GPU upgrade

[V12]

Only for info :
I upgraded GPU from GTX 1060 6GB to RTX 3060 12 GB. As I expected, framerate in FG is not significantly higher, probably 1-2%. Jittering and stuttering is unchanged too.
Result :
GTX 1060 is enough for FG. Later I will upload test video.

[wkitty42]

FWIW: we could have told you this... apparently your system is a bit CPU bound...

FWIW2: tests like this are good but we're missing a lot of relevant information... namely LOD, graphics quality settings, and other similar things that deal with data movement in memory as well as to/from the GPU... with a 12G GPU, your system can certainly handle more graphics... the question is if your CPU can calculate the changes in all that scenery and feed it to the GPU fast enough...

[V12]

Here is the video :

Quality - all shaders at maximum
LOD bare 210 km
Active OSM roads, pylons, railways, OSM buildings
Vegetation very high density
Active AI traffic

P3D runs much better with the new GPU, so I'm satisfied.

[TheEagle]

Well, with bare LOD at 210 km i'd wonder if there's any (affordable) GPU that can run 24 FPS … if you're unsatisfied with FG's performance, turn that down maybe !

[V12]

Well, with bare LOD at 210 km i'd wonder if there's any (affordable) GPU that can run 24 FPS … if you're unsatisfied with FG's performance, turn that down maybe !

GTX1060 can handle visibility limit 200 nm with clouds from horizon to horizon. But in other sim

I do not complain on FPS, but lags and stutters.

[erik]

I do not complain on FPS, but lags and stutters.

I have the feeling this has mostly to do with object placement masks. e.g. where the trees are placed within a texture.
This is done on the CPU.

Erik

[MariuszXC]

I have the feeling this has mostly to do with object placement masks. e.g. where the trees are placed within a texture.
This is done on the CPU.

Is this backed up by benchmarking/code performance analysis?
Dividing the object placement mask into subareas, handled by separate threads, would seem like a standard solution then. That is, if the underlying data structures tolerate multithreaded access..

[erik]

I have the feeling this has mostly to do with object placement masks. e.g. where the trees are placed within a texture.
This is done on the CPU.

Is this backed up by benchmarking/code performance analysis?

No, not yet.

Update:
A simple test would be to go to an area covered mostly by forest (Alaska?) and see if things improve.

Erik