the 13min test flight):
Here we can see, that the NVCP-limiter has much more fluctuations compared to the internal limiter.
Also visible are the occuring stutters, because they are always at the exact same position (causing long frames with 6-15fps)
I would guess, this is some kind of loading issue during the flight (maybe an airport coming into range).
Looking at these results I would definitely recommend using RTSS with the "scanline sync x/2 = -1" option for users with a 60Hz monitor (limiting internally or not has to be tested, but it's no a big difference...maybe autogen is loading better with limiting in sim).
The NVCP-Limiter is not doing a great job, so I would rather stick with the internal limiter in this case 😜