Vertex/Geometry Shader Performance

There's more to the DirectX pipeline than just pure pixel shader performance though – here, we use a combination of geometry and vertex shader-bound tests to determine Cypress's performance in these areas.

Tests

• Direct3D 10: RightMark 3D 2.0: GS Galaxy
  
• Direct3D 10: RightMark 3D 2.0: GS Hyperlight
  
• Direct3D 10: 3DMark Vantage: GPU Cloth
  
• Direct3D 10: RightMark 3D 2.0: VTF Earth
  
• Direct3D 10: RightMark 3D 2.0: VTF Waves

Test Kit

• Intel Core i7 965 processor (3.2GHz: 133MHz x 24)
  
• Asus P6T V2 motherboard (Intel X58 Express)
  
• 6GB Corsair TR3X6G1333C9 DDR3 memory (3x 2GB, 1,333MHz, CL9)
  
• Corsair X128 SSD
  
• Corsair HX1000W PSU
  
• Windows Vista Home Premium x86-64 (with Service Pack 2)
  
• Antec Twelve Hundred Chassis
  
• HD 5870: Catalyst 8.66 RC6 Windows Vista/7 (September 11th Build)
  
• All other ATI: Catalyst 9.9 WHQL
  
• ] Nvidia: ForceWare 190.38 WHQL

Our geometry shader-targeted tests don't reveal anything out of the ordinary on the Radeon HD 5870 - GS Galaxy is typically CPU limited, while Hyperlight is a more accurate reflection of pure geometry shader throughput. 3DMark Vantage's own geometry shader test is something of a curve-ball though, as it clearly favours Nvidia hardware (we've disabled PhysX in the driver just in case the driver tries to do something funky, for what it's worth). The Radeon HD 5870 is on a par with the GTX 285, so we can expect GT300 to push the boat out in this particular test.



Dropping back into RightMark3D for the vertex texture fetch tests swings things back the other way - the Radeons dominate proceedings here. The complexity of the scenes favours the AMD hardware, which is able to better cope with the loads being placed on the GPU. We also ran the same tests at lower detail and found that the GeForces fared much better.