So I was watching Chewwy's stream earlier today and he had really, really bad FPS for the 787.  He was getting from 25 to 30 FPS usually on ground and even during takeoff on 2K resolution.  Chewwy is using an Intel i9-9900K, GeForce RTX 2080Ti, and 32 GB of DDR4 RAM, so he does have a higher end computer.  There was heavy cloud cover earlier so that may have impacted his FPS.  Chewwy then worked with the chat to optimize his settings in the game but after optimizing his settings, he tried some GA aircraft that improved his FPS. It's uncertain whether the optimization he did to the settings or whether switching to GA aircraft improved his FPS.  He then flew the A320 later in the day (his settings were optimized at this point) but he was getting low 30 FPS sometimes.

So my question is, is it the airliners in MSFS that are dropping the FPS so much?  Do GA aircraft give 1.5x to 2x better FPS than the airliners?

 

 

Edited August 14, 20205 yr by abrams_tank
2K resolution originally used by Chewwy

24 minutes ago, abrams_tank said:

So I was watching Chewwy's stream earlier today and he had really, really bad FPS for the 787.  He was getting from 25 to 30 FPS usually on ground and even during takeoff.  Chewwy is using an Intel i9-9900K, GeForce RTX 2080Ti, and 32 GB of DDR4 RAM, so he does have a higher end computer.  There was heavy cloud cover earlier so that may have impacted his FPS.  Chewwy then worked with the chat to optimize his settings in the game but after optimizing his settings, he tried some GA aircraft that improved his FPS. It's uncertain whether the optimization he did to the settings or whether switching to GA aircraft improved his FPS.  He then flew the A320 later in the day (his settings were optimized at this point) but he was getting low 30 FPS sometimes.

So my question is, is it the airliners in MSFS that are dropping the FPS so much?  Do GA aircraft give 1.5x to 2x better FPS than the airliners?

 

 

👍

I don't understand why the FPS dropping so much in airliner cockpits. The systems and aerodynamics simulations are still done when in external view, but fps is 2x better.

I can only see one reason : draw calls. They are a using a single draw call per knob/switch inside cockpit. I'm betting the 2D geometry over glass panels isn't batched either.

If this is the reason they should definitely batch all the knobs and switches into single draw call, either directly perform transforms from CPU over the mesh, or use shader uniforms and index each element from the mesh

Edited August 14, 20205 yr by cepact

5 minutes ago, cepact said:

The systems and aerodynamics simulations are still done when in external view, but fps is 2x better

The gauges do not have to be rendered nor refreshed in external view, which depending upon the plane can save alot of overhead. Virtual 3d cockpits have always been tougher on resources than an external view. Thats why when testing fps you should never do it in external view as its meaningless there (unless you always fly externally).

Edited August 14, 20205 yr by hangar

Just now, cepact said:

I don't understand why is FPS dropping so much in airliner cockpits. The systems and aerodynamics simulations are still done when in external view, but fps is 2x better.

I can only see one reason : draw calls. They are a using a single draw call per knob/switch inside cockpit. I'm betting the 2D geometry over glass panels isn't batched either.

If this is the reason they should definetely batch all the knobs and switches into single draw call, use the mesh (or shader uniform) for transformation

All the knobs and switches in the cockpit are stagnant and they aren't changing dynamically so that would be very weird if that is what is dropping the FPS. If that is the reason though, I hope they can optimize it further.

But if GA aircraft give good FPS, I would still be happy to fly them.  Chewwy is getting subpar FPS again flying the A320 - he is getting about 30 FPS at times.  Having said that, he is flying at 4K resolution now.  But I believe a lot of his settings are still at medium & high, even though he is flying at 4K.

1 minute ago, hangar said:

The gauges do not have to be rendered nor refreshed in external view, which depending upon the plane can save alot of overhead. Virtual 3d cockpits have always been tougher on resources than an external view. Thats why when testing fps you should never do it in external view as its meaningless there (unless you always fly externally).

Oh, is that the reason why the FPS takes a huge hit?  I didn't know that. Hopefully Asobo can optimize this further.  I'll still be happy if the GA aircraft give good FPS.

If you want 4k and you want ultra settings then get a $2k video card and top processor and get used to flying in only forgiving scenery in default planes.

From what have seen thus far I have come to the conclusion that a 2k monitor/resolution running with a GTX1080 and something along the lines of an i7 8700k proc should be ok at MEDIUM settings with even demanding scenery areas such as JFK/KLGA (as long as you dont over complicate the scene, i.e. bad thunderstorms/rain on windshield & AI traffic all turned up at same time) .

Use a give & take sort of common sense, otherwise as history has always proven, you'll be in for a rough ride.

Edited August 14, 20205 yr by hangar

Airliners have always had more impact than GA in the VC. This has been the case for at least 15 years.

Edited August 14, 20205 yr by sanh

yea with 3rd party thats true...and thats because in 3rd party airliners theres alot more programming going on with the fmc's and alot more functioning switches on the panels and overhead.

I'm also betting that all those G1000 units will be eating up alot of frames if you turn on the synthetic vision and map insets. Those sorts of picture within picture features has always been notorious for eating frames. So, shut them OFF whenever you can, especially in dense scenery areas.

Edited August 14, 20205 yr by hangar

Flying airliners+4K+heavy clouds/rain is going to be lower FPS on any pc with todays technology. I have a great gaming pc, but use a 2k monitor because 4K takes a lot more resources to get decent fps.

Edited August 14, 20205 yr by Pugilist2

29 minutes ago, hangar said:

The gauges do not have to be rendered nor refreshed in external view, which depending upon the plane can save alot of overhead..

There was a video showing the gauges updating in external view though. Perhaps only at higher graphics settings. This is why it surprises me that the cockpit view would create so much more overhead.

those gauges are an overlay arent they? I think they are rendered much easier and also might not get redrawn as fast as the real gauges in the cockpit...but im only taking an educated guess here.

@abrams_tank WIthout knowing Chewy's resolution it's hard to say, if it's 4K that's going to be taxing on anything.

Also I see "2K" being bandied about in this thread, is everyone clear that basically means normal FullHD 1080p? The resolution between 1080p and 4K is 1440p, which is about 2.5K vertically.

Edited August 14, 20205 yr by ckyliu

8 minutes ago, ckyliu said:

@abrams_tank WIthout knowing Chewy's resolution it's hard to say, if it's 4K that's going to be taxing on anything.

Also I see "2K" being bandied about in this thread, is everyone clear that basically means normal FullHD 1080p? The resolution between 1080p and 4K is 1440p.

Chewwy said he was on 2K resolution when he first flew the 787.   I think later on when he flew the Airbus, he increased it to 4K resolution.  But Chewwy said for him, there is little difference in performance from 2K to 4K resolution.

It really depends upon the scene that's being rendered/tested and your exact viewpoint at the time as well...some scenes will use more gpu and other scenes may need more cpu...if he's not seeing much difference it's likely that the scene is very forgiving and not demanding on his gpu OR he is bottlenecked at the cpu.

If you're going to benchmark GPU, then you first need to turn off all AI traffic then find and save a heavily gpu dependent and demanding scene and appropriate viewing angle for it. If you want to benchmark CPU then you find an empty field in the middle of nowhere with no trees and turn up all your AI in theI air and on the ground (you get the idea).

Edited August 14, 20205 yr by hangar