Microsoft Flight Simulator 2020 in 2021
Interesting, thanks!
I like this part:
…
“MS Flight Simulator 2020 has been a bit of an enigma when it comes to the sheer low performance. We know the GPU is hardly the problem; we also know the CPU is not the problem, we measured that 16, 32, and 64 GB system memory makes no difference. So that leaves the DX11 game engine, game complexity, and programming. The sheer scale of the sim is also a massive factor here. However, if we analyze the frametime behavior, we see weird things. The same can be stated for CPU utilization, as clearly the sim finds itself most comfortable with a handful of cores that run faster over many cores.”
…
2 Likes
" … 16GB really is enough …"
I have some doubts 
… in special they proof that with a screenshot of used memory , but have 64GIG installed… they simple forgot the commited memory, etc… see all the issues for users <32GiG RAM.
I can also give him a proof:
I think some users in existing “limited by main thread” discussions possible want not believe that 
In special within the airplane, the CPU have hard work… outside the GPU is more stressed.
And they forgot to mentioned the one-core issue…
It’s just like the most of “these” perf-tests… a good starting point for, but the thruth can you find only within the forums 
I agree 
But in fact, when monitoring CPU usage with “Limited by Mainframe” we can see some pretty weird results that I never saw when benchmarking other apps.
E.g. In some testing situation I can see the “Limited by Mainframe”, having my GPU under 70% usage, and my 8 CPU physical cores (when doing tests with Hyperthreading OFF) don’t exceed 70/80% usage each.
So where does those CPU “headroom” goes?
IMHO I think this is what they tried to express here by saying “We know the GPU is hardly the problem; we also know the CPU is not the problem…So that leaves the DX11 game engine, game complexity, and programming”.
I thought about the monitoring tools granularity (one check per 500ms, 1s, 2s, etc.) not been enough precise and missing CPU spikes. It’s probably true for some stuffs like the instruments refreshing panels (settings you can change from low to high in menu) as they don’t appear clearly in core usage curves. Hence a great post from CptLucky8 asking for better dispatching such management.
Anyway, we have other threads talking about “Low cpu and gpu usage in MSFS” and I don’t want to reinvent the wheel (and don’t feel I can)
1 Like
na I think it is because the one core issue,… the CPU can not fully utilized and so here graph show “70%”. In contrast to the GPU. But might be you are right that they would mentioned this with the DX11 hint.
yes… again correct and same for me
I was also only a bit shocked about the “16GIG is enought” . This is definitiv not the case… Of course the game run , but all memory above these 16gig ( normaly range till 32GIG commited ) comes then from the pagefile and this kills “in the long run” the SSDs…
1 Like
I always monitor all core separately, and with HT OFF when doing benchmark, so it’s not a 70% CPU average usage. Some power are lost, but could be granularity as I said : monitoring tools don’t catch quick spike usage of one or more cores…
Yeah the memory I agree, I saw it when upgrading to 32GB.
1 Like
RX6800 vs. RTX 3070
This review benchmarks them equally at 1440p ; 47 fps .
If running an AMD CPU would you go for the RX6800 ?
Doesn’t really matter for MSFS. Seeing as both are pretty hard to find, just go with whichever one you can actually get your hands on for a reasonable price.
1 Like