I believe MS said that the new coding and multicore support was going to provide massive performance improvements over FS2020, even with the rigs we are currently using.  I thought I could remember them quoting figures like 2x or 3x more FPS.

I hope that also applies to PCs with standard quad core CPUs (four cores/four threads).

Edited June 23, 20241 yr by Christopher Low

2 hours ago, cianpars said:

I believe MS said that the new coding and multicore support was going to provide massive performance improvements over FS2020, even with the rigs we are currently using.  I thought I could remember them quoting figures like 2x or 3x more FPS.

Uhmm, if anybody has a link to this exact statement by someone on the MSFS team, I would love to read it.

I say this because MSFS 2020 has already gone through several rounds of optimization. And they have done wonders in those several rounds of optimization over the last 4 years.  I would be very, very, surprised if MSFS 2024 can generate up to 3x more FPS than MSFS 2020. Heck, I would be surprised if it could generate 2x more FPS than MSFS 2020. And remember that the ground is much more detailed in MSFS 2024, with rocks and pebbles, etc, which are essentially more polygons that will reduce the FPS.

I think even in the best case scenario, with the extra detail on the ground, that MSFS 2024 could perhaps give 20% more FPS than MSFS 2020. 20% more FPS is just an improvement of 1.2x.  But improving FPS by a factor of 2x or 3x, well, I am very skeptical of that. I would love to see the link to that exact statement, if someone on the MSFS team did say that.

Edited June 23, 20241 yr by abrams_tank

33 minutes ago, abrams_tank said:

Uhmm, if anybody has a link to this exact statement by someone on the MSFS team, I would love to read it.

I say this because MSFS 2020 has already gone through several rounds of optimization. And they have done wonders in those several rounds of optimization over the last 4 years.  I would be very, very, surprised if MSFS 2024 can generate up to 3x more FPS than MSFS 2020. Heck, I would be surprised if it could generate 2x more FPS than MSFS 2020. And remember that the ground is much more detailed in MSFS 2024, with rocks and pebbles, etc, which are essentially more polygons that will reduce the FPS.

I believe this was more about the reduction in processor usage rather than fps. I know there was a lot of discussion about the subject at the time.

2 hours ago, cianpars said:

I thought I could remember them quoting figures like 2x or 3x more FPS.

no way. don't get them all excited with such far fetched figures without links to the source. I would be surprised/impressed if they could squeeze out another 20%-30% percent increase.

1 hour ago, abrams_tank said:

Uhmm, if anybody has a link to this exact statement by someone on the MSFS team, I would love to read it.

At last year's FS Expo interviews, they spoke about more efficient multi-threading and fuller usage of CPU and GPU for various calculations including physics in MSFS 2024, and in turn being able handle all the extra computations coming in MSFS 2024. The 4x to 5x performance increase is in the area of the core physics calculations, not overall FPS.

https://www.helisimmer.com/interview/msfs-2024-helicopters-future-msfs-2020

Sergio: Sebastian talked about many of the changes and one of them had to do with multi-thread improvements. And CFD (Computational Fluid Dynamics) highly benefits from it. In case you don’t know it, CFD is the basis of Microsoft Flight Simulator’s flight dynamics so if it gets improved, flight dynamics will get better.

Seb: "You know, CFD is very important for helicopters. And so with the new system, we're very much multi-threaded, taking work of the main thread about all the simulation aerodynamics works.

The CFD can be a lot more detailed than it could be for, without an impact on the FPS. Because basically, we stopped at the detail we had with the current implementation, because if you go further, it slows the FPS down, and we didn't want that.

But currently, you're sort of limited in how far the ground effect goes, because that's the limit of the CFD, and how accurate the vortex effects can be. And so, for example, on a dual rotor helicopter, if you take the same CFD and you stretch it, you sort of, it is so big that with two rotors, you can't capture the, the donuts are sort of almost gone.

And so […] for that, we can now increase the size without a performance impact. And so have a bigger CFD, more accurate, longer ground effects, all that is possible now, thanks to the architectural changes we're doing. And this with new architecture is super cheap, right? It's just offloaded, you don't see it."

Sergio: All that makes a lot of sense. With more computational power, more stuff can be done. But I started to wonder about limits. How many cores can we eventually use for this? Sebastian answered that as well.

