Fenix A320 Block 2 - FPS

[Christopher Low 5,354]

2 hours ago, RJC68 said:

.....I think the other thing people do not consider is monitor size. I have flown with a 55 inch Sony X90K for over a year now and when I first got my 4090 I tested various resolution/FPS combos to see what worked best. After running at 100Hz, 97fps & GSYNC I can definitely see a big difference if I go back to 60fps, on a big screen like this it just is not smooth when I pan the camera no matter what I do. Running the TV at 100Hz or 120Hz with fps to match is night and day different as you have witnessed yourself. It brings in a level of smoothness I had never witnessed before.....

That is a good point. I run MSFS @ 1080p resolution on a 24" widescreen monitor, so differences in framerate will almost certainly be harder to perceive on this than they would be on your 55" monitor.

[aniiran 431]

The Fenix has never performed as well as the PMDG 737 series.  But I think it really depends on the scenario.  The Fenix and the PMDG do about the same at iniBuilds LAX, of course the frame rate is low to begin with.  At a minor airport or stock airport the PMDG has a significantly better frame rate, it always has.

I have noticed a bit of a performance penalty with the IAE model specifically, but not with the CFM...   strange.  I think one of the differences is that there are no 4k repaints.  I never use 8k liveries always 4k, so for me I think that has something to do with my stuttering in the outside view and wing views.

The cockpit textures seem sharper to me, maybe they are 8k now as well?

I get a new system soon so all of this won't matter.

[RJC68 823]

9 minutes ago, aniiran said:

The Fenix has never performed as well as the PMDG 737 series

Depends on your hardware,

I get equally great performance in both the Fenix & PMDG. Zero issues with either aircraft here.

[polosim 850]

19 minutes ago, aniiran said:

The Fenix has never performed as well as the PMDG 737 series.

Well, I'm not so sure about this statement.

[carlanthony24 576]

35 minutes ago, aniiran said:

 

I have noticed a bit of a performance penalty with the IAE model specifically, but not with the CFM...   strange.  I think one of the differences is that there are no 4k repaints.  I never use 8k liveries always 4k, so for me I think that has something to do with my stuttering in the outside view and wing views.

The cockpit textures seem sharper to me, maybe they are 8k now as well?

I get a new system soon so all of this won't matter.

4K repaints on the livery manager and on fs.to the livery manager will say 8K but when you download them it will be in 4K

[Cpt_Piett 4,672]

37 minutes ago, aniiran said:

The cockpit textures seem sharper to me, maybe they are 8k now as well?

Same here. At least I think so. It’s the first thing I noticed after launching B2.

[F737MAX 5,608]

29 minutes ago, polosim said:

Well, I'm not so sure about this statement.
[image]
[image]

You and your facts. How are we supposed to throw shade at the Fenix's performance now...?! 😁

Also, interesting how both freeware/default-modded airliners struggle with 0.2% lows.

  

3 hours ago, Cpt_Piett said:

Did a long flight yesterday without full stop landings, only touch n' go's. FPS 75-80 the whole time (using AutoFPS). Route: LFPO-EHRD-EHAM-EGKK-EGLL-EGCC-EGPH-ENBR-ENGM

That's different!
One way to see a lot of airports and certainly saves a lot of time without the faff of taxi, shutdown, turn-around, start-up and push, and taxi again.
 

[psychedelic_tortilla 170]

The biggest FPS killer in this sim (and I suppose in any sim) isn't your own plane, but other planes. I get beautiful FPS when spawning at the gate, but once I connect to VATSIM on a busy evening, there go 15 to 20 FPS straight away. AutoFPS has ameliorated that somewhat, but what has made airports like iniBuilds' Heathrow or FlyTampa's Amsterdam usable for me in the first place is the FSR Frame Generation mod. Without that, these airports where absolute slideshows with lots of traffic.

[Christopher Low 5,354]

AI traffic must hit performance a lot harder in MSFS than it did in P3D. I have not tested the theory yet, but framerates at IniBuilds EGLL London Heathrow and FlyTampa EHAM Amsterdam Schiphol on my PC from yesteryear with only static planes are very good.

Edited March 6 by Christopher Low

[Cpt_Piett 4,672]

4 hours ago, F737MAX said:

That's different!
One way to see a lot of airports and certainly saves a lot of time without the faff of taxi, shutdown, turn-around, start-up and push, and taxi again.
 

Probably the most fun flight in the Fenix so far! I just re-routed in the MCDU after each touchdown. Great practice too for those go-arounds!

EDIT: Oh, it seems I'm in the Fenix B2 fps thread now...

3 hours ago, psychedelic_tortilla said:

The biggest FPS killer in this sim (and I suppose in any sim) isn't your own plane, but other planes.

Agree! Btw, what is that creature in your avatar?

3 hours ago, Christopher Low said:

AI traffic must hit performance a lot harder in MSFS than it did in P3D.

Supposedly it hits even harder in SU15 beta, at least according to some of the community managers.

Edited March 6 by Cpt_Piett

[psychedelic_tortilla 170]

11 hours ago, Christopher Low said:

AI traffic must hit performance a lot harder in MSFS than it did in P3D. I have not tested the theory yet, but framerates at IniBuilds EGLL London Heathrow and FlyTampa EHAM Amsterdam Schiphol on my PC from yesteryear with only static planes are very good.

