Interesting video regarding the poor use of CPU and why CPU bottleneck is going to be unavoidable with the current GPUs if this stays the same.
I’d like to know if MSFS is using the full CPU potential at this moment, I thought it was but apparently it isn’t.
Thoughts?
The methodology is not good IMHO. When you work on CPU monitoring you try to create a situation where the GPU don’t interfere with results.
So he should have set the Graphics options including render scale and resolution to the lower possible, and certainly not all at ULTRA like he said.
Here we have a situation where we cannot be sure the 60fps is the maximum due to the CPU or due to the CPU+GPU (as it said ULTRA it could mean also render scale at 200%, on 1080p display it mean nearly 4K).
I guess we will have seen a greater value than 60fps at first, and then going down at each steps (possibly).
That don’t mean MSFS use only one or 2 cores. That mean it can work fine with one of 2 core. I need to retest this at it’s a long time I did it, but I can see MSFS been dispatched more than that on my 8 cores/ 16 threads.
MSFS will utilize all the cores of your CPU, if there’s enough workload for it (especially around heavy photogrammetry areas, since the CPU handles the 3D stuff).
The main thread, which syncs up all the tasks offloaded to the other CPU treads again, will be the bottleneck for many users, that’s why it still helps to have a CPU with good single core performance.
here’s my 5800X working hard in NY photogrammetry with stupid high LOD (on DX11, a couple of months ago);
MSFS (or any game or sim at that rate) is however not a workload that’s suited for asynchronous multithreading (like rendering, 3D modeling, video editing etc). So you’ll likely never see a game/sim utilize all cores/threads on your productivity focused CPU.
yeah I’ve been seeing this in my i7 9700k so I thought MSFS was indeed using all cores but the results of that video are strange tbh
If you do a test like that, please share your results 
This is incorrect. You are viewing performance by thread, not by core, and you are confusing background Windows tasks with MSFS tasks. Many of the threads on your charts are just background Windows tasks while gaming.
No, you’re incorrect. This is the background windows tasks (browser and discord). As you can see, those background tasks don’t put in dent into my 5800X (2% load)
2 threads per core, and MSFS uses all cores, and all threads.
Well, it’s all about the test scenario;
he said he was sitting still on the runway in a default Cessna 172. So all 3D information had been processed already, at which point the CPU doesn’t have that much to do anymore. He should rerun his test while actually flying, preferably in an airliner, over photogrammetry.
That way the CPU has to calculate the plane’s systems, the 3D scenery loading in/out while flying, the aerodynamics etc etc.
This is also while completely idle. There is more going on in the background while running MSFS besides the main MSFS threads, including network activity, and other Windows thread schedules. The Windows scheduler can shift usage around as it deems necessary for performance improvement, making actual thread usage harder to visualize.
This topic has been discussed many times on this forum, and you’re certainly not reinventing the wheel here. I would caution saying MSFS uses all cores, as that’s not entirely true when there is a lot going on in the background that’s constantly being shifted around. What I see on your task manager is only a few primary threads being utilized. Regardless, this core usage doesn’t equate to improved performance.
There are so many things wrong with the testing methodology in this video. And there’s so much information left out. Is it the steam gauge 172 or G1000. DX11 or DX12? Was AI traffic on? Was PG on? Was multiplayer on?
I can’t shake the feeling that the point of the video was to show that XP12 scales better with more cores than MSFS does and the methodology was chosen to try to show that.
I agree the testing methodology is close to non existant but the conclusion was that both sims behave the same way, check the numbers for XP, he also made a video for the XP part.
As I said he needed to get ride of the GPU in the equation and he did the reverse.
Also, the way I tested this sometimes ago was by using Process Lasso or a similar app (can’t remember). You see all cores used for a specific application, and you can restrict the use of core one by one until only one stand.
This is the way (IMHO)
Edit: An other person did the same process than me and created a nice report posted on this forum, it was months ago when Asobo changed the CPU management to use more core. I’ll try to retrieve his report.
I know I’m not inventing the wheel here, but if MSFS is the only other task running, it’s MSFS that’s loading up the cores. Of course there’s all sort of management going about which core/thread does what (most handled by MSFS).
About the performance; already covered that as well. the main thread will always be the limiting factor on the CPU, and you can do nothing to help that. The rest of the load will be spread on all available cores. If you have a 12core/24 thread CPU, the load will be spread out over most of those, and you’ll see very low utilization of those. If you have a 4core/8thread CPU, you’ll see the load being spread over those 8 threads. More cores does not mean more performance (like with any non-asynchronous multithreaded workloads). And yes, after a certain point (I would say 6core/12thread) there’s not much if any performance benefit to be seen. All logical, and part of the type of workload MSFS is (and all other games as far as that’s concerned).
Many use cases in MSFS (like the one in the video above, sitting stationary on a runway in a 172) are likely fine on 4 threads. Start flying fast over photogrammetry in a complex airliner though, and you’ll see utilization of the other threads shoot up (like in my screenshot).
If you want to test this yourself, open up the usercfg.opt file, change the LOD factors from 2.000000 to 8.000000 and watch your CPU struggle. And no, that’s not ‘windows background tasks’.
Seeing that that guy has done a lot of stuff together with Austin, I would not be surprised.
fully aggree… we have some topics where users mention that MSFS have only one thread, use only one core, others comes with 4 threads, etc… It so simple to open the Taskmanger and see the amount of threads which can spread around the cores. The problem is , some threads have to wait for each other and of course the well known dx11 limitation with the “mainthread” ( a problem what not only MSFS have ). But if MSFS have the chance, the cpu is fully used - also with dx11.
I9-9900K, HT off, 5.0 all cores
2080TI at 4K capped to 30 fps
LOD is way up, I use Vancouver by ORBX. I recorded “nothing” for more than 4 minutes, but there are no such “background tasks”. My PC is for gaming only.
4:20 into the video I run MSFS, the only other task running is Afterburner.
So, CPU usage comes from MSFS only, right?
I will not enter the methodology of the test on the use of CPU cores but I can say that I recently upgraded my PC by mounting an i5-12700K with an RTX 3090.
When I run some graphics benchmarks I see that GPU reaches 95% with a significant increase in temperatures and consumption of WATTS.
When instead I run in MSFS with ULTRA settings, the temperatures and WATTS go up much less and the GPU never goes through 60% load.
I’m curious to see if the i9-13900K when available, will be able to engage the GPU over 60%.
This calls for a Systools suite using Promon64 to see what tasks Msfs2020 is really using.
Let’s have a standoff between the two sims.
This is likely from being CPU-limited. Turn down terrain LOD to the minimim and increase render scaling >100%. Eventually, you will hit 100% GPU usage provided there is no CPU bottleneck.
