Beta 

I load the mkstudios KSEA airport (with zzz Static KSEA FSLTL) and the Flysimware C414 with DLSS at High, FG on, TLOD 305  OLOD 200 and all other settings on High except Buildings and Clouds which are Ultra.  I have Max Frame Rate set at 30 in the sim.  I have FSLTL running along with GAIST V7 and GAIST Trees and Scenery.  OR - just a whole bunch of stuff.

I start Free Flight with the Project Kart off road vehicle.  I can drive all over the airport, into the Cargo bays, around the Passenger terminals with a jet at almost every gate and the sim stays at a steady, rock solid 60fps. Both GPU time and GPU time are at 16 or 17 ms.  VRAM usage is 11.2 out of 12 GB and GPU busy is around 80%.  The scenery details are incredible, shadows are perfect and there is no melting of other strange things with the scenery or planes. 

I follow a 737 from the gate to the runway when they are ready to take off.  At the same time other aircraft are taking off and landing on the other two parallel runways.  Following under their tail or at a distance of 100-yards still a rock steady 60 fps.

The 737 accelerates for takeoff and I follow as close as possible and stay up with the jet to about 70 mph, and still 60fps.  As the jet pulls away I keep accelerating but they soon rotate and fly away.  At the same time a jet is taking off on a parallel runway.  Still 60 fps with no changes in the VRAM or CPU usage.

I then return to an empty GA gate where I load the Flysimware Cessna C414.  Shift -C into the cockpit and FPS drops to 35.  Shift-C out of the cockpit and look around the FPS drops to 25.   At this point the CPU time per frame has jumped up to over 100ms but the GPU stats are unchanged.

I return to the cockpit and prepare the C414 for flight.  As soon as I start to taxi - the FPS drops to single digits and the CPU time exceeds 250ms.  Still - the GPU stats remain the same.  As I prepare to takeoff the GPU time per frame climbs to about 60ms while the CPU exceeds 350 ms.  But, the VRAM usage does not change.

In summary:

Driving around a very detailed airport in a motor vehicle consumes ALL the VRAM but the FPS stays at the requested 60 per second. 

Following an AI plane in flight uses all the VRAM but the FPS stays at 60.

Loading a detailed aircraft, cold and dark, does not change the VRAM usage but does increase the CPU time by almost a factor of 10 and cuts the frame rate in half. 

Moving an aircraft increases CPU time by another factor of 2, doubles the GPU time per frame but does not change the VRAM usage.

When I reach 10,000' in the C414 and level out over Seattle, TLOD at 305 and OLOD at 200, the VRAM usage is still more than 11.2 GB out of 12 but the FPS is back to 60 and is very steady.

 

 

 

Edited June 5, 20251 yr by TacomaSailor

Could be that the plane is slightly more complex than the go-cart? Just a guess...

The same happens with the default TBM and many not so complex default liners in FS24.

I had the stutters once over the last three months – flying into NZCH which is an NZA payware.
And I initially thought it was traffic, but then I don't have any stutters with NZAA, which is arguably busier and also a payware airport.

Interestingly I don't get it in any other payware - NZTG, NZGS, NZWN, basically any of the NZA or Flightbeam catalogues (you know, other than NZCH).

I wonder if it's a set of specific sim objects / vehicles / etc that are causing issues with their LODs?

Edited June 6, 20251 yr by Georgleboui

6 hours ago, TacomaSailor said:

Beta 

I load the mkstudios KSEA airport (with zzz Static KSEA FSLTL) and the Flysimware C414 with DLSS at High, FG on, TLOD 305  OLOD 200 and all other settings on High except Buildings and Clouds which are Ultra.  I have Max Frame Rate set at 30 in the sim.  I have FSLTL running along with GAIST V7 and GAIST Trees and Scenery.  OR - just a whole bunch of stuff.

I start Free Flight with the Project Kart off road vehicle.  I can drive all over the airport, into the Cargo bays, around the Passenger terminals with a jet at almost every gate and the sim stays at a steady, rock solid 60fps. Both GPU time and GPU time are at 16 or 17 ms.  VRAM usage is 11.2 out of 12 GB and GPU busy is around 80%.  The scenery details are incredible, shadows are perfect and there is no melting of other strange things with the scenery or planes. 

