System Setup

NVIDIA G80 Performance Test Platform

  Hardware Component 
Graphics Hardware  NVIDIA GeForce 8800 GTX
NVIDIA GeForce 8800 GTS 640MiB
NVIDIA GeForce 7950 GX2
GeForce 7900 GTX
ATI Radeon X1950 XTX 
Processor  AMD Athlon 64 FX-62
2.8GHz, Socket AM2 
Mainboard  Foxconn C51XEM2AA
NVIDIA nForce 590 SLI 
Memory  SuperTalent DDR2
2GiB (2 x 1GiB), DDR2-800
4-4-3-8 2T 
Power Supply  PC Power and Cooling TurboCool
1000W, Quad-SLI 
Hard Disk  Western Digital WD2500KS 250GB
7200rpm, SATA2 

We used the Force 97.92 WHQL driver release for the GeForce 8800 boards, 93.71 WHQL for the GeForce 79xx hardware, and CATALYST 7.1 for the Radeon X1950 XTX. Performance testing was conducted in February 2007, prior to the release of 100-series ForceWare and CATALYST 7.2.

We used the following games for real-world performance testing:

  • Prey v1.2
  • F.E.A.R - in-game performance test, v1.08 retail
  • Company of Heroes - v1.4, in-game performance test, 16xAF forced via graphics CP
  • Call of Duty 2 v1.3
  • Quake 4 v1.3
  • Half Life 2: Episode 1 b2992
  • Half Life 2: Lost Coast b2992
  • Splinter Cell: Chaos Theory v1.05 SM3
  • Battlefield 2 v1.4 - 16xAF forced via graphics CP
  • GTR2 v1.1.0.0 - FRAPS recording
  • Far Cry v1.4
  • Pacific Fighters v4.04 - Rendering errors with VTF water enabled on 7-series GeForces

Notes

All benchmarks were run three times and the average result taken. We use an in-house tool to measure simple shader rates using HLSL, with its basic shader fragments available on reasonable request. For measuring pixel and textured fillrates we used a combination of Marko Dolenc's fillrate tool, D3D Rightmark and an in-house utility. Other applications like nDAW, bogobench and RoOoBo's BenchTest were used to verify fillrate measurements.

We start with game performances.