A SU2 Beta Performance Comparison

A SU2 Beta Performance Comparison

The following are the results of four performance comparison flights in MSFS 2024 SU2 v1.4.7.0

Prior to each test (except the last), MSFS was exited and the various caches (including the rolling cache) deleted to ensure that the results were not skewed by any data that might be stored in caches.

HWiNFO64 was used to collect the performance data. The collection interval was 1 second, starting at the point the engines were started at on the apron at San Gabriel Valley & ending when the aircraft reached the parking spot at French Valley. Each flight took ~ 32 minutes.

Details on hardware, graphical & other settings are provided at the end of this post.

Like the previous comparison flights in SU1, the VFR flight was flown at ~ 2,000 ft using the Cubcrafters NX Cub from San Gabriel Valley (KEMT) to French Valley (F70) following freeways & major roads. A LittleNavMap flight plan is enclosed for those that wish to recreate it:
Beta Test Flight - San Gabriel Valley (KEMT) to French Valley (F70) v2.pln (35.7 KB)

  • The first comparison flight is between MSFS 2024 SU2 v1.4.7.0 and MSFS 2024 SU1 v1.3.24.0

  • The purpose of the second comparison flight was to compare the first comparison flight - where Frame Generation was off (None) + HDR 10 was On - with Frame Generation set to On (AMD FSR 3) + HDR 10 set to Off. All other settings remained the same.

  • The purpose of the third comparison flight was to compare the second flight (with Frame Generation On (AMD FSR 3) + HDR 10 Off)but with the Dynamic Settings now enabled with a Frame Rate Target of 40 FPS. All other settings remained the same.

  • Finally, the fourth comparison flight is the same as the third, except this time the Rolling & other caches were not removed prior to the flight.

Framerates

This chart shows the framerates for the four flights:

Chart Legend Description
v1.3.24.0: SU1 performance results
FrameGen Off: 1st Flight: Frame Generation Off (None), HDR 10 On
FrameGen On: 2nd Flight: Frame Generation On (AMD FSR 3), HDR 10 Off
DS-40, RC-Empty: 3rd Flight. Dynamic Settings enabled, Frame Rate Target of 40 FPS. Caches deleted before flight.
DS-40, RC-Full: 4th Flight. Dynamic Settings enabled, Frame Rate Target of 40 FPS. Caches not removed prior to the flight.

  • SU2 shows a modest increase in framerates (even with temporary memory logging enabled, as per the release notes.)
    However, as with SU1, stuttering was very noticeable throughout the flight, and it made a smooth approach & landing at French Valley difficult.
  • Enabling Frame Generation significantly improves performance, with FPS increasing from the 30s into the 40s - 50s. The median increase was almost 60%!
    There appeared to be a slight reduction in stuttering, but it was still noticeable throughout the flight. Approach & landing at French Valley was still negatively impacted.
  • Enabling Dynamic Settings improves performance even further, with FPS now regularly staying in the high 50s - an almost 80% increase over the SU1 median value!
    Furthermore, stuttering was virtually eliminated. Changing views, panning in external& drone-mode, and conducting a smooth and controlled appraoach & landing was now possible!
  • Having data stored in the rolling cache did not make an appreciable difference compared to an empty rolling cache, although the variance value did half from 20.71 to 10.23, which you can see in the chart as the fourth flight had a slightly more compressed range than the third.
    As with the third flight, stuttering was virtually eliminated.
Framerate v1.3.24.0 FrameGen Off FrameGen On DS-40, RC-Empty DS-40, RC-Full
Median Framerates 32.90 36.20 52.10 58.70 58.90
% Change from SU1 N/A 10.0% 58.4% 78.4% 79.0%

Total CPU Usage (%)

There appears to have been some optimizations made in SU2, as shown in this chart:

  • SU2 sees a median reduction in total CPU usage of 6%.
  • Enabling Frame Generation reverses that benefit - there is a 6% increase in median total CPU usage compared to SU1.
  • Enabling Dynamic settings makes a huge difference - a 17% (RC-Empty) to 24% (RC-Full) decrease from SU1.
CPU Usage v1.3.24.0 FrameGen Off FrameGen On DS-40, RC-Empty DS-40, RC-Full
Median 30.9 29.0 32.7 25.6 23.3
% Change from SU1 N/A -6.1% 5.8% -17.2% -24.6%

