Static geometry, unlike dynamic geometry, will directly benefit from hardware transformation. NVIDIA's TreeMark is a synthetic benchmark of transformation and lighting (henceforth referred to as T & L) based on static geometry. It uses the OpenGL pipeline to accelerate a procedurally generated tree with adjustable polygon complexity, between 700 and 140000 polygons per frame.
Static Geometry: the tree used in TREEMARK
Wire-Frame (left) and Textured (right)
To test polygon throughput with TREEMARK, GeForce (120/183) and TNT2 (150/183) were put through their paces on similarly configured 300 MHz and 450 MHz CeleronA systems.

One dynamic point light, 700 to 140000 polygons per frame, 640x480x16
The graph shows polygon throughput versus polygons per frame. The peak polygon throughput, as derived from the plateau of the curves, is approximately:
| CPU |
3D Accelerator M |
aximum Polygon Throughput |
| 300 MHz CeleronA |
GeForce
|
2.2 million polygons per second |
| 450 MHz CeleronA |
GeForce
|
3.3 million polygons per second |
| 450 MHz CeleronA |
TNT2
|
0.5 million polygons per second |