Rendering performance benchmarks comparing Leadwerks 5 Beta, Leadwerks, and Unity.
Four tests were created to isolate and measure different aspects of the graphics pipeline in each engine.
The original tests are in the "Leadwerks 5" subdirectories. Updated builds are also provided with improved hardware compatibility.
This measures the performance of object culling speed in a worst-case scenario where frustum culling removes no objects. 32,768 instanced boxes are shown.
This demonstrates the efficiency of the animation pipeline. Each of the 1024 characters has a unique skeleton and a relatively low polygon count, because the vertex shader speed is probably going to be similar in each engine. This is meant to test the efficiency of skeleton data transfer, and determine if animation introduces any overhead to the renderer.
This measures the performance of a large number of lights against a large number of identical objects. I was surprised to find that Unity redraws every shadow map every single frame, instead of only updating shadows when an object moves. 25 point lights are displayed in a scene made of 801 boxes.
This test is the most representative of an actual game level, with many different unique objects. It measures the overhead of draw calls and the efficiency of the mesh batching system. 4096 unique boxes are shown.
- Windows 10 or 11
- Discrete GPU with support for Vulkan 1.3 (Integrated graphics are not yet supported)
- Visual Studio Redistributable (You probably already have this installed)
- The Leadwerks 4 builds require OpenAL to be installed
Note that AMD cards will display some visual artifacts in the current build.
TechPowerUp GPU-Z can be used to measure GPU utilization. Windows task manager does not correctly measure this metric.
Built-in FPS counters are provided. FRAPS can also be used to measure the framerate, but is reportedly not displaying on some Vulkan drivers. Some other FPS counter overlays such as OpenGPU OCAT will reduce performance significantly enough to alter your readings, by a lot, and are not suitable for measuring high-frequency framerates.
In each test, Leadwerks 5 outperforms Leadwerks 4, Unity, or both by a factor greater than 10x. GPU utilization is consistently higher with Leadwerks 5. All measurements were recorded using a PC with a quad-core Intel Core i7-7700K CPU @4.20 GHz and an Nvidia GeForce 1080 GTX GPU with driver 471.41 installed, on Windows 10.
| Engine | GPU Utilization | Framerate |
|---|---|---|
| Leadwerks 4 | 8% | 52 |
| Unity (DX11) | 4% | 40 |
| Unity (Vulkan) | 4% | 40 |
| Leadwerks 5 | 95% | 1206 |
| Engine | GPU Utilization | Framerate |
|---|---|---|
| Leadwerks 4 | 1% | 5 |
| Unity (DX11) | 45% | 62 |
| Unity (Vulkan) | 45% | 64 |
| Leadwerks 5 | 96% | 1179 |
| Engine | GPU Utilization | Framerate |
|---|---|---|
| Leadwerks 4 | 49% | 704 |
| Unity (DX11) | 33% | 90 |
| Unity (Vulkan) | 30% | 74 |
| Leadwerks 5 | 96% | 1456 |
| Engine | GPU Utilization | Framerate |
|---|---|---|
| Leadwerks 4 | 10% | 49 |
| Unity (DX11) | 5% | 2 |
| Unity (Vulkan) | 4% | 2 |
| Leadwerks 5 | 63% | 6213 |