FSL Concorde - Core usage confusion

[SimeonWilbury 220]

The renderasync tweak has made a world of difference on my machine too,

Crucially I now have a decent framerate on night approaches into Heathrow, which was the one major performance issue that was bugging me. Many more Concorde flights are now on the table it seems.

As an aside, I notice the gauges still have quite a low FPS, but the actual sim now runs far smoother and the Autopilot is much more competent than it was when my frames were chugging.

Many thanks Maxime!

Edited February 20 by SimeonWilbury

[kmax59 213]

On 2/19/2024 at 7:09 PM, Ray Proudfoot said:

That equates to 0101010101010101

I can see you have SMT enabled, hence the 16 logical processors. So reading from right to left you have one half of each processor assigned to P3D.

What are the assignments in the JobScheduler section?

That's right, but as you can see there are cores that should not be used but they still are. I'm using these settings: 

[JobScheduler]
AffinityMask=65535
P3DCoreAffinityMask=21845
MainThreadScheduler=0
RenderThreadScheduler=1
FrameWorkerThreadScheduler=2

 

On 2/19/2024 at 9:26 PM, Afterburner said:

I am glad to hear that you have finally had a chance to enjoy your flight with the Concorde. Do you have any fps cap enabled, and are you able to maintain that cap?  Also, if you use other airplanes, do you see odd cores active and even ones idle on your task manager CPU load graphs (the way it should be according to your AF settings)?  If yes, then it must be indeed the case that Concorde "plays" with the AF settings internally.

Yes my fps are capped to 30 fps using the Nvidia Control Panel. That's the most stable solution for me. I am usually able to maintain that cap with all other aircraft but with Concorde (with the renderAsync line set to 1), fps sometimes drop below 30 (around 25 which is still acceptable considering I am using a true Gsync monitor that prevents visual stuttering due to the fps not being synced with the monitor). Concorde is still very heavy on my system and I really hope perfs will be improved in the next update but at least I can finally enjoy it.

I haven't had the opportunity to confirm that the CPU behaves as per my jobscheduler settings with other planes but from what I remember, that was the case. I'll have a look and will post my result here when I can do that 😉

Edited February 20 by kmax59

[Ray Proudfoot 8,980]

3 minutes ago, kmax59 said:

That's right, but as you can see there are cores that should not be used but they still are. I'm using these settings: 

Same on mine Maxime but I’m not enough of an expert to prove beyond reasonable doubt our AM settings are being manipulated by outside influences. And even if we could the silence from the developers would be deafening.

Just as it is for the official response to your find. I find that very rude. Three days in and not a word. Wow!

[kmax59 213]

7 minutes ago, Ray Proudfoot said:

Just as it is for the official response to your find. I find that very rude. Three days in and not a word. Wow!

I don't really understand why nobody is reacting to this indeed, may be they are investigating, let's see by the end of this week if Norman or someone else reacts 😅

I'm extremely surprised that nobody in the team, even the testers, didn't find this earlier than me.

I must say I always loved FSlabs products, I've owned them all from the very first Concorde-X in boxed version to the entire 32X series and then Concorde, but something I always quite disliked is their deafening silence in many occasions

Edited February 20 by kmax59

[Ray Proudfoot 8,980]

Maxime, great products but questionable attitudes towards customers. You gave to have bought their aircraft before you can post. But I’ve seen posts removed without explanation. I don’t know why they behave this way. I was threatened with moderation by the head man. That’s why I won’t post there unless circumstances require it.

As for the lack of reaction I can only think customers aren’t flying her and aren’t visiting the forum. Ironic really because if they did see your post it might make them change their mind.

[Afterburner 374]

It is disappointing to see that the developers at FSL still haven't provided an explanation for why changing the AllSync setting in the ini-file improves the performance for a lot of users. If they are not giving an apology to Maxime for the long silence, at least they should thank him for coming up with the helpful tip for free.

