Hi,
As the sim is more multithreaded, I haven’t really seen any dedicated reputable benchmarks or discussions regarding if having more cores with benefit the sim more over less with Vcache.
Average framerate, 1% lows , utilization, smoothness all in one, if anyone has any info or personal experience, please feel free to post as this year I’ll replace my 7700X
Moved to User Support Hub as OP is requesting community advise and assistance.
this is a bit above my paygrade, i’ll let one of the hardware heavyweights take over, but for the little i understand of the subject the main benefit of the v-cache when applied to msfs is to nullify the system ram bandwith bottleneck, making it as if you had lightning fast ram
can’t say if the number of cores has any relation to that
What I can say from testing around the Los Angeles PG area on my Ryzen 5800X3D with the 8 cores is that I can easily fill all 8 logical cores with SMT off flying low over LA in a small G1000-based plane at 95-100% usage per core, but have only occasional spikes that far if I turn SMT on and have 16 logical cores.
This is at 4K native-size TAA Ultra settings, sometimes stressing it by forcing TLOD higher (which tanks frame rate under 30 in the busiest areas)
More generally at 16 logical cores it’s floating around 50-70% usage across all cores with the same scenarios, sometimes spiking up to 100% perhaps as it’s processing new scenery.
I suspect this means that you can expect to be bottlenecked on the main thread still if you go to 16 physical cores too (eg a 7950X with no CCD restrictions on MSFS).
Whether that still ends up beneficial from the bits that are multithreaded being faster, I’m not sure.
I have a feeling the CPU boost speed is critical for the main thread.
I’ve seen some screenshots from a 16 core x3D CPU where FS2024 didn’t saturate any cores beyond 8, possibly because Windows or a driver assigns the FS24 threads to the vcache cores only, and going cross-CCD would hurt performance. This is just anecdotal and largely speculation though.
But it does makes sense when you consider that the Xbox only has 8 cores.
Would be interesting whether it still behaves this way with a 16 core CPU that doesn’t have any vcache cores. I’m sure someone here could easily test that.
Take a look at this thread.
Limited but something anyway
On my 5800X3D I can saturate all cores flying low over Los Angeles at 4K Ultra setttings in a GA plane with SMT off (8 logical cores) but not with SMT on (16 logical cores) except for short bursts which look like scenery loads.
This makes me think that going to 16 separate physical cores of equal performance still won’t saturate all processors. It feels like it can make much better use of the additional space for processing scenery but it can still bottleneck on the main thread, coherentgt thread, or “manipulators” and additional cores mostly add to your idle time beyond “about 8”.
Begotten poet’s testing answers this. He used process lasso to park cores on a 7950X3D. You can see that even using all 16 cores it does not give as good as fps as just using the 8 cores with vcache. It seems the best cpus are the 8 core AMDs with vcache, just like 2020.