Can BeyondATC's injection be used in PSX Traffic

[KL Oo 700]

Hi Nico,

Feel free to close down if you like, but I thought I'd flag this video & Topic where BeyondATC are presenting their method of injecting traffic. They appear to be claiming that there is NO FPS hit in the sim as the sim isn't calculating any AI Traffic position data. They're clearly injecting in a method not seen before and I wondered if you had any idea how they might be doing it? On the surface, I thought it would be the same method as you - manually telling the sim where to display an AI Aircraft at any given frame and that would induce the heavy FPS hit we get with PSX Traffic, but it appears not.... 

 

[kiek 1,215]

6 hours ago, KL Oo said:

and I wondered if you had any idea how they might be doing it? 

Yes, they are using exactly the same method as PSXtraffic is doing for 8 years now. Bypassing the MSFS engine and telling MS at each frame where to position the aircraft.

So nothing new.

The fps hit is not only coming from the AI models you are using, but also from your MSFS settings. I don't know what settings they are using in that (well controlled)  demo. 

Personally I don't see why 90 fps is so much better then 45? With 45 everything looks smooth enough to me. Giving a real live, real time, picture is much more important to me, and that is what is missing in their program, it is still canned traffic, not representing the positions of real live aircraft.

If you go to EGLL with PSXTraffic you will see up to 15 aircraft lined up in three or four rows waiting for departure, or go to KSEA and you may see 15 aircraft in one row waiting, nicely sequenced. That all comes for free from RealTraffic and it will be very hard for them to achieve that. Not to speak of disturbances, such as blocked run- or taxiways, or an aircraft that has to turn around, or, ....

The way they build that 3D model of an airport finding out the taxiroutes and so on is a lot of work and really impressive (chapeau), but for PSXTraffic that is not needed. All that information is coming from the RealTraffic app.

 

Edited November 10, 2023 by kiek

[SierraDelta 1,342]

7 hours ago, KL Oo said:

and that would induce the heavy FPS hit we get with PSX Traffic

Something’s wrong, with your system you shouldn’t see any heavy fps hit. Are all your AIG and/or FSLTL models totally up-to-date?

[KL Oo 700]

42 minutes ago, SierraDelta said:

Something’s wrong, with your system you shouldn’t see any heavy fps hit. Are all your AIG and/or FSLTL models totally up-to-date?

There's nothing wrong with my system...its a simple case of adding AI Traffic puts pressure on the main thread. The FPS drop is different for each user however. In my case, if I'm running my triple monitor 4K Setup....I can get an easy 35 FPS at EGLL....add in 50 AI traffic and its drops down to low 20s.

You telling me you don't see any performance hit when you add 200 models to your sim?

[kiek 1,215]

1 hour ago, KL Oo said:

There's nothing wrong with my system...its a simple case of adding AI Traffic puts pressure on the main thread. The FPS drop is different for each user however. In my case, if I'm running my triple monitor 4K Setup....I can get an easy 35 FPS at EGLL....add in 50 AI traffic and its drops down to low 20s.

You telling me you don't see any performance hit when you add 200 models to your sim?

I have sometimes 160 live and about 100 static parked aircraft at EGLL INIBUILDS and still get say between 35 and 40 fps depending on the position of the camera. But I have a single monitor setup running MSFS windowed (and a  powerful cpu).

[SierraDelta 1,342]

9 hours ago, KL Oo said:

There's nothing wrong with my system...its a simple case of adding AI Traffic puts pressure on the main thread. The FPS drop is different for each user however. In my case, if I'm running my triple monitor 4K Setup....I can get an easy 35 FPS at EGLL....add in 50 AI traffic and its drops down to low 20s.

You telling me you don't see any performance hit when you add 200 models to your sim?

No, there is obviously some hit, but not what I would characterize as a heavy hit - maybe 10 frames out of 75-80.

One thing - on my system Win11 and the 13900k does a fine job of assuring that all 8 P-cores are working at the same level, same goes for the E-cores, there's no such thing as a main thread.

[mmcmah 409]

1 hour ago, SierraDelta said:

No, there is obviously some hit, but not what I would characterize as a heavy hit - maybe 10 frames out of 75-80.

One thing - on my system Win11 and the 13900k does a fine job of assuring that all 8 P-cores are working at the same level, same goes for the E-cores, there's no such thing as a main thread.

Not sure what to say, other than everyone's experience can vary for so many reasons. As said above, there is definitely a hit on the CPU from the processing of the AI traffic. PSXTraffic itself consumes very few resources, but injecting the traffic into the sim, even though MSFS doesn't need to control them, results in a significant drop in frames - FOR ME.

