Maximum pre-rendered frames (i5 users get in)

[MrFuzzy 2,118]

Have you guys done any test with this setting? For AMD cards it should be "Flip queue size".

https://gfxspeak.com/2019/08/26/nvidia-speeds-rendering/#:~:text=Back in 2015 or earlier,The default value is 3.

This is the number of frames that the CPU can render ahead for the GPU to process and values range between 0 and 8 in nVidia Profile Inspector.

Zero corresponds to "Ultra low latency" in nVidia control panel and it is recommended to minimize input lag, at the price of a reduced framerate. 

Higher values (1,2,3...) should improve the framerate and the smoothness (frametime consistency) at the price of an increased input lag, which is not that important for a "slow" game like MSFS.

When I had an i5 and a GTX 1080 (CPU bottleneck), I tested this parameter with Battlefield 1 and changing it from 1 to 2 really made a difference (20-25 fps).

I tried to play with this setting and jump into the A320 at JFK but honestly I don't see perceivable differences with my system. 

You may want to give it a try...

Cheers

[Colonel X 1,428]

14 minutes ago, MrFuzzy said:

Higher values (1,2,3...) should improve the framerate and the smoothness (frametime consistency) at the price of an increased input lag, which is not that important for a "slow" game like MSFS.

This is true and definitely worth a try. But you'd think if it was useful, somebody would have noticed by now.

[wthomas33065 994]

I really think for us Kaby Lake and lower i5's the lack of hyperthreading REALLY hurts us.  Kaby Lake i7's can hyperthread which allows the cores to be used more efficiently.  8th generation and above i5's have more than 4 cores, and as such has cores that can do the housekeeping.

Where we run into issues, IMHO, is when we need to load scenery details (usually controlled by LOD detail settings (object and terrain). we have to pull in that data from the HDD or SSD or the internet.  The CPU is already pegged running the sim.  So, something has to give.  People have tried to change affinity and such but the thing is, we just run out of available resources.

This is also why when we start the Sim, we often get terrible stutters at the start of the flight.  We're still processing all that initial flight data.  That can be mitigated by waiting to taxi or take off until you can POV around the cockpit without stutter.  But that doesn't help us midflight when we enter a dense area of terrain objects or mesh.  the ILS 27R approach into KATL is practically unplayable for me at any setting on my i5 7400.

I'm not sure there is a good solution other than staying out of those areas.  I am not ready to upgrade my entire rig for a single title yet, so I've gotten an i7 and will put it in and enable hyperthreading to see if that helps.  

People with more than 4 cores, or i7's with hyperthreading can run background fetches in the background without interrupting the main thread, which is usually what gets limited during those stutters, (look in the FPS counter and it will tell you when your are limited by the GPU or main thread).

I have set my GPU to lock in at 30FPS to provide the maximum consistency in frames and for most of the flight I can change details from low to ultra and maintain consistent 30FPS, The RX580 actually does a pretty good job at keeping up even in Ultra.  UNTIL I do a scenery load due to moving into a new area, Then the thing becomes a stuttery mess and the FPS counter says limited by main thread.

Hopefully my new I7 will mitigate that a bit.  We'll see.

[samcan 37]

I hope that my Kaby Lake i5 is the main issue; I notice that it's frequently pegged at 100%, although I've still done pretty well with it. I'm planning on upgrading to the new Ryzen 4000 desktop CPUs once they're available, so good to know that I should look for more than 4 cores.