GPU Memory Usage (%)

SU2 can provide significant benefits if Dynamic settings is enabled:

  • SU2 sees a small median reduction in total GPU usage of ~ 2%
  • Enabling Frame Generation causes a 7% increase.
  • Enabling Dynamic settings reduces memory usage by 20% (RC-Empty) to 27% (RC-Full) from SU1.
GPU Memory v1.3.24.0 FrameGen Off FrameGen On DS-40, RC-Empty DS-40, RC-Full
Median 87.60 85.30 94.10 69.70 64.20
% Change from SU1 N/A -2.6% 7.4% -20.4% -26.7%

GPU Total Usage (%)

The Intel Arc 770 was basically running flat-out for each of the flights. As the table below shows, there was no real difference in usage between any of the flights:

Total GPU v1.3.24.0 FrameGen Off FrameGen On DS-40, RC-Empty DS-40, RC-Full
Median 97.9 98.5 98.5 98.6 98.5
% Change from SU1 N/A 0.61% 0.61% 0.72% 0.61%

Network

Because the SU1 test was conducted over 3 weeks ago, the 17% reduction in the median download rate may have been caused by differing connectivity conditions, so I’m not reading to much into that observation.

Current DL rate [KB/s] v1.3.24.0 FrameGen Off FrameGen On DS-40, RC-Empty DS-40, RC-Full
Median 3,938 3,253 4,033 1,574 1,510
% Change from SU1 0.00% -17.41% 102.40% -60.03% -61.65%
% Change from previous flight N/A -17.41% 23.99% -60.97% -4.06%

Of more interest are the four SU2 flights that were conducted within a 3-hour time-period.

  • Enabling Frame Generation causes a 24% increase in median download rate.
  • Enabling Dynamic settings causes a 61% decrease in median download rate with Frame Generation enabled
  • Despite the rolling cache being retained for the last flight, the network download rates between Flight #3 and #4 are remarkably similar.

Conclusion

The addition of Frame Generation in SU2 has provided significant increase in performance (FPS) to my PC system.

The increase in framerates to above 40 FPS for the majority of a flight has now made enabling Dynamic settings a viable proposition (previously, when framerates rarely reached the high 30’s setting, configuring Dynamic Settings with a Frame Rate Target of 40 FPS made no sense).

Enabling Dynamic settings has resulted in virtually stutter-free flights that now match the smoothness of flights seen in MSFS 2020 SU15.

Further testing of SU2 is needed to see if these benefits carry over to airliners with complex displays during taxiing, taking off, appraoch and landing at major hub airports, as well as how enabling AI traffic, weather & multiplayer affect the results.

Expand to see flight settings
Item Setting
Aircraft Cubcrafters NX Cub
Flightplan KEMT to French Valley F70
Flight configuration - Weather Clear Skies
Flight configuration - Air Traffic Off
Flight configuration - Multiplayer Off
Expand to see basic PC specs:
Item Setting
CPU i7-12700K (no overclocking or undervolting)
RAM 32 GB 5200 MHz DDR5 CL40 RAM
GPU Intel Arc 770 (16GB, v32.0.101.6314)
OS Windows 11 Pro 24 H2 v26100.3194 + all OS updates.
MSFS Edition MS Store Premium Deluxe
Build Version SU1 Beta v1.4.7.0
3rd-Party & Community Addons None

No performance-improving applications are installed (e.g. for CPU or GPU improvements).
Neither the Intel Arc Control Service or Arc Control App is running.

Expand to see graphics settings
Item Setting
Display Mode Full Screen
HDR10 On
Full Screen Res 3840 x 2160
Anti-Aliasing TAA
Render scaling 65 (Resolution: 2496 x 1404)
Frame Generation None (NEW: Was not present in SU1)
AMD FidelityFX Sharpening 150
VSync Off
Frame Rate Limit N/A
Dynamic Settings Off
Frame Rate Target N/A
Global Rendering Quality Custom
Terrain LOD 400
Off Screen Terrain Pre-Cache Ultra
Displacement Mapping On
Buildings High
Trees Medium
Plants Medium
Rocks High
Grass Medium
Object LOD 200
Volumetric Clouds Ultra
Texture Resolution Medium
Anisotropic Filtering 4x
Water Waves Medium
Raytraced Shadows On
Shadow Maps 2048
Terrain Shadows 512
Contact Shadows High
Windshield Effects Medium
Ambient Occulsion High
Cubemap Reflections 256
Raymarched Reflections Medium
Light Shafts High
Depth of Field Off
Motion Blur Off
Glass Cockpit Refresh Rate High
Characters Quality Medium
Traffic Airport Quality Medium
Air Traffic Medium
Road Traffic Medium
Sea Traffic Medium
Fauna High
Expand to see other settings
Item Setting
Online Tab
Photogrammetry On
Live Real-World Air Traffic Off
Live Weather Off
Multiplayer Off
Multiplayer Servers West USA
Rolling Cache On @ 16GB
Bandwidth Unlimited
Advanced Options
Developer Mode Off
26 Likes