Seb: "So, we don't have currently a hard limit set, it's just the limit of the memory. At some point, it's gonna drop the performance, right?

What we measure right now, and we're not finished optimizing, we're already at a four to five times performance increase. And I think we're gonna maybe get 10 […] you can easily add 10 times more surfaces, right? It's very quickly, right?

And so, I think if people add 20 times more surfaces than before, well, then you may be gonna offload, you know, load the other threads. And the more you load, the more you take away compute power for the terrain, or other stuff, right? Which is using thread.

And at some point, it's gonna have an impact. I don't think we're gonna set a limit. We're gonna limit ourselves, we're gonna have recommendations, but just like today, a third-party aircraft can do a lot of stuff. If it wants to load up the machines, it can.

Memory can be a limit because that would crash. But performance, well... There's no limit.

Also, the new aerodynamics is a lot cheaper on memory than before. Like, this is three or four times cheaper, right? In terms of memory footprint. Which is also necessary because otherwise, if you put more stuff, then you run into memory issues. So, we optimize that as well."
 

Edited June 23, 20241 yr by lwt1971

1 hour ago, lwt1971 said:

At last year's FS Expo interviews, they spoke about more efficient multi-threading and fuller usage of CPU and GPU for various calculations including physics in MSFS 2024, and in turn being able handle all the extra computations coming in MSFS 2024.

https://www.helisimmer.com/interview/msfs-2024-helicopters-future-msfs-2020

Sergio: Sebastian talked about many of the changes and one of them had to do with multi-thread improvements. And CFD (Computational Fluid Dynamics) highly benefits from it. In case you don’t know it, CFD is the basis of Microsoft Flight Simulator’s flight dynamics so if it gets improved, flight dynamics will get better.

Seb: "You know, CFD is very important for helicopters. And so with the new system, we're very much multi-threaded, taking work of the main thread about all the simulation aerodynamics works.

The CFD can be a lot more detailed than it could be for, without an impact on the FPS. Because basically, we stopped at the detail we had with the current implementation, because if you go further, it slows the FPS down, and we didn't want that.

But currently, you're sort of limited in how far the ground effect goes, because that's the limit of the CFD, and how accurate the vortex effects can be. And so, for example, on a dual rotor helicopter, if you take the same CFD and you stretch it, you sort of, it is so big that with two rotors, you can't capture the, the donuts are sort of almost gone.

And so […] for that, we can now increase the size without a performance impact. And so have a bigger CFD, more accurate, longer ground effects, all that is possible now, thanks to the architectural changes we're doing. And this with new architecture is super cheap, right? It's just offloaded, you don't see it."

Sergio: All that makes a lot of sense. With more computational power, more stuff can be done. But I started to wonder about limits. How many cores can we eventually use for this? Sebastian answered that as well.

Seb: "So, we don't have currently a hard limit set, it's just the limit of the memory. At some point, it's gonna drop the performance, right?

What we measure right now, and we're not finished optimizing, we're already at a four to five times performance increase. And I think we're gonna maybe get 10 […] you can easily add 10 times more surfaces, right? It's very quickly, right?

And so, I think if people add 20 times more surfaces than before, well, then you may be gonna offload, you know, load the other threads. And the more you load, the more you take away compute power for the terrain, or other stuff, right? Which is using thread.

And at some point, it's gonna have an impact. I don't think we're gonna set a limit. We're gonna limit ourselves, we're gonna have recommendations, but just like today, a third-party aircraft can do a lot of stuff. If it wants to load up the machines, it can.

Memory can be a limit because that would crash. But performance, well... There's no limit.

Also, the new aerodynamics is a lot cheaper on memory than before. Like, this is three or four times cheaper, right? In terms of memory footprint. Which is also necessary because otherwise, if you put more stuff, then you run into memory issues. So, we optimize that as well."
 

Thanks for this source @lwt1971!

Anyways, maybe @cianpars has a different source that he/she saw the 2x or 3x increased FPS from?