I follow a 737 from the gate to the runway when they are ready to take off.  At the same time other aircraft are taking off and landing on the other two parallel runways.  Following under their tail or at a distance of 100-yards still a rock steady 60 fps.

The 737 accelerates for takeoff and I follow as close as possible and stay up with the jet to about 70 mph, and still 60fps.  As the jet pulls away I keep accelerating but they soon rotate and fly away.  At the same time a jet is taking off on a parallel runway.  Still 60 fps with no changes in the VRAM or CPU usage.

I then return to an empty GA gate where I load the Flysimware Cessna C414.  Shift -C into the cockpit and FPS drops to 35.  Shift-C out of the cockpit and look around the FPS drops to 25.   At this point the CPU time per frame has jumped up to over 100ms but the GPU stats are unchanged.

I return to the cockpit and prepare the C414 for flight.  As soon as I start to taxi - the FPS drops to single digits and the CPU time exceeds 250ms.  Still - the GPU stats remain the same.  As I prepare to takeoff the GPU time per frame climbs to about 60ms while the CPU exceeds 350 ms.  But, the VRAM usage does not change.

In summary:

Driving around a very detailed airport in a motor vehicle consumes ALL the VRAM but the FPS stays at the requested 60 per second. 

Following an AI plane in flight uses all the VRAM but the FPS stays at 60.

Loading a detailed aircraft, cold and dark, does not change the VRAM usage but does increase the CPU time by almost a factor of 10 and cuts the frame rate in half. 

Moving an aircraft increases CPU time by another factor of 2, doubles the GPU time per frame but does not change the VRAM usage.

When I reach 10,000' in the C414 and level out over Seattle, TLOD at 305 and OLOD at 200, the VRAM usage is still more than 11.2 GB out of 12 but the FPS is back to 60 and is very steady.

 

 

 

For starters your TLOD and OLOD are too high. Try lowering them and you'll see an improvement in regard to stutters around airports.

Lee

33 minutes ago, Midnight Music said:

For starters your TLOD and OLOD are too high. Try lowering them and you'll see an improvement in regard to stutters around airports.

I think I agree here. I have mine set around 100 and do not experience stutters at airports.

I have TLOD and OLOD at 250/200 respectively in MSFS 2020 using my 2015 specification PC in DX11 mode, and I do not experience stutters at airports (even very large and detailed ones like Aerosoft EDDF Frankfurt and EGLL London Heathrow). I know that this is MSFS 2024, but surely a 9800X3D/RTX 4070 Super 12GB powered system should be able to handle TLOD/OLOD settings only slightly higher than mine, even in DX12 mode?

EDIT: Framerates take a hit near Terminal 2 on the ground at EDDF Frankfurt, but that is an optimisation feature affecting framerates. There are no stutters.

Edited June 6, 20251 yr by Christopher Low

TLOD/OLOD == 100/100, used to be 50 / 50.

Made an EGLL-LPPT flight a couple days ago with the FSLabs 321, BATC and FSLTL AI, and for the first time I experienced a few stutters, nothing really special, taking off from RW 27R...

I've 8GB VRAM, 3060Ti GPU, and 32 GB RAM with 5600x CPU.

                             

Edited June 6, 20251 yr by jcomm

What screen resolution are you running MSFS 2024, jcomm? A third party developer stated to me recently that you really need 10GB VRAM just to run MSFS 2024 properly @ 1080p resolution, so I am curious to know how well it runs generally with only 8GB VRAM.

35 minutes ago, Christopher Low said:

What screen resolution are you running MSFS 2024, jcomm? A third party developer stated to me recently that you really need 10GB VRAM just to run MSFS 2024 properly @ 1080p resolution, so I am curious to know how well it runs generally with only 8GB VRAM.

FullHD (1920x1080),  60 Hz , 0,5 ms monitor.

Has been running smooth as silk, mostly stutter free, with default aircraft and the Fenix, now the FSLabs too. Also BATC and FSLTL AI aircraft models @ 2k (have had it @ 512, but decided to go 2k).

FPS limited to 30, now in SU3 beta you can do it in the GUI but I still have my limit in "UserCfg.opt".