Wow! Clearly, a whole lot of effort went into testing and documenting/formatting the results. Thank you for putting this together.

3 Likes

When passionate nerds play technical games, THIS is the A+ data you get. Well done and its clear you did a TON of work on this.

3 Likes

The following are the results of a performance comparison flight for MSFS 2024 SU2 v1.4.10.0

Testing was conducted in the same manner as recorded in the first post in this topic.

Note that for this test, HDR 10 was off. In the v1.4.7.0 flights, turning on HDR 10 caused a “grey wash” effect, so was kept off. This issue appears solved on my PC in v1.4.10.0, but HDR 10 was left off to keep the settings between the v.7 & v.10 flights the same.

Framerates

The Release Notes for v1.4.10.0 show no changes under “Stability & Performance” that would suggest that performance would be significantly different in this build.

This chart shows the framerates for the flight:

Chart Legend Description
v1.3.24.0: SU1 performance results from the first post (No Frame Generation, Dynamic Settings disabled).
v1.4.7.0: 3rd flight in the first post (“DS-40, RC-Empty”). Frame Generation On (AMD FSR 3), HDR 10 Off, Dynamic Settings enabled, Frame Rate Target of 40 FPS.
v1.4.10.0: Frame Generation On (AMD FSR 3), HDR 10 Off, Dynamic Settings enabled, Frame Rate Target of 40 FPS, Rolling Cache empty.

As expected, there is little difference in framerates between the v.7 & v.10 builds. The v.10 flight was smooth & virtually stutter-free. There is some improvement to the variance ( from 20.71 to 11.75), which you can see in the chart as the v.10 flight had a slightly more compressed range than the v.7 build.

Framerate v1.3.24.0 v1.4.7.0 v1.4.10.0
Median Framerates 32.90 58.70 58.30

Total CPU Usage (%), GPU Memory Usage (%) and GPU Total Usage (%)

Median values between the two Beta builds are also similar:

CPU Usage v1.3.24.0 v1.4.7.0 v1.4.10.0
Median 30.9 25.6 23.6
GPU Memory v1.3.24.0 v1.4.7.0 v1.4.10.0
Median 87.60 69.70 62.10
Total GPU v1.3.24.0 v1.4.7.0 v1.4.10.0
Median 97.9 98.6 98.5

Conclusion
There has been no degradation in performance in this build. On my PC, the v1.4.10.0 build provides the same benefits as the v1.4.7.0 build.

Expand to see flight settings
Item Setting
Aircraft Cubcrafters NX Cub
Flightplan KEMT to French Valley F70
Flight configuration - Weather Clear Skies
Flight configuration - Air Traffic Off
Flight configuration - Multiplayer Off
Expand to see basic PC specs:
Item Setting
CPU i7-12700K (no overclocking or undervolting)
RAM 32 GB 5200 MHz DDR5 CL40 RAM
GPU Intel Arc 770 (16GB, v32.0.101.6314)
OS Windows 11 Pro 24 H2 v26100.3194 + all OS updates.
MSFS Edition MS Store Premium Deluxe
Build Version SU2 Beta v1.4.10.0
3rd-Party & Community Addons None

No performance-improving applications are installed (e.g. for CPU or GPU improvements).
Neither the Intel Arc Control Service or Arc Control App is running.

