Thanks! I see the little drop-down arrows now.
No CUDA activity on mine either.
I know CUDA is used by video editors like Adobe Premiere to allow parallel processing of certain tasks (as opposed to sequential processing without them) as well as offloading computational tasks from the CPU. It seems like using CUDA cores would be useful for the sim.
Adobe calls it the ‘Mercury Playback Engine GPU Acceleration’ and it has to be enabled.
Maybe they didn’t include it because AMD and Intel cards don’t have CUDA cores. Don’t know about the Xbox.
I think it would be a good thing to include in MSFS 2028.
Yeah, it is pretty basic for AI. The main astrophotography software is called PixInsight at uses AI for noise reduction, star removal and a thing called deconvolution. All of these are trained on large datasets etc and the model applied in the software… it works but is laborious… the CUDA all parallel processing, the simulator may be more sequential than I think (I used to write code for this a decade ago).
And thermals are also a lot improved. I expect higher base/boost clocks voor both new cpu’s as well. But man, januari , that’s way too early … all the money is gone with these holidays
It’s simple. MSFS is primarily optimized for XBOX, which uses AMD with 8 cores. And what’s the best gaming CPU according to all the benchmarks? The 9800X3D—AMD, with 8 cores. So, one can either buy this or keep waiting and analyzing forever
That shows a workload that would likely run in about the same speed in half the CPUs (most CPUs are idle most of the time, meaning anything not the bottlenecking thread has plenty of time to complete).
It’s not that MSFS2020 or MSFS2024 are specifically meant to target a specific number of CPU cores, but that some of the workload is well distributed and other parts are not. You usually end up bottlenecked on one of the single-thread things – the core of the main thread, or the wasm extensions (now on their own thread in 2024 I believe?), or the HTML gauges on the coherentgt thread, or the “manipulators” futzing around with in-game items.
With this bottleneck, it’s less important how many other cores you have as long as there are enough to handle all the threads that need to get their own stuff done and all the multithreaded bulk-operations in the simulation and graphics prep work.
It’s not super surprising that the performance characteristics of the sim fit more or less within the most common gaming processors’ ability to perform. That includes Xboxes, sure, but also the huge number of low-end and medium gaming PCs that are out there.
I get that. My point was that the sim can use all 16 cores. My testing shows the 7950X3D CPU utilization is (by far) best when I assign it to the Vcores only, and everything else to the non-Vcores.
That’s exactly what people mean when they say it’s not effectively making use of more than 8 cores: it’s not faster with 16 than it is with 8. It doesn’t mean that CPUs numbered after 8 can never have workload assigned to them.
I think people are missing the point. When the sim is allowed to use 16 cores it has to share resources with every other process. I wonder what the results would be if I assigned the sim to only 9 cores, or 10, or 12, or 14?
I think it’s more appropriate to say that Windows is not optimized to run the sim better on all 16 cores.
Maybe someone can hack an Xbox and replace it’s CPU with a 16-core processor, then report back whether they see a performance increase. (that’s a joke…)
I’ll pass. I’ve done enough testing to know that the sim runs best when I assign the sim to cores 0-7, and everything else to cores 8-15, assigning priorities to processes the way Saviratrax showed in his video.
Which means the question of whether the sim is using all 16 cores is also irrelevant.
My point was that’s not the case in MSFS (2020). The majority of cases are CPU limited. So no they can’t keep the GPU fed and it’s why you have to go big on the CPU for 2020. Like I said not seen enough data yet for 2024 to call the CPU/GPU split there but 2020 is as I said predominantly a CPU limited title.
Again, my point was and is that as long as your CPU is sufficient to feed the data to the GPU when the GPU needs it, that’s all you need. My experience is that with both a 12900K and now a 14900K, either processor is capable of meeting that task, The 12900K has an average score of 41,400 on Passmark, so any processor that scores higher than that (and perhaps some lower than that) should be sufficient for MSFS 2020 and 2024. Even the Ryzen 7 7800X3D or the i7-12-700K at just over 34,000 Passmark points each are capable of not limiting the GPU’s ability to produce max FPS.
What you are saying is common sense really and I’m not questioning that. It was your statement that the CPU is less important than the GPU that I was commenting on. If that is what you want to believe or your own personal experience then fine but I’ll tell you for specifically 2020 in most cases the CPU is the limitation and not GPU like you are stating. I have a 7800X3D which before it’s 9 series replacement launched was regarded as the best CPU for flight sim and games in general and I’m still cpu limited in 2020. For the bigger picture the facts that back this up are right here in the forum.
@Sling380@Wdseith
In the case of flight simulators like MSFS the main bottleneck is often the CPU. This is due to the nature of the game, which relies on advanced physics algorithms, atmospheric simulations which require significant CPU processing power. In MSFS, rendering the scenery and flight simulation calculations also put a heavy load on the CPU.
A graphics card like the RTX 4090 is very powerful and at resolutions like 2K or 4K, it won’t be the limiting factor. Higher resolutions, such as 8K, could lead to situations where both the CPU and GPU are heavily stressed. At 8K, the GPU has to process a much larger number of pixels, which could impact performance, especially when paired with the demanding simulation engine.
So, at 2K and 4K resolutions, the CPU is more likely to be the bottleneck in and the RTX 4090 will handle the GPU load with ease.
At 8K resolution, both the CPU and GPU could become performance bottlenecks.
Then I guess it’s time to upgrade to a 5090 when they become available later this month. With a 3090 and the in-game settings on Ultra, I’m definitely GPU limited and not CPU limited with my 14900K/3090 combo at 5120x1440p.
No doubt your understanding of the way the game is coded is much better than mine, but I’ll throw out this observation. While “Main Thread” is frequently listed as the limiting factor when I turn on the in-game FPS display, the load on my CPU never goes above 40%. If the CPU is the limiting factor on my setup, why does its workload only fluctuate between 25 and 35% and maintain a chill 50-60 degrees C while the GPU is continuously pushing 100% usage, 80 degrees C, and its fans are ready for takeoff?
I’m thinking all of us are misinterpreting “Main Thread” to mean CPU when it is not.
There is only one thing that can process the main thread, and it’s the cpu. The reason you aren’t seeing 100% cpu utilization is because the game isn’t optimized to do that, nor should it. It’s unreasonable to expect 100%, your cpu has to do other things like run the OS and there is only so much optimization the devs can do before they reach diminishing returns, especially when so much has to be simulated. The gpu has one job, and it’s to render the image as fast as possible. So it’s much easier for the gpu to top out before it runs out of performance or hits an fps cap.
, that’s way too early 
… all the money is gone with these holidays 