I have a 13900K with dual monitors at 4K being driven by a 4090 on Windows 11 and Process Lasso to keep the MSFS thread on the P cores rather than the E cores. I have 64GB of RAM and NVME drives. Even so, here are two examples of the hit to frames (obviously the developer mode doesn't account for the additional frames that NVIDIA generates) where the only difference was the injection of PSXT traffic at EGLL with the default Citation Longitude:

No PSXT injection: https://imgur.com/a/INByHqI

With PSXT injection: https://imgur.com/a/M2OaMqj

I'd call 21 FPS a pretty significant difference.

[mmcmah 409]

Turning off the Process Lasso core manipulation and some background tasks (to try to make a best case scenario) resulted in a baseline FPS increase of about 4 FPS, but that only translated into a 1 FPS increase with PSXT running, so the delta became 24 FPS:

Without PSXT: https://imgur.com/a/n6yO7PN

With PSXT: https://imgur.com/a/s2xhNUm

[somiller 694]

1 hour ago, mmcmah said:

Turning off the Process Lasso core manipulation and some background tasks (to try to make a best case scenario) resulted in a baseline FPS increase of about 4 FPS, but that only translated into a 1 FPS increase with PSXT running, so the delta became 24 FPS:

Without PSXT: https://imgur.com/a/n6yO7PN

With PSXT: https://imgur.com/a/s2xhNUm

Your demonstration seems typical with the number of aircraft injected into MSFS - most certainly a substantial strain compared to no traffic. I see the same with of course MUCH lower fps caused by my meager CPU struggling much of the time at 100% utilization, of which 80%-95% is MSFS. I'm usually in the 15-25 fps range with 80+ AI aircraft, PMDG 737, and an airport like EGLL.

[KL Oo 700]

So my take homes from this thread are that

1. Yes, AI Traffic does have an impact on FPS (who would have thought 😁); and

2. Nico believes that BeyondATC are injecting traffic in the same means as he already is. Which paints a disappointing picture for BeyondATC's ability to not hit FPS.

 

[F737MAX 5,612]

29 minutes ago, KL Oo said:

So my take homes from this thread are that

1. Yes, AI Traffic does have an impact on FPS (who would have thought 😁); and

2. Nico believes that BeyondATC are injecting traffic in the same means as he already is. Which paints a disappointing picture for BeyondATC's ability to not hit FPS.

I don't see how BeyondATC could ever inject AI without there being *any* FPS impact.

The sim's graphics engine has to render these complicated moving AI objects, as well as generate their animations, sounds, lights, effects, etc.
With Frame Generation turned on and a limited number of AI aeroplanes in the tech concept video, the impact can intially be limited as to be neglible. At some point though, say 100 AI aeroplanes, rendering all these elements starts to have an effect.

Throw in a complex user airliner add-on and a detailed payware scenery and there is limited PC resource remaining for traffic computations and model rendering - no different to what we experience now.

I like what BeyondATC is working towards as it should be far, far better than what the MSFS default AI engine is capable of. Yet what I really want to use the program for, is being directed by competent and great-sounding ATC.

[V1ROTA7E 718]

PSXT uses a huge amount of traffic, I had to change my matching to ‘2’ in order to get better performance. It’s still awesome though. I am curious though, JustFlight’s Traffic engine does give me much better performance than FSLTL or PSXT. I’m wondering if having the PSXT engine running on a different affinity mask will help with performance. Not sure if it’s the sim, processing the moving objects that causes the drop, or if it’s the program itself using CPU that is causing the issue. 

[knich 113]

16 hours ago, kiek said:

Personally I don't see why 90 fps is so much better then 45? With 45 everything looks smooth enough to me.

 

Except when you want to 2x and 4x speed.  🙂

Edited November 10, 2023 by knich

[kiek 1,215]

7 hours ago, V1ROTA7E said:

PSXT uses a huge amount of traffic, I had to change my matching to ‘2’ in order to get better performance. It’s still awesome though. I am curious though, JustFlight’s Traffic engine does give me much better performance than FSLTL or PSXT. I’m wondering if having the PSXT engine running on a different affinity mask will help with performance. Not sure if it’s the sim, processing the moving objects that causes the drop, or if it’s the program itself using CPU that is causing the issue. 

Psxt is not using much cpu time, it is below 0.5 percent. To me It seems to be  the Simobjects "proocesssing" of MSFS and/or the rendering.

[kiek 1,215]

6 hours ago, knich said:

Except when you want to 2x and 4x speed.  🙂

That you must not do (!!), see the instructions in the Manual. The datastream of RT remains at 1x.