Expand to see graphics settings
Item Setting
Display Mode Full Screen
HDR10 Off
Full Screen Res 3840 x 2160
Anti-Aliasing TAA
Render scaling 65 (Resolution: 2496 x 1404)
Frame Generation AMD FSR 3
Framerate Multiplier 2x
AMD FidelityFX Sharpening 150
Exposure Compensation EV 0
VSync Off (N/A)
Frame Rate Limit 100% MRR (N/A)
Dynamic Settings On
Frame Rate Target 40
- -
Global Rendering Quality Custom
Terrain LOD 400
Off Screen Terrain Pre-Cache Ultra
Displacement Mapping On
Buildings High
Trees Medium
Plants Medium
Rocks High
Grass Medium
Object LOD 200
Volumetric Clouds Ultra
Texture Resolution Medium
Anisotropic Filtering 4x
Water Waves Medium
Raytraced Shadows On
Shadow Maps 2048
Terrain Shadows 512
Contact Shadows High
Windshield Effects Medium
Ambient Occulsion High
Cubemap Reflections 256
Raymarched Reflections Medium
Light Shafts High
Depth of Field Off
Motion Blur Off
Glass Cockpit Refresh Rate High
Characters Quality Medium
Traffic Airport Quality Medium
Air Traffic Medium
Road Traffic Medium
Sea Traffic Medium
Fauna High
Expand to see other settings
Item Setting
Online Tab
Photogrammetry On
Live Real-World Air Traffic Off
Live Weather Off
Multiplayer Off
Multiplayer Servers West USA
Rolling Cache On @ 16GB
Bandwidth Unlimited
Advanced Options
Developer Mode Off
4 Likes

I get slightly higher framerates. But performance in v1.4.10.0 is the best of all builds imo. Stutters are much less frequent.

2 Likes

Very cool testing! Thank you!
I wander why did you put dynamic settings target framerate as 40… If I do the same I get really awful terrain graphics.

1 Like

Hi @PertainedMaple7 and welcome to the forums!

My Arc 770 is a mid-range GPU. With the graphic settings I’ve chosen, 40 FPS gives good performance with TLOD = 400.

40 gives a smooth virtually stutter-free flight without over-taxing the GPU memory at my render scaling on my 55" TV. I could probably bump the target framerate to 60, but might encounter issues on the odd occasion when I fly an airliner into a major hub airport.

As I mentioned in my 1st post, further testing of SU2 is needed to see if these benefits carry over to airliners with complex displays during taxiing, taking off, approach and landing at major hub airports, as well as how enabling AI traffic, weather & multiplayer affect the results.

2 Likes

Same here. For me the dynamic settings are way too aggressive. I set it to 40 FPS and it is like setting my LOD to 50. I can turn dynamic off and get a smooth experience with LOD set to 150.

Thank you for making that clear! I have a 3070ti GPU, it’s only 8 Gb vram… in the end I turned off dynamic settings and I set TLOD at 100. That way my GPU can push about 50fps without tricks like frame gen.

Since the release notes for v.14.12.0 show no changes that were specifically targeted to change performance, and the performance in my flights so far in this build had all been very similar to the v.10 build (reported previously), I thought I’d see how performance stacks up using a more complex aircraft than the NC Cub, flying from and to larger airports.

For this series of tests, I chose the Cessna Citation CJ4, leaving from Gate 3 at Fresno Yosemite (KFAT) and flying direct to Los Angeles (KLAX). Three flights with different weather presets (Clear Skies, Broken Clouds & Rain) were conducted. Flight time was ~ 42 minutes at 10,000 feet. In addition, a short “Taxiing around KLAX” flight was done. AI traffic was also enabled.

For those interested, here is the flight plan:
Beta Test Flight (SU2) Fresno (KFAT) to Los Angeles (KLAX).pln (1.4 KB)

As with my previous tests, MSFS 20204 was restarted before each test in order to delete the caches (including the rolling cache), so not to skew results with cached data.

A full list of hardware & game settings is provided at the end of this post, but the main graphical settings for the v1.4.12.0 flights were:

  • Frame Generation On (AMD FSR 3)
  • HDR 10 On
  • Dynamic Settings enabled with a Frame Rate Target of 60 FPS
  • Air Traffic set to “AI”
  • Multiplayer set to “Off”
  • Time set to: April 12th 2025, 07:00

Note that because the v1.3.24.0 and v1.4.10.0 flights were conducted using different settings, aircraft and flightplan, one cannot make a direct performance comparison between these and the 1.4.12.0 flights. I’ve included these two flights in the charts mainly as reference points.