Also, I think when Seb says MSFS 2024 can get "four to five times performance increase," he is probably referring to 4x to 5x performance increase on a specific area that is being computed, but not for the overall FPS. For example, the performance for calculating the CFD may increase 4x or 5x, but unless CFD is the child thread that is holding up the main thread, and the next slowest child thread is 4x or 5x faster, that probably won't result in a 4x to 5x increase in FPS.  As an example, let's say the next slowest child thread is only 2x faster than the CFD thread, and the CFD thread is the slowest thread among all the child threads. By improving the CFD thread by a factor of 4x or 5x, that means the CFD thread is no longer the slowest child thread, but the next slowest child thread was only 2x faster than the CFD thread, so the overall FPS will improve by a factor of 2x, not 4x/5x.

Anyways, my overall impression is that MSFS 2020 is already a marvel when it comes to optimization and performance. The reason is, given the level of graphics, the performance of MSFS 2020 is out of this world. On a scale out of 10, if I were to rate the performance of MSFS, considering the level of graphics, I would give MSFS 2020 11/10.  Yeah, that's right, I am saying MSFS 2020 performance is really out of this world, it's better than I possibly imagined.

In general, the higher the level of graphics, the lower the FPS. This is almost a universal rule in the video game industry.  Now as a comparison, you can look at the level of graphics and the performance of the competition. The graphics are objectively worse with the competition to MSFS, and yet the performance of the competition is generally worse.  This is why I give MSFS 2020 an 11/10 when it comes to performance, given the level of graphics MSFS 2020 produces.

So I would be very, very surprised if MSFS 2024 can yield 2x or 3x more FPS, especially because the graphics level of MSFS 2024 is even higher than MSFS 2020, with the increased detail in the ground. I think some of the increased performance in MSFS 2024 from the architectural changes, may be offset by the increased level of graphics, especially on the ground in certain locations where the ground has increased detail versus MSFS 2020. I already give MSFS 2020 an 11/10 for performance, if MSFS 2024 performs 3x better than MSFS 2020, I have to give MSFS 2024 a 33/10 🤣

Would be awesome if @cianpars has an actual source, I would love to check out that source, if there is an expected 2x or 3x increase in FPS.

Edited June 23, 20241 yr by abrams_tank

4 minutes ago, abrams_tank said:

Also, I think when Seb says MSFS 2024 can get "four to five times performance increase," he is probably referring to 4x to 5x performance increase on a specific area that is being computed, but not for the overall FPS.

Oh yes for sure, the 4x-5x improvement is in the core physics CFD calculations, which then allows them to do more of it, and with more precision, which then results in higher fidelity aerodynamics like Seb showed with the A330 (due to the more precise definition of the fuselage/wings/engines surfaces and shapes which the CFD calculations act on). Bottom line, they're able to do more processing without impacting FPS. Side note: currently, increasing the precision of the CFD calculations impacts processing time non-linearly as its time complexity is O(n^3), see https://docs.flightsimulator.com/html/Content_Configuration/SimObjects/Aircraft_SimO/flight_model/aerodynamics.htm.. so this 4x to 5x computational performance improvement really comes in handy.

Also more efficient multi-threading and fuller use of the CPU and GPU are definitely a good thing. I've never heard them speak of 2x or 3x overall FPS improvements, that would require a CPU and graphics card from the future 🙂.
 

4 minutes ago, lwt1971 said:

Oh yes for sure, the 4x-5x improvement is in the core physics CFD calculations, which then allows them to do more of it, and with more precision, which then results in higher fidelity aerodynamics like Seb showed with the A330 (due to the more precise definition of the fuselage/wings/engines surfaces and shapes which the CFD calculations act on). Bottom line, they're able to do more processing without impacting FPS. Side note: currently, increasing the precision of the CFD calculations impacts processing time non-linearly as its time complexity is O(n^3), see https://docs.flightsimulator.com/html/Content_Configuration/SimObjects/Aircraft_SimO/flight_model/aerodynamics.htm.. so this 4x to 5x computational performance improvement really comes in handy.

Also more efficient multi-threading and fuller use of the CPU and GPU are definitely a good thing. I've never heard them speak of 2x or 3x overall FPS improvements, that would require a CPU and graphics card from the future 🙂.
 

Hehe, absolutely agree 👍

A 2x or 3x overall FPS improvement would definitely require a CPU and GPU from the future 🤣

Edited June 23, 20241 yr by abrams_tank

 

12 hours ago, abrams_tank said:

A 2x or 3x overall FPS improvement would definitely require ... GPU from the future

Jen-Hsun Huang to the rescue. RTX 5090 !