Find out whether Prepar3D is CPU- or GPU-bound with a repeatable test, then change the settings that actually improve FPS and reduce stutters.
Prepar3D is CPU-bound when lowering display resolution and anti-aliasing barely changes FPS, but reducing traffic, autogen or scenery load helps. It is GPU-bound when a lower resolution or lighter anti-aliasing gives a repeatable FPS gain. Test the same saved flight with VSync and frame caps temporarily disabled.
Is Prepar3D normally CPU- or GPU-bound?
Prepar3D has no single, permanent bottleneck. Complex aircraft, busy airports, AI traffic and dense scenery commonly expose a main-thread CPU limit, while high resolutions, anti-aliasing, shadows, clouds and dynamic lighting can shift the limit to the GPU.
The bottleneck can change during one flight. A large airport may be CPU-bound at the gate, while cloud, lighting and anti-aliasing loads can make the same system GPU-bound after departure. That is why hardware utilisation from one quiet location tells us very little.
How to test whether Prepar3D is CPU- or GPU-bound
- Create a repeatable test. Save a flight at a demanding airport with the aircraft, weather, traffic and cockpit view you normally use. Reload that flight for every comparison and allow initial scenery loading to settle.
- Remove artificial limits. Temporarily disable VSync, external frame limiters and Prepar3D's own frame cap. If FPS remains fixed at the monitor refresh rate or another exact value, the test is still capped.
- Record the baseline. Display Prepar3D's frame rate and use Windows Task Manager or an existing monitoring utility to watch individual CPU logical processors, GPU 3D utilisation and dedicated GPU memory. On systems with two graphics processors, make sure you are monitoring the GPU running
Prepar3D.exe. - Run the GPU test. Lower only the display resolution, then repeat with lighter anti-aliasing. Leave traffic, scenery and autogen unchanged. A clear FPS increase that repeats over several runs identifies a GPU limitation.
- Run the CPU test. Restore the graphics settings, set AI and vehicle traffic to zero, then reduce autogen, scenery complexity or object draw distance. If these changes help while resolution changes did not, Prepar3D is CPU-bound in that scenario.
- Confirm elsewhere. Repeat the comparison at a second demanding location. Tune for the situation where poor performance matters, rather than an empty default airfield that already runs well.
Change one setting group at a time. A mistake we see constantly is lowering resolution, traffic, clouds and scenery together; FPS improves, but the simmer learns nothing about the actual constraint.
How do the results identify the bottleneck?
| Observed result | Likely limit | What to adjust first |
|---|---|---|
| Lower resolution or anti-aliasing produces a repeatable FPS gain | GPU rendering load | Resolution, anti-aliasing, shadows and dynamic lighting |
| Resolution changes little, but reducing traffic or scenery helps | CPU main thread | AI traffic, autogen, scenery complexity and demanding add-ons |
| GPU memory is full and panning causes pauses or texture replacement | VRAM pressure | Texture resolution, anti-aliasing, shadows and dense scenery |
| Both tests improve FPS | Balanced or mixed limit | Reduce the settings responsible for the worst flight conditions |
| Neither test changes a fixed frame rate | Frame cap or VSync | Remove the cap before testing again |
Which Prepar3D settings improve CPU-bound performance?
For a CPU-bound installation, reduce settings that increase the number of objects, simulations and draw calls the main thread must prepare.
- AI and vehicle traffic: airline traffic, airport vehicles, road traffic and boats can be expensive around large airports. Test with all traffic disabled before deciding that scenery is responsible.
- Autogen and scenery complexity: lower these one step at a time. Dense buildings and vegetation can constrain the CPU even when the graphics card appears lightly loaded.
- Scenery and object distance: excessive radius settings make Prepar3D prepare more terrain and objects. Our guide to controlling Prepar3D scenery draw distance explains which reductions preserve nearby detail.
- Complex aircraft and add-ons: detailed systems, gauges, airport modules and external utilities all consume processor time. Compare the problem flight with a default aircraft and default airport to isolate an add-on bottleneck.
- Shadows: these can load both the CPU and GPU because objects must be prepared and their shadow maps rendered. Reduce shadow-casting categories before removing shadows entirely.
Low total CPU usage does not rule out a CPU bottleneck. Prepar3D can be held back by one critical thread while the remaining cores are partly idle, and Windows may move that thread between logical processors. A faster per-core processor can therefore matter more than adding cores, though the benefit depends on the Prepar3D version and add-ons used.
Which settings improve GPU-bound performance?
For a GPU-bound system, reduce pixel-heavy effects before sacrificing scenery density or AI traffic.
- Display resolution and anti-aliasing: these are the cleanest diagnostic controls and usually the first settings to lower. Supersampling options are especially demanding where offered.
- Dynamic lighting and shadows: night lighting, numerous light sources and high-quality shadows can place a heavy load on the graphics card.
- Cloud and reflection quality: lower these when poor FPS occurs mainly in weather or around reflective surfaces.
- Texture resolution: reduce it when dedicated GPU memory is nearly exhausted or stutters appear while turning the view. If VRAM has ample headroom, lowering textures may save memory without producing much extra FPS.
For sensible starting values before fine-tuning, use our balanced Prepar3D graphics baselines. A broader Prepar3D performance-tuning sequence covers frame pacing, background software and the order in which to adjust settings.
Can stuttering occur without a CPU or GPU bottleneck?
Yes. Full VRAM, insufficient system memory, scenery loading, storage delays, unstable add-ons and thermal clock reductions can cause stutters even when neither utilisation graph is continuously full.
Watch for the pattern. Stutters only while panning suggest texture or VRAM pressure; pauses near one airport point towards scenery or an airport module; regular pauses across every flight suggest an add-on process or background task. Test with a default aircraft, default scenery and clear weather before making permanent global reductions.
Once uncapped performance is understood, restore VSync or set a sustainable frame target if it gives steadier frame pacing. A limiter does not increase maximum FPS, but a stable target often feels better than repeatedly swinging between CPU- and GPU-bound states.