Framerates

Chart Legend Description
v1.3.24.0: SU1 performance results (No Frame Generation, Dynamic Settings disabled, Air Traffic: Off, KEMT to F70).
v1.4.10.0: SU2 Beta (Frame Generation On (AMD FSR 3), HDR 10 Off, Dynamic Settings enabled, Frame Rate Target of 40 FPS, KEMT to F70).
v1.4.12.0-CS SU2 Beta - Clear Skies.
v1.4.12.0-BC SU2 Beta - Broken Clouds
v1.4.12.0-RN SU2 Beta - Rain
v1.4.12.0-TX SU2 Beta - KLAX Taxiing Test. Gate 18B to Parking 321 via taxiways D, K and C. This “flight” took just under 9 minutes.

Framerate v1.3.24.0 v1.4.10.0 v1.4.12.0-CS v1.4.12.0-BC v1.4.12.0-RN v1.4.12.0-TX
Median Framerates 32.90 58.30 60.50 55.80 57.10 49.00

This data shows:

  1. The clear performance (FPS) benefit Frame Generation provides to my mid-range PC hardware.
  2. Changing the aircraft from the “simple” NX Cub to the “more complex” Cessna Citation CJ4 did not cause a reduction in performance. In fact, there was a slight uptick (3.8%) in median framerates in the clear-skies flight with the Cessna Citation CJ4.
  3. Enabling weather results in a slightly lower framerate (5.6 to 7.8 % decrease). The fact that “Broken Clouds” had a lower framerate than “Rain” is probably due to the fact that the cruise portion of the “Rain flight” was flying in clear air between two cloud layers, whilst in “Broken Clouds” the cruise portion was flying in and out of clouds throughout the flight.
  4. The scenery density & complexity of KLAX significantly reduces framerates (the median dropped by 19%).

Subjectively, whilst monitoring the Game Bar FPS counter during the flight, I was seeing the following frame rates:

Flight Segment Clear Skies Broken Clouds Rain
Fresno taxiing & take-off Low 50’s High 40’s Low 50’s
Cruise altitude Constant at 59-60 Constant at 57-58, drops to 54-56 in external & drone view Constant at 58-60, drops to mid 50’s in external & drone view, drops a few FPS when in clouds
Los Angeles approach, landing & taxiing Low 50’s Mid-to-high 40’s Low 50’s
Taxiing Test High 40s to low 50s. Popping of taxiway signs (less obvious now with fade-in), blast fences (very obvious) & vehicle LOD boxes (very bad, as you can see in the cropped screenshot below):

For all flights, the entire flight from start to finish was very smooth with no stutters.

Visuals.
Here are a couple of screenshots of from the Clear Skies flight:

GPU Memory Usage (%)

GPU Memory v1.3.24.0 v1.4.10.0 v1.4.12.0-CS v1.4.12.0-BC v1.4.12.0-RN v1.4.12.0-TX
Median 87.60 62.10 79.00 83.10 77.00 83.60
% Change from .10 41 N/A 27 34 24 35

There is an clear increase in GPU Memory usage in the .12 builds - the median usage increases by 24% - 35%. However, even when taxiing around LAX, the median remains below 84%. Flying through clouds causes the most memory usage for brief periods (96%).

The chart below shows how each flight uses the GPU in relation to the other flights. The v1.4.10 build has the lowest usage (but remember, this is with the NX Cub, a different flight plan and no AI aircraft). None of the .12 builds were below 70%.

Of the three 1.4.12.0 flights, the Clear skies & Rain flights use less VRAM than the Broken Clouds. the GPU memory taxiing around LAX for nine minutes consistently used 85% +/-

Visuals:
Here are a couple of screenshots of from the Broken Clouds flight:

GPU Total Usage (%) and Total CPU Usage (%)

Total GPU v1.3.24.0 v1.4.10.0 v1.4.12.0-CS v1.4.12.0-BC v1.4.12.0-RN v1.4.12.0-TX
Median 97.9 98.5 98.40 98.50 98.50 98.80
CPU Usage v1.3.24.0 v1.4.10.0 v1.4.12.0-CS v1.4.12.0-BC v1.4.12.0-RN v1.4.12.0-TX
Median 30.9 23.6 27.20 25.60 26.00 35.40

No surprise that I’m GPU limited with my Intel Arc 770. The GPU is basically running flat-out throughout the flight. My i7-12700K is not having any problems handling the various processing requests.