 

 

Edited June 6, 20251 yr by jcomm

I guess I need to be more explicit in my observation.

The  density/detail/amount of scenery objects does not cause stuttering as shown by rapid movement around the airport in a kart.

Full usage of VRAM does not cause stuttering as shown by 60FPS with 11.2GB in use at both the airport and in flight.

A complex aircraft does not cause stuttering as shown by 60FPS over Seattle with a very detailed Seattle addon.

TLOD @305 and OLOD @200 does not cause stuttering as shown by 60 FPS while driving around KSEA with a very detailed addonor 60 FPS on final in a complex airplane at a complex airport with complex addon scenery. 

Stuttering only occurs when a complex airplane moves in a visual environment created by a detailed airport.  Lowering the TLOD and OLOD to 50 does not improve the single digit FPS.

My preliminary PIX analysis shows the problem occurs when DX12 "resource barriers" are used incorrectly to protect UAVs (unordered access views).  During periods with high frame rates there are few resource barriers created.  Resource barriers number in the thousands per frame during periods of low rates.   The resource barriers I see during times of very low FPS are all protecting UAVs.

Here are some notes from the DX12 App Development website.

ResourceBarrier notifies the graphics driver of situations in which the driver may need to synchronize multiple accesses to the memory in which a resource is stored.

An unordered access view (UAV) is a view of an unordered access resource (which can include buffers, textures, and texture arrays, though without multi-sampling). A UAV allows temporally unordered read/write access from multiple threads. This means that this resource type can be read/written simultaneously by multiple threads without generating memory conflicts. 

 

Performance implications

When recording explicit ResourceBarrier transitions on resources in the common state, it is correct to use either D3D12_RESOURCE_STATE_COMMON or any promotable state as the BeforeState value in the D3D12_RESOURCE_TRANSITION_BARRIER structure. This allows traditional state management that ignores automatic decay of buffers and simultaneous-access textures. This may not be desirable though, as avoiding transition ResourceBarrier calls with resources known to be in the common state can significantly improve performance. Resource barriers can be expensive. They are designed to force cache flushes, memory layout changes and other synchronization that may not be necessary for resources already in the common state. A command list that uses a resource barrier from a non-common state to another non-common state on a resource currently in the common state can introduce a lot of unneeded overhead.

Also, avoid explicit ResourceBarrier transitions to D3D12_RESOURCE_STATE_COMMON unless absolutely necessary (e.g. the next access is on a COPY command queue which requires a resources begin in the common state). Excessive transitions to the common state can dramatically slow down GPU performance.

 

 

Edited June 6, 20251 yr by TacomaSailor

17 hours ago, Christopher Low said:

What screen resolution are you running MSFS 2024, jcomm? A third party developer stated to me recently that you really need 10GB VRAM just to run MSFS 2024 properly @ 1080p resolution, so I am curious to know how well it runs generally with only 8GB VRAM.

It looks like you, jcomm and I  are in the same configuration range with the same challenges so here is some feedback.

I am beginning to ease into MSFS 2024 by establishing an availability and performance baseline.  My configuration is mid-range at best so I know I cannot just add a component simply because I like it.  It's add, measure and decide !  Thus far I have had a stutter-free experience testing with the Cessna Citation Longitude, FSLTL (some flights).  Checked the FPS 2-3 times.  It's consistently in the 37-42 FPS range.  I am not the FPS type.  I just care about stutter-free performance.  All scenery is what comes with the product, my Community folder has only FSLTL.  I am now waiting for the PMDG B77W and for Fenix to release their BFU.  Hopefully my experience with these products will be the same as the Longitude and be able to use them.  If not, my flight simulation future will be lots of MSFS 2020 and a little MSFS 2024.  I have no intention of upgrading my current configuration, which has been performing very well for 2.5 years.

Here is my configuration:

• CPU  i7-12700K
• Memory 32GB DDR5-4400
• GPU Nvidia 3070ti 8GB VRAM
• Monitor 2560 x 1440 (Max refresh rate 144 MHz)
• HAGS is OFF
• HDR is OFF

So you are running MSFS 2024 @ 2560x1440 resolution with 8GB VRAM? Your PC is actually a fair bit more powerful than mine, but that screen resolution would bring the comparison a bit more into line.