System setup and notes
Hardware
| Graphics cards | Inno3D GeForce 8800 GT 512MiB Overclock Edition | MSI NX8800GT-T2D512E-OC 512MiB | ASUS GeForce 8800 GTS 320MiB | ATI Radeon HD 3870 512MiB |
|---|---|---|---|---|
| Current pricing, including VAT | £185 | £170 | £180 | £150 |
| Shader Model | 4.0 | 4.1 | ||
| Stream processors | 112 | 96 | 320 | |
| GPU clock speed (MHz) | 702 | 660 | 513 | 777 |
| Shader clock speed (MHz) | 1674 | 1650 | 1188 | 777 |
| Memory clock speed (MHz) | 2000 | 1900 | 1584 | 2252 |
| Memory bus width (Bits) | 256 | 320 | 256 | |
| CPU | Intel Core 2 Extreme QX6850 LGA775 (3.0GHz, 8MiB L2 cache, quad-core) | |||
| Motherboard | eVGA NF68 (nForce 680i SLI) | ASUS P5E3 Deluxe WiFi-AP (Bearlake X38) | ||
| Motherboard BIOS | P31 | 0504 | ||
| Mainboard software | NVIDIA device driver 15.08 | Intel Inf 8.4.0.1016 | ||
| Memory | 4GiB (4 x 1GiB) DDR2-1066 | 4GiB (4 x 1GiB) DDR3-1066 | ||
| Memory timings and speed | 5-5-5-15 2T @ DDR2-1066 | 7-7-7-20 2T @ DDR3-1066 | ||
| PSU | FSP Epsilon 600W | |||
| Monitor | Dell 30in 3007WFP - 2,560x1,600 | |||
| Disk drive(s) | Seagate 160GiB SATAII (ST3160812AS) | |||
| Graphics driver | NVIDIA ForceWare 169.04 | CATALYST 7.11 BETA | ||
| Operating system | Windows Vista Business, 64-bit | |||
Software
| 3D Benchmarks |
Company Of Heroes: Opposing Fronts v2.103: DX9 and DX10 Enemy Territory: Quake Wars v1.2 (demo_00010.dem, map Valley) Crysis SP demo 64-bit GPU benchmark Lost Planet: Extreme Condition v1.004 built-in benchmark: DX10 |
|---|
Notes
The Inno3D card, once flashed with the supplied BIOS, as per our sample, runs with faster clocks than the also-overclocked MSI GeForce 8800 GT 512MiB, so we expect to benchmark a few per cent faster as a result. The interesting comparisons will be against the GeForce 8800 GTS 320 - priced at around the same level as the Inno3D card but sporting a smaller frame buffer - and the cheaper Radeon HD 3870, equipped with 512MiB of RAM and Radeon HD 2900 XT-matching performanceOur transition to Microsoft Vista Business 64-bit hasn't been without its share of problems. The most pressing issue is related to obtaining reproducible numbers on a consistent basis. We have seen that Vista performance is inherently more variable than that of Windows XP, such that the deviation between sets of runs on the same card has been as high as 15 per cent. Running practically countless iterations has helped eliminate the majority of the variance.
We've run our four games with the in-game settings fixed at high/very high and at suitably card-taxing resolutions, with antialiasing/anisotropic filtering applied. We'll note any testing anomalies during benchmark commentary, of course.