Visuals:

Conclusion
Overall, I’m very satisfied with the performance of these flights. I’m not seeing any stuttering on take off or landing and visually the ground textures look good on my 55" TV at a resolution of 2496 x 1404. The main visual problem I see with my settings & hardware is the LOD culling of airport objects such as aircraft without landing gear, box-like vehicles, aircraft popping in & out of existence, etc. (all of which have been reported & bug-logged elsewhere).

I’d still like to run at 4K as I can do with 2020. Time will tell if that will ever be possible.

Expand to see flight settings
Item Setting
Aircraft Cessna Citation CJ4
Flightplan Fresno Yosemite (KFAT) to Los Angeles (KLAX)
Flight configuration - Weather Clear Skies; Broken Clouds; Rain, Time: April 12th 2025, 07:00
Flight configuration - Air Traffic AI
Flight configuration - Multiplayer Off
Expand to see basic PC specs:
Item Setting
CPU i7-12700K (no overclocking or undervolting)
RAM 32 GB 5200 MHz DDR5 CL40 RAM
GPU Intel Arc 770 (16GB, v32.0.101.6314)
OS Windows 11 Pro 24 H2 v26100.3194 + all OS updates.
MSFS Edition MS Store Premium Deluxe
Build Version SU2 Beta v1.4.12.0
3rd-Party & Community Addons None

No performance-improving applications are installed (e.g. for CPU or GPU improvements).
Neither the Intel Arc Control Service or Arc Control App is running.

Expand to see graphics settings
Item Setting
Display Mode Full Screen
HDR10 On
Full Screen Res 3840 x 2160
Anti-Aliasing TAA
Render scaling 65 (Resolution: 2496 x 1404)
Frame Generation AMD FSR 3
Framerate Multiplier 2x
AMD FidelityFX Sharpening 150
Exposure Compensation EV -1
VSync Off (N/A)
Frame Rate Limit 100% MRR (N/A)
Dynamic Settings On
Frame Rate Target 60
- -
Global Rendering Quality Custom
Terrain LOD 400
Off Screen Terrain Pre-Cache Ultra
Displacement Mapping On
Buildings High
Trees Medium
Plants Medium
Rocks High
Grass Medium
Object LOD 200
Volumetric Clouds Ultra
Texture Resolution Medium
Anisotropic Filtering 4x
Water Waves Medium
Raytraced Shadows On
Shadow Maps 2048
Terrain Shadows 512
Contact Shadows High
Windshield Effects High
Ambient Occulsion High
Cubemap Reflections 256
Raymarched Reflections Medium
Light Shafts High
Depth of Field Off
Motion Blur Off
Glass Cockpit Refresh Rate High
Characters Quality Medium
Traffic Airport Quality Medium
Air Traffic Medium
Road Traffic Medium
Sea Traffic Medium
Fauna High
Expand to see other settings
Item Setting
Online Tab
Photogrammetry On
Aircraft Traffic Type AI Offline
Live Weather On
Multiplayer On
Multiplayer Servers West USA
Show MP aircraft in close proximity On
Rolling Cache On @ 16GB
Bandwidth Unlimited
Advanced Options
Developer Mode Off
7 Likes

I completed another test flight over the weekend in v1.4.12.0, reproducing the v1.4.12.0-BC (Broken Clouds) test mentioned in my previous post, but replacing the Cessna Citation CJ4 with the default Boeing 747-8.

The purpose was to see if my PC & settings could handle a large jet conducting flight operations at a large airport - in this scenario, landing at LAX with AI traffic enabled. This flight took slightly longer than the flight in the CJ4.

Framerates

Chart Legend Description
v1.4.12.0-BC SU2 Beta - Broken Clouds. These are the same results from the test reported in my previous post.
v1.4.12.0-B747 SU2 Beta - Broken Clouds. Identical to v1.4.12.0-BC, but with the 747-8 rather than the CJ4.

Flight v1.4.12.0-BC v1.4.12.0-B747
Median Framerate 55.80 57.00

Subjectively, whilst monitoring the Game Bar FPS counter during the flight, I was seeing the following frame rates:

Flight Segment v1.4.12.0-BC v1.4.12.0-B747
Fresno taxiing & take-off High 40’s Cockpit view: Mid 50’s; Pushback: 49-50; Taxi:45; Take-off: Low to mid 40’s.
Cruise altitude Constant at 57-58, drops to 54-56 in external & drone view Cockpit: 58-60; Exterior 49-53; Drone 46-49
Los Angeles approach, landing & taxiing Mid-to-high 40’s Mid to high 40’s

The median performance of the Boeing 747 flight was marginally better, but I don’t think 1.2 FPS is statistically significant. The main takeaway here is that even though there is more variability in framerates - especially during take-off - the complexity of the 747 did not detrimentally impact performance. There were two slight stutters on final at LAX, but not long enough to impact the control of the aircraft.