@kmax59  What I am interested to hear is whether your CPU load on the main core is at 100% when your fps drops below 30. If yes, it indicates that the CPU can't keep up with that frame rate (as it is normally the case in a CPU-limited situation). If not, then something else slows down the performance of your PC.

Edited February 21 by Afterburner

[kmax59 213]

11 hours ago, Afterburner said:

 

@kmax59  What I am interested to hear is whether your CPU load on the main core is at 100% when your fps drops below 30. If yes, it indicates that the CPU can't keep up with that frame rate (as it is normally the case in a CPU-limited situation). If not, then something else slows down the performance of your PC.

No the CPU load is not at 100% on the main core, but I don't think that it's the case with other aircraft either though, I'm not sure why honestly. I'll definitely try with another aircraft soon, probably not before next week though. I feel the urge to fly Concorde now that it finally works with an acceptable framerate. I flew AFR001 this evening and that was so satisfying to be back at it ! I definitely need to be back at training as well however, I forgot a lot of the cockpit flows during these last 6 years 😰

I'll slightly go off-topic with this but let's be quick: I realized during a short internet shutdown that the HP200LX totally relies on an internet connection, and if you don't have one it seems you simply cannot plan a proper flight, am I right? I also did not find any way to select the descent speed between 325, 350 and 380 which was used by AF, it seems the HP200LX only calculate the decel point with the 350 kt procedure and as I could not find any way to change the decel point or the TBP point on a scheduled flight, it seems we're forced to use the 350kt procedure for now. I'm not sure about these so I've asked these as well on their forum.

Edited February 21 by kmax59

[MachTwo 386]

22 hours ago, Afterburner said:

It is disappointing to see that the developers at FSL still haven't provided an explanation for why changing the AllSync setting in the ini-file improves the performance for a lot of users. If they are not giving an apology to Maxime for the long silence, at least they should thank him for coming up with the helpful tip for free.

Dear Konstantin,

The issue is, that this is an unsupported, undocumented parameter and it has been disabled for a reason. While it might enhance performance for certain users, extensive testing revealed that it also triggers threading issues with some of our instrumentation. Consequently, we made the decision to disable it to maintain system stability.

In recent weeks, we've communicated that updates to our Concorde title will be delivered to our customers in due course. However, it's essential to acknowledge the limitations we face in optimising performance on the outdated platform engine that Prepar3D utilises. Despite our efforts, squeezing out present-day expectations from this aging engine presents significant challenges.

Thank you for your understanding.

[Ray Proudfoot 8,980]

Hi Andrew,

For 10 years I flew the 32-bit FSL Concorde on two computers. Until 2018 I had to be careful with settings to keep fps reasonable and VAS within limits.

Frames per second was never the issue on the Chillblast i7-8086K / 1080Ti combo I acquired in 2018. The computer was extremely capable and even with fuel managed by the VFE performance was fine. The only problem was VAS. You needed to keep settings low to stay within the 3.8Gb limit.

But with a 64-bit version of P3D that limitation was removed. I’m not sufficiently knowledgeable to know what limitations exist within a 64-bit version. It would be interesting to know what they are and why Concorde seems to be the only aircraft affected. To my knowledge it’s the only aircraft with a built-in virtual crew.

There are clearly some processes that are consuming considerable resources to the extent that even on a system as powerful as mine the sim is incapable of maintaining a pretty reasonable 30fps at the busiest airports with the same amount of Ai as I use for the PMDG737. I’m curious what those processes are.

Perhaps it’s the Virtual Crew System whilst impressive but could be the issue. Or perhaps it’s the integration of the FE panel into the Virtual Cockpit.

Those two features are the main differences between the 32-bit and 64-bit versions.

If it was known during development these would severely impact performance was consideration given to scaling them back? Or perhaps optional so users could toggle them off if performance was adversely impacted.

The 2D FE panel in the 32-bit Concorde did everything this one does but using a fraction of the resources. Thinking back to when development was ongoing it was suggested by many that a 64-bit version of that Concorde retaining the popup panels would be very acceptable. It would have shortened the development time and many would have been satisfied.

