Wonder how long until we will depart from one core usage in this sim. Is that being worked on ?
With so much going on in the background, traffic, photogrammetry etc., it will help a great deal.
2 Likes
They are already using more than one. The problem is that realtime applications are usually bandwidth/latency limited.
You can do lots of things on the CPUās but that information has to make it in realtime to the screen. In other words, a flight sim is not trivially optimizable by throwing more cores at the problem.
Optimizations of this kind of application require more thought and are usually based on clever data reduction mechanisms. Distance based level-of-detail is one of those techniques.
If you could find a way to do reductions of geometry on the CPU, than that could be parallelizable.
The application could also offload work to the GPU if it is maxing out the CPU, but that creates a big divide in your application. Two modes with their own class of bugs.
3 Likes
Does this sim, on PC (not XBOX) , implement Direct Storage 1.1 ?
From what I have seen in tests, Direct Storage eliminates the need for time-consuming decompression via the CPU. This can unload the CPU significantly as graphics data is transferred directly from the SSD via RAM to the GPU to be processed.
While Direct Storage comes pre-installed on Windows 11, its usage is only optimal if games were programmed using Direct Storage.
Has Asobo implemented Direct Storage in MSFS for PC ?
2 Likes
MSFS is in no ways tied to a single core except mainthread which is where it all gets put together before being dispatched to the gpu. Iām sure you will find that all MIMO games work in this manner and for any MIMO app not to do so would only introduce more sync issues and more latency than necessary.
2 Likes
MSFS is unpacked to ram and vram and runs from there, except for initial load times DS is unlikely to make much difference as NVMe 4.0. transfer speeds are actually no quicker.
1 Like
Periodically someone asks this question or a variant of it. There does seem to be this persistent idea that somehow it should be possible to max out all the cores of your CPU with just the sim and get a massive performance boost, if only the developers would āfixā the sim. The problem is that concurrent programming just doesnāt work like that. The need for a single thread that coordinates the activity of all the others - and thereās just no way around that - will always impose performance limits due to latency, order of work, task dependencies and frame timing.
Iām quite certain things could be improved - they always can, because humans are just not great at concurrent programming - but the persistent idea that there is some huge reservoir of unused CPU power that can be pressed into use is as much of a myth as the idea that we only use 10% of our brains.
13 Likes
Was looking recently at DCS announcement re splitting the work for the engine to reduce the workload on the GPU. Wonder if something similar can be done on MSFS ?
To a great extent thatās already been done. Use DX12 and in task manager right click on the cpu graph selecting logical view. For sure thereās probably a bunch more that can be done but I guess not before MS/Asobo drop all support for 4 core/ 8 thread CPUs.
thanks.. and yes, it just because lots of users have no idea how all that things work, but need to say how worst all is.
I also wish that my car will drive with banana peel , like in the movies, but I have also no idea how a flux-compensator work 
As you Iāam pretty sure there is room for optimizations, but more like to reduce little stutters and not give users 500% more fps with the same hardware.
And more strange for me is, that I have no idea how all that wrong informations about the performance of the other flight-simulators on the market come, where users dreams from that this get us much more fps, etc.. I own them and none of these have really better fps if we compare the level of details.
2 Likes
Usually, the ones who ask about when we are getting multi-core, donāt know anything about programming.
3 Likes
For sureā¦lets run away and stop asking questions then :-). This is what the forum is for and the level of knowledge and expertise varies among the members here. I guess you dont have to be a programmer to be able to readā¦ālimited by main threadā in dev modeā¦and it`s a well known fact that the simulation is CPU dependantā¦primarily. This is why DLSS 3 did so much by taking the load of the CPU.
1 Like
If you look at the per-core CPU usage in Task Manager over different conditions, youāll see that you do get heavy usage of all cores in busy scenery areas, like flying low over Manhattan or Los Angeles.
This is because a lot of work that can be parallelized well is already parallelized well ā but thereās also a lot of work that must be serialized, which can bottleneck things.
There may be additional future improvements, but generally speaking itās already doing a good job at breaking up total computational load in areas that need it (areas with lots of scenery that must be processed on the CPU)
Note that some folks have seen an improvement in main thread times from use of DX12, which promises better ability to break graphics calls to separate threads. (I donāt use DX12 though because memory management is very broken on Nvidia cards with only 8 GiB VRAM.)
5 Likes
it did nothing with the cpu load.. these overhyped frame gen create frames which not real exists, lets say fake frames. This is a completly different story than split āactionsā into separate parts which can parallel computed on different cores. Frame Gen technic compute nothing which is relevant.
1 Like
Nothingās real, itās all fake ā itās a simulator. 
Frame Generation is unrelated to parallelization, but it is a method for increasing frame rates. Itās more similar to Motion Reprojection in VR in how it works.
2 Likes
This is a really bad misconception. MSFS will use several cores, but typically there is one core (the mainthread) with the heavy usage. If you use HWInfo Sensors while MSFS runs, you will see the load bounce around on several cores.
itās not an easy task
if Asobo could just magically make the sim run better they would. to an extent thatās what SU5 was, no?
1 Like
Recently DCS has a Multithreading beta option. The performance increases from 50 to 80 FPS on my computer. So i hope Asobo can do the same with MSFS.
Yes, same here, fantastic improvement in VR with recent DCS MT beta. I also hope Asobo will look at this. Perhaps vulcan implementation might also help but something must be done as performance in VR is not great, the least to say.
since one day now and with a lot of known issues ( therefore its a beta within the beta )
Comparing of DCS code-base and MSFS code-base is like compare apples with nuts.
Also, as DCS developers wrote: if you former was GPU limited, you are still also with the new multithread-implementatin in DCS GPU limited. Thats similar in MSFS. The MSFS developers work also on optimzations, hopefully. The stepwise switch to DX12, which is mainly same as Vulcan ( a low level api ) , is a part of that. Implement fundamendal changes in multithreading is just a dangerous step, like we see in DCS where some parts currently not works ( eg. known the garmin ) or stutters can occur now.
Yes i know, DCS and MSFS are complety different programs, but we can still hope 
.
For the most of the users the DCS beta sim is useable with MT (see the DCS forum), for me even more stable then the beta 12 for MSFSā¦
I love VFR flying in a cesna (MSFS), but also in a Apache and fire rockets 
Also MSFS is on my computer with a intel 13700K still CPU limited and not GPU limited (3080TI)
Still my CPU is only at 25-30% (task manager showed that the hypertreaded cores are not used). My GPU is mostly at 70%.
In DCS my CPU is almost at 80 procent and my gpu also at 80-90%). So there is much room for improvement.