Visuals.
Here are a couple of screenshots:

GPU Memory Usage (%)

Flight v1.4.12.0-BC v1.4.12.0-B747
Median GPU Memory 83.10 82.10

Again, the median performance of the Boeing 747 flight was marginally better, but not statistically significant. What is noticeable is that 90%+ of the available 16 GB of VRAM is utilized whilst at both Fresno & Los Angeles airports. The chart clearly shows that immediately after take-off, memory is freed-up, and remains for the most part at or below 85% for the cruise portion of the flight, then rises again to over 90% as one is on final, landing & taxiing.

Taxiing after landing at KLAX.
This test followed the same taxiways as the previous test (D, K and C).
At this point, 15.3 of the 16 GB of VRAM was in use:

LittleNavMap shows the aircraft displayed on arrival at the gate:

As one moves along the taxiway, more aircraft are displayed (i.e. pop into existence):

Note that the AI aircraft are only added, not removed as one moves away from them:

Visual:

Despite this, taxiing was generally in the mid to low 40’s, and stutter-free.

By the time I reached my parking spot, this had increased to 15.5 GB:

Network Usage.

The loading time for this flight was a little over 1 minute. During that time, 0.96 GB of data was downloaded at an average speed of 14 MB/Sec. The entire flight (ending when parked at the gate) downloaded 3.39 GB. The maximum download rate reported was 27.40 MB/sec.

5 Likes

The following are the results of three performance comparison flights for MSFS 2024 SU2 v1.4.14.0

Testing was conducted in the same manner as recorded in the first post in this topic. Full PC & graphics specs are provided at the end of this post. Flight times were around 32 - 35 minutes.

The purpose of these tests was:

  1. To compare the performance of the .14 build with that of the .10 build.
  2. Investigate if changing Anti Aliasing from TAA with Render Scaling at 65% to AMD FSR3 Quality (67% render scale) improves performance.
  3. Investigate what impact a few changes to the graphics settings has on performance.

Framerates

This chart shows the framerates for the flight:

Chart Legend Description
v1.4.10.0-TAA: Base measurement. TAA (Render scaling @ 65), Frame Generation On (AMD FSR 3), Dynamic Settings enabled, Frame Rate Target @40 FPS, HDR 10 Off
v1.4.14.0-TAA TAA (Render scaling @ 65), Frame Generation On (AMD FSR 3), Dynamic Settings enabled, Frame Rate Target @40 FPS, HDR 10 Off
v1.4.14.0-AMD AMD FSR 3 (Quality), Frame Generation On (AMD FSR 3), Dynamic Settings enabled, Frame Rate Target @40 FPS, HDR 10 Off

Framerate v1.4.10.0 v1.4.14.0-TAA v1.4.14.0-AMD v1.4.14.0-AMD-Plus
Median 58.30 60.5 57.5 54.8

All the v.14 flights were smooth & stutter-free. Of the 3 flights, v1.4.14.0-TAA had the highest framerates. This is more clearly seen in the Frequency chart below, where the v.14-TAA version outperforms the v.10 and v.14-AMD flights. Note, however, that the AMD FSR 3 Quality setting is rendering at 67%, 2% greater than TAA’s 65%, which could explain the 3 FPS median difference between the two flights.

The v.14-TAA framerates are very similar to the v1.4.12.0-CS framerates reported earlier using the Cessna Citation CJ4.

GPU Memory Usage

This chart shows the frequency count for the percentage of the 16 GB of VRAM used on my Arc 770:

GPU Memory % v1.4.10.0 v1.4.14.0-TAA v1.4.14.0-AMD v1.4.14.0-AMD-Plus
Median 62.2.0 67.6 77.7 76.6

The v.10-TAA build has the lowest usage, with the v.14-TAA build using around 5 percent more. AMD FSR 3 has the greatest usage of the three, about 10 % more than the .14-TAA build and fifteen percent more than the .10-TAA build.