If P3D really is the reason why performance is so poor do these issues go away with the MSFS version? Is 60fps at a 3rd party Heathrow with 100+ Ai a pipe dream or a realistic possibility in that sim?

For now I shall continue to leave the value at 1 giving me much improved performance and will monitor the gauges on the FE panel.

How long is “in due course”? Can you be more specific? Before August - the first anniversary - or longer? You’ll appreciate performance is the highest priority with any aircraft or other addon. Not looks, but the ability to fly the aircraft and land it with acceptable fps compared to other complex aircraft.

One final question, I raised a ticket a week ago regarding the fuel amount in tank 9 on approach. Can I expect an answer soon?

[CO2Neutral 278]

On 2/22/2024 at 1:46 AM, MachTwo said:

Despite our efforts, squeezing out present-day expectations from this aging engine presents significant challenges.

It would be nice if you provided more details, because "threading issues" on a gauge is pretty generic and has nothing to do with P3D "aging engine" and everything to do with gauge code implementation.  To be frank, it sounds like you have not discovered the actual cause of your issue and falling back on P3D as the problem.  The threading capabilities of P3D are far superior to the threading capabilities of MSFS (even Microsoft acknowledge that and have made threading a feature task to improve for MSFS 2024).

Regardless, I would really like to see some more details at how you arrived at your conclusion and why the issue isn't present in A3xx series under P3D?

Edited February 25 by CO2Neutral

[kmax59 213]

Hi @Afterburner @Ray Proudfoot,

As we speaked about it earlier in this post, here is the CPU behaviour with AffinityMask 21845 but this time with the PMDG 747 : it seems to be defintely following the affinity setting that Concorde does not follow: 

[Ray Proudfoot 8,980]

Maxime, was this screenshot taken under similar circumstances to your Concorde one?

Neither shows any VP working flat out so 30 fps should be achievable. Was your Concorde one on the previous page taken with the tweak to improve performance?

A reply from @MachTwo would be helpful on earlier points raised.

[SierraDelta 1,342]

On 2/25/2024 at 6:09 AM, CO2Neutral said:

The threading capabilities of P3D are far superior to the threading capabilities of MSFS

Sorry, but I dont agree, here's P3D with very uneven workload distribution:

and here's MSFS with much more uniform workload distribution:

[Ray Proudfoot 8,980]

Two different aircraft in two different sims. How can you make a fair comparison especially when it’s suspected FSL are manipulating AM settings.

[Ray Proudfoot 8,980]

I flew Heathrow to Las Palmas this morning. During the vast majority of the flight cores 4 and 5 were running flat out at 100%. Even with the aircraft over the western Atlantic. How bizarre.

I use Aivlasoft EFB to log my flights. To get credit for passing a waypoint I need to pass within 5nm of them.

I had to switch to HDG HOLD when 45nm from my active waypoint because the INS had started to turn to the waypoint beyond it. The heading change was a shallow 23°. At 15nm from my active waypoint I reengaged INS and passed with 2.5nm of it.

If I hadn’t intervened I would have missed the active one by several miles. This proves in my mind that the code is flawed and requires correction. But I was told a few months ago the calculation is correct. The person who told me that is a Concorde expert but I wonder if he’s an expert in trigonometry. I doubt it.

I also intervened in the decel phase as it wanted to start slowing me down 40nm before my calculated decel point using the supplied descent tables. I arrived at my assigned altitude just 10 miles early. Had I not intervened I would have been subsonic 40nm too early. 40nm at Mach 0.95 instead of Mach 2. 😟

Whilst it’s an excellent aircraft there are a few key things that require fixing. But with attention switching to MSFS we have been given no indication of when these fixes will be made to the P3D version. Will attention only return to P3D when the MSFS version is released and any bugs fixed? That could be another year.

The lack of news is disappointing but not surprising.