Fundamental performance question

[lukehall 10]

Something fundamental that I don't understand...

At a gate, KSFO, clear weather, aircraft in turnaround config. FPS limited in Nvidia CP to 60, VSync on.

GPU Usage approx 45%, CPU Usage approx 60%, GPU memory usage approx 50%, system memory approx 40%. All temps at around 50c and no CPU cores maxed out. Network activity 0%. GPU at approx 60% power consumption.

Why then am I only seeing 21fps when no component of the system is working at more than about half of its capacity? As soon as I get airborne, my FPS return straight back to 60 or higher if I deregulate them. It demonstrates that my system can produce far higher performance but something is preventing it on the ground. It has to be something in the core of MSFS, surely?

Under the same circumstances, if I change to an outside view, and pan around, the FPS are still glued to 21 and don't fluctuate regardless of scenery rendering, or for that matter, changing my settings. I currently have most things on ultra and lowering some of the more influential ones doesn't increase performance. This further suggests to me that the program is limiting FPS, not the capacity of the hardware. 

Can anyone give any insight into this?

Cheers! 🙂

[redjr 20]

Isn't it safe to say generally, that on the ground there are more objects - big and small - that require much more detail, hence more rendering. When you're high up in the air, you typically don't have a lot of different hard shapes and a lot of two colors; blue sky and white/gray clouds.  Objects on earth are seen as little specks, hence less rendering required.  Try flying over NYC and you'll likely see a drop in FPS - depending upon graphics card.

[Cpt_Piett 4,613]

My guess is that you’re main thread limited due to a bug which causes terrible performance on the ground, particularly in the US.

It might be worthwhile deleting photogrammetry cities from the content manager. 

More info in this thread:

 

[lukehall 10]

2 hours ago, redjr said:

Isn't it safe to say generally, that on the ground there are more objects - big and small - that require much more detail, hence more rendering. When you're high up in the air, you typically don't have a lot of different hard shapes and a lot of two colors; blue sky and white/gray clouds.  Objects on earth are seen as little specks, hence less rendering required.  Try flying over NYC and you'll likely see a drop in FPS - depending upon graphics card.

Yes, in the normal way of things that's correct. However, the point I'm making is that that whether I'm on the ground or in the air, my processing and rendering hardware still has lots of spare performance headroom (apparently) and there don't appear to be any bottlenecks to performance, so I can't understand where the performance drop is being invoked. If something was maxed out, it would be a different story of course! 🙂

[lukehall 10]

2 hours ago, Cpt_Piett said:

My guess is that you’re main thread limited due to a bug which causes terrible performance on the ground, particularly in the US.

It might be worthwhile deleting photogrammetry cities from the content manager. 

More info in this thread:

 

Thanks for your reply @Cpt_Piett. I've been following the thread you've linked to and tried deleting and deactivating several things but it hasn't made any significant difference. I'm guessing there is something deeper and uglier happening within the core simulator... probably explains why MS/Asobo have remained very quiet about it thus far!

[pmb 5,137]

I reported this specifically for KSFO on the official forum w.r.t. the first SU9 beta on April, 3rd

https://forums.flightsimulator.com/t/performance-at-ksfo-massively-down-in-su9beta/508831

A number of users confirmed it and it's marked "bug-logged". There were a number of similar reports by others.

Asobo is still investigating. I hope they will be successful.

Kind regards, Michael

[Cpt_Piett 4,613]

1 hour ago, lukehall said:

Thanks for your reply @Cpt_Piett. I've been following the thread you've linked to and tried deleting and deactivating several things but it hasn't made any significant difference. I'm guessing there is something deeper and uglier happening within the core simulator... probably explains why MS/Asobo have remained very quiet about it thus far!

Hopefully you’ll be able to sort out the issue, one way or another. Good luck!