Total CPU Usage (%) and GPU Total Usage (%)

There is little difference between any of the builds for these two metrics. As reported previously, my PC is GPU-limited, with the GPU running flat-out for the entire flight, and the CPU cruising along with plenty of additional processing capacity available if needed:

Total GPU v1.4.10.0 v1.4.14.0-TAA v1.4.14.0-AMD v1.4.14.0-AMD++
Median 98.5 98.5 98.4 98.6
CPU Usage v1.4.10.0 v1.4.14.0-TAA v1.4.14.0-AMD v1.4.14.0-AMD++
Median 23.6 23.5 22.4 22.3

Changing some settings while using AMD FSR 3

The third flight shows the results of changing a few of the graphics settings. As expected, increasing these settings causes a small but noticable decrease in framerates:

Chart Legend Description
v1.4.14.0-AMD AMD FSR 3 (Quality), Frame Generation On (AMD FSR 3), Dynamic Settings enabled, Frame Rate Target @40 FPS, HDR 10 Off
v1.4.14.0-AMD-Plus v1.4.14.0-AMD, but with the following changes: HDR 10 On; Exposure CEV: -1; Grass: High, Rocks: Medium; Windshield Effects: High; Light Shafts: Ultra; Fauna: Ultra.

Framerate v1.4.14.0-AMD v1.4.14.0-AMD-Plus
Median 57.5 54.8

Conclusion

For my PC specs & settings, TAA is the better Anti-Aliasing option than AMD FSR 3.

The v1.4.14.0 build provides the same performance as the v1.4.12.0 build, with neither dramatic improvements nor regression. Both are the best performing builds in the SU to-date on my PC.

Expand to see flight settings
Item Setting
Aircraft Cubcrafters NX Cub
Flightplan KEMT (Parking # 48) to French Valley F70 Rwy 18
Flight configuration - Time & Weather 04/19/25 @ 16:00 hrs; Clear Skies
Flight configuration - Air Traffic Off
Flight configuration - Multiplayer Off
Expand to see basic PC specs:
Item Setting
CPU i7-12700K (no overclocking or undervolting)
RAM 32 GB 5200 MHz DDR5 CL40 RAM
GPU Intel Arc 770 (16GB, v32.0.101.6314)
OS Windows 11 Pro 24 H2 v26100.3194 + all OS updates.
MSFS Edition MS Store Premium Deluxe
Build Version SU2 Beta v1.4.14.0
3rd-Party & Community Addons None

No performance-improving applications are installed (e.g. for CPU or GPU improvements).
Neither the Intel Arc Control Service or Arc Control App is running.

Expand to see Base graphics settings
Item Setting
Display Mode Full Screen
HDR10 Off
Full Screen Res 3840 x 2160
Anti-Aliasing TAA
Render scaling 65 (Resolution: 2496 x 1404)
Frame Generation AMD FSR 3
Framerate Multiplier 2x
AMD FidelityFX Sharpening 150
Exposure Compensation EV 0
VSync Off (N/A)
Frame Rate Limit 100% MRR (N/A)
Dynamic Settings On
Frame Rate Target 40
- -
Global Rendering Quality Custom
Terrain LOD 400
Off Screen Terrain Pre-Cache Ultra
Displacement Mapping On
Buildings High
Trees Medium
Plants Medium
Rocks High
Grass Medium
Object LOD 200
Volumetric Clouds Ultra
Texture Resolution Medium
Anisotropic Filtering 4x
Water Waves Medium
Raytraced Shadows On
Shadow Maps 2048
Terrain Shadows 512
Contact Shadows High
Windshield Effects Medium
Ambient Occulsion High
Cubemap Reflections 256
Raymarched Reflections Medium
Light Shafts High
Depth of Field Off
Motion Blur Off
Glass Cockpit Refresh Rate High
Characters Quality Medium
Traffic Airport Quality Medium
Air Traffic Medium
Road Traffic Medium
Sea Traffic Medium
Fauna High
Expand to see other settings
Item Setting
Online Tab
Photogrammetry On
Live Real-World Air Traffic Off
Live Weather Off
Multiplayer Off
Multiplayer Servers West USA
Rolling Cache On @ 16GB
Bandwidth Unlimited
Advanced Options
Developer Mode Off
5 Likes

Great stuff - thank you!