Well, static planes get rendered by the GPU, which is orders of magnitude faster than a CPU cycle. They don't need additional CPU resources to compute their movement, the path they will take, their altitude, position and orientation etc. This is what adds up and slows performance, the additional load on the CPU on top of it having to execute all the code for simulating your plane before the GPU can render it.

Maybe a little excursion, I don't know if people know this, but a single CPU (a single core, to be exact) can only execute exactly one instruction at any given time. The scheduler in your OS decides what to execute when, and how long the queue can be. Switching between processes/instructions is just so insanely quick, it only seems like everything runs in parallel. But as we all know as flight simmers, if a single process has many demanding instructions, that illusion can vanish rather quickly. This is where multithreading is so important, since offloading CPU cycles to another core (which is it's own CPU with its own memory map, cache etc.) allows parallel processing to relieve the bottleneck. 

This is also why GPUs are so word not allowed fast. They are specifically optimized for matrix operations (computer graphics - just like ML - is just lots and lots of linear algebra), and many of these can be done in parallel on pieces of a matrix and then just frankensteined together once done. 

 

7 hours ago, Cpt_Piett said:

Agree! Btw, what is that creature in your avatar?

 Haha, that is a sloth in a space suit. Found it on KnowYourMeme ages ago, it's so ridiculous, I love it :D. 

[willy647 193]

On 3/5/2024 at 7:58 PM, RJC68 said:

I disagree, I can consistently hit 80fps in this version and the previous one. That is @4K with Ultra settings & FG, shadows at 6144 and plenty of FSLTL traffic. I was at EGLL this morning with a ton of live traffic and I only lost 10fps until I was in the air, that was not the Fenix that was just the extra load due to the scenery and traffic. After that flight I left Gibraltar with a solid 80fps on the ground and it hasn't budged. I can get 97fps consistently as well but locking at 80fps is just as smooth and gives my GPU more headroom.

No performance issues whatsoever with either versions for me 

Not everybody can spent over €/$ 1000 on just a graphics card

[Cpt_Piett 4,672]

1 hour ago, psychedelic_tortilla said:

Maybe a little excursion, I don't know if people know this, but a single CPU (a single core, to be exact) can only execute exactly one instruction at any given time. The scheduler in your OS decides what to execute when, and how long the queue can be. Switching between processes/instructions is just so insanely quick, it only seems like everything runs in parallel. But as we all know as flight simmers, if a single process has many demanding instructions, that illusion can vanish rather quickly. This is where multithreading is so important, since offloading CPU cycles to another core (which is it's own CPU with its own memory map, cache etc.) allows parallel processing to relieve the bottleneck. 

That’s very interesting, thanks for sharing! I guess there’s an argument for using hyperthreading as number of threads are doubled. At least w/ DX12 multithreading is better.

1 hour ago, psychedelic_tortilla said:

Haha, that is a sloth in a space suit. Found it on KnowYourMeme ages ago, it's so ridiculous, I love it 😄

It’s awesome 🙂 Reminds me of a movie I saw a long time ago with many weird creatures 😉 

Edited March 7 by Cpt_Piett

[Aamir 2,907]

16 hours ago, aniiran said:

The Fenix has never performed as well as the PMDG 737 series.  But I think it really depends on the scenario.  The Fenix and the PMDG do about the same at iniBuilds LAX, of course the frame rate is low to begin with.  At a minor airport or stock airport the PMDG has a significantly better frame rate, it always has.

I have noticed a bit of a performance penalty with the IAE model specifically, but not with the CFM...   strange.  I think one of the differences is that there are no 4k repaints.  I never use 8k liveries always 4k, so for me I think that has something to do with my stuttering in the outside view and wing views.

The cockpit textures seem sharper to me, maybe they are 8k now as well?

I get a new system soon so all of this won't matter.

We've seen a lot of speculation around the textures being the 'culprit' for bad performance since Block 2; so far our investigation has found several code-side items leading to poor performance on some systems, which we be doing our best to address. On the other hand, we've yet to find a scenario where lowering our already very low texture memory usage makes a measurable impact for any of those reporting lower performance than Block 1. To this end here's some hard numbers to help visualise just how efficient Block 2 is in terms of resources:

PMDG's 737 has 421 cockpit textures totalling 709MB, of those 4 are 4k.

Fenix A320 has 142 cockpit textures totalling 271MB, of those 4 are 4k.

To put the above into perspective, the default A320 NEO as shipped uses 235MB of textures for the cockpit - we're using just 36MB more. In fact, Block 2 actually falls within the stringent memory requirements to be shipped on XBOX if we so wished.

Beyond textures, our drawcalls are significantly reduced compared to Block 1, and in-line with other aircraft even though we have a significant amount of extra interactive items such as nearly 300 circuit breakers. Our cockpit is a little under 4000 drawcalls 'at scene' compared to the 737 which has about 3700 'at scene' - there's not much in it. When you get towards CPU-intensive aspects of the art like bone animations, we're also doing pretty well here, using 238 bones to the 737's 396.

Ultimately, numbers are meaningless if the performance isn't there, but I wanted to clear up any misconceptions that we've just "thrown more polygons and textures" at the A320 for Block 2 - the reality is in almost every measurable way we've done the opposite, and utilised some seriously innovative tech to bring that level of visual quality whilst keeping within the aforementioned XBOX budget. Beyond the bigger code-side performance bugs, we're continuing to investigate every aspect of the product to make little nip and tucks to eek out as much performance as we can across a variety of systems.

[RockOla 233]

Very informative and, above all, factual post, thank you Aamir. 👍