Systems setup and notes
| NVIDIA nForce 590 SLI Intel System | Intel i975X System | ATI CrossFire Xpress 3200 AMD System | NVIDIA nForce 590 SLI AMD System | |
|---|---|---|---|---|
| Processor | Intel Core 2 Duo Extreme X6800 (2.93GHz, 4MiB L2 cache, LGA775) | AMD Athlon 64 FX-62 (2.8GHz, 2MiB L2 cache, AM2) | ||
| Motherboard | NVIDIA nForce 590 Intel Edition - reference | ASUS P5W DH i975X | ATI CrossFire Xpress 3200 - reference | ASUS M2N32 SLI nForce 590 SLI AMD |
| Memory | 2GiB (2 x 1GiB) Corsair XMS2 PC8500 EPP | |||
| Memory timings | 4-4-4-12-2T @ 800MHz | |||
| Graphics card(s) | ASUS GeForce 7900 GTX 51MiB (SLI for Splinter Cell) | Sapphire Radeon X1900 XTX 512MiB (CrossFire for SC) | ASUS GeForce 7900 GTX 51MiB (SLI for Splinter Cell) | |
| Disk drive(s) | Seagate 160GB 7200.9 SATA 3Gbps | |||
| BIOS revision | 17/07/06 - 2.05342 | 30/06/06 - 0602 | 05/04/06 - 08.00.13 | 14/06/06 - 0504 |
| Mainboard software | NVIDIA 9.37 | Intel Inf 8.0.1.1002 | 5.10.1000.5 | NVIDIA 9.34 |
| Graphics driver | NVIDIA ForceWare 91.31 | ATI Catalyst 6.6 BETA | NVIDIA ForceWare 91.29 | Operating System | Windows XP Pro SP2 32-bit |
Testing software
We ran the NVIDIA nForce 590 SLI Intel Edition reference motherboard through the usual barrage of tests:
- ScienceMark Memory Bandwidth
- ScienceMark Memory Latency
- Pifast calculation to 10M places
- KribiBench v1.1
- HEXUS WAV encoding
- HEXUS DivX encoding
- Cinebench 2003
- HDTach 3.0.1.0
For 3D performance it was the intrepid trio of HEXUS custom benchmarks:
- Far Cry v1.33
- Quake 4 v1.04
- Splinter Cell: Chaos Theory v1.05
CPU speeds
As is often the case, one chipset can appear to be faster than another, yet the performance gain is induced by an overclocked FSB/HTT clock. That's why we explicitly list CPU speeds. You need to bear these in mind when comparing upcoming results.2938.0MHz - 267.1MHz FSB - ASUS P5W DH i975X - Intel Core 2 Extreme X6800 CPU
2933.3MHz - 266.7MHz FSB - NVIDIA nForce 590 SLI Intel Edition - Intel Core 2 Extreme X6800 CPU
2813.0MHz - 200.9MHz HTT - ATI CrossFire Xpress 3200 AM2 - AMD Athlon 64 FX-62 CPU
2812.8MHz - 200.9MHz HTT - ASUS M2N32 SLI nForce 590 SLI AMD Edition - AMD Athlon 64 FX-62 AM2 CPU
The ASUS i975X has a slight lead over the NVIDIA reference board. That's a little naughty, really, as it's the simplest method of making the motherboard faster in benchmarks.
Notes
We're comparing the performance of the NVIDIA reference board against a class-leading, Core 2 Duo-supporting i975X motherboard from ASUS. The latter supports CrossFire, ATI's multi-GPU technology, so we've run it with a single Sapphire Radeon X1900 XTX in our 2D benchmarks and gaming apps. We've then installed a master card, enabled CrossFire, and benchmarked