5 minutes ago, Ray Proudfoot said:

Odd that needs doing. Does it have a benefit? Why don't LM include it?

Anyway, I've found a solution to the problem of fps dropping below 30 on approach. I changed Autogen and Scenery Draw Distance from High to Medium and 30 was maintained throughout. Core 0 was below 100% for the circuit except for the last 3 miles when it hit 100% but fps remained at 30.

It would be interesting to know what other users have that setting at.

It's a manual tweak to adjust the time spent between rendering and scenery loading. It allows refinement of the SIM performance to suit different use cases.  It is documented in the learning center I believe.

Then it appears that the problem was a bottleneck with the sim trying to load distant autogen with insufficient cores/bandwidth available. Reducing the setting reduced the demand to within the hardware capability.  

I found going from a 6core 8700 to a 10 core 10850 made a significant difference in scenery loading.

Edited July 27, 20214 yr by kevinfirth

22 minutes ago, Ray Proudfoot said:

Can't see any reference to that in p3d.cfg. I have never experimented with that.

In stock P3D the FFTF is 0.33 - Prepar3D.cfg:

[MAIN]
FIBER_FRAME_TIME_FRACTION=0.33

That gives one third of the time to loading scenery. By reducing the fraction to 0.1 that will free up more time in the main task of rendering. It's not usually necessary to change it but it can be useful as a test to determine where the fps is going.

Some use it to limit the scenery loading to maintain fps with higher detail in the closer objects.

With more cores and or HT enabled allocating more LPs the scenery can load faster and the fraction can be maintained lower.

27 minutes ago, micstatic said:

my 9700k I have idling around 30C.  P3D 60-73C.  To me 90c is too hot.  Considering that chip starts throttling around 100c I don't see much head room there.  Just my opinion. 

Nice and low, I remember I started at around 75c max after moving to a water cooler.

The temps went up ~5c after moving to unlimited which eliminated the stutters for me.

I may need to clean the case...

17 minutes ago, SteveW said:

In stock P3D the FFTF is 0.33 - Prepar3D.cfg:

[MAIN]
FIBER_FRAME_TIME_FRACTION=0.33

That gives one third of the time to loading scenery. By reducing the fraction to 0.1 that will free up more time in the main task of rendering. It's not usually necessary to change it but it can be useful as a test to determine where the fps is going.

Some use it to limit the scenery loading to maintain fps with higher detail in the closer objects.

With more cores and or HT enabled allocating more LPs the scenery can load faster and the fraction can be maintained lower.

Thanks. I have no instance of that. In SimStarterNG tweak section there is an option to change it. I'll experiment.

Just tried a circuit from EGLL 27R. Far more demanding than EGKK. 30 was maintained on the downwind and base legs. It dropped briefly to 20 around 3nm out presumably where a lot of loading  was being done.

It soon recovered to 25 and stayed there for the landing and rollout. I can't really complain about that performance given the demands of SimWimgs EGLL.

I may tweak other scenery settings but I'm not going to get wholesale improvements without more overclocking which I'm reluctant to do.

43 minutes ago, simfan1983 said:

Is that max temp for individual cores? Are you water cooled?

I have a be quiet Dark Rock 3 Quiet CPU Cooler but the case has limited airflow as it's a be quiet Silent Base 600 - I listen to a lot of music and like the ability to make the pc almost silent, but it's not the best for keeping everything cool, especially when it's 25 degrees plus in the room. I also probably need to clean the filters on the case!

Yes, and yes water cooled. My case is a bequiet dark base pro 900. My temps are at an ambient between 20 and 25c.

42 minutes ago, kevinfirth said:

I found going from a 6core 8700 to a 10 core 10850 made a significant difference in scenery loading.

My CPU is now 3 years old. Not quite ancient but it just goes to show time stands still for no man (or woman! ).

33 minutes ago, Ray Proudfoot said:

Anyway, I've found a solution to the problem of fps dropping below 30 on approach. I changed Autogen and Scenery Draw Distance from High to Medium and 30 was maintained throughout. Core 0 was below 100% for the circuit except for the last 3 miles when it hit 100% but fps remained at 30.

It would be interesting to know what other users have that setting at.

Autogen settings have some of the biggest impacts on frame rates for me. My World settings in 4.5 are as follows:

Settings - Image on Pasteboard

I loaded the FSLabs at EGKK 08R and did a circuit in rainy / misty weather with all my usual apps running, frame rates were pretty good for my system, held around 30 with a minimum of 25 and no stutters. A few images including fps in top left:

https://pasteboard.co/Kd6ndZd.png

https://pasteboard.co/Kd6s7Bk.png

https://pasteboard.co/Kd6thF3.jpg

https://pasteboard.co/Kd6uHHF.jpg

 

55 minutes ago, simfan1983 said:

My temps are similar if you are measuring average temps. Can you try using CoreTemp and see what the max temps are on individual cores?

Pasteboard - Uploaded Image

https://imgur.com/a/lESj8yQ

here you go.  Probably could have let it run longer and heat up some more

Just now, micstatic said:

https://imgur.com/a/lESj8yQ

here you go.  Probably could have let it run longer and heat up some more

That's pretty cool - are you using water cooling?

1 minute ago, simfan1983 said:

That's pretty cool - are you using water cooling?

I'm using a 240mm AIO.  I feel like if I ran the test longer they would have gone up 5c or so

3 minutes ago, micstatic said:

I'm using a 240mm AIO.  I feel like if I ran the test longer they would have gone up 5c or so

Water cooling is pretty effective, particularly compared to my case with only a couple of small fan!
A nice little utility when you are testing BIOS settings is CPU-Z - it's similar to CoreTemp but shows frequencies for both CPU and Memory and has a handy little 'Bench' tab that lets you stress specified cores.
Another thing for people to check is that they have an XMP profile set in BIOS for their memory as quite often default BIOS runs memory underclocked without an XMP profile. I personally always overclock memory a little as it gives a significant performance gain at little effort.

11 minutes ago, simfan1983 said:

Water cooling is pretty effective, particularly compared to my case with only a couple of small fan!
A nice little utility when you are testing BIOS settings is CPU-Z - it's similar to CoreTemp but shows frequencies for both CPU and Memory and has a handy little 'Bench' tab that lets you stress specified cores.
Another thing for people to check is that they have an XMP profile set in BIOS for their memory as quite often default BIOS runs memory underclocked without an XMP profile. I personally always overclock memory a little as it gives a significant performance gain at little effort.

I do as well.  It's amazing that people ignore OC'ing their ram.  I also use CPU-Z and GPU-Z like you mention.  

@simfan1983, thanks for conducting the test. V4 and an Airbus is different to my v5.2 and PMDG Boeing so hard to draw comparisons.

But I note you had the same setting as me for Autogen and Scenery draw distance. And with fps dropping to 25 it confirms there is a reduction on final approach.

I turned HT off and flew a 08R circuit at EGKK using exactly the same settings as my earlier ones.

There is still a drop in fps when 2-3 miles out but it goes no lower than 25. I had to look at a rolling graph to see the drop.

Here's what TM-Performance looks like. Core 0 still maxxed out so I'm not sure there are any benefits to HT being off. Certainly no significant changes to it being on.

Back with HT on now but I did increase the core up to 4.7 from 4.6. Insignificant I know but the long term plan is to try to get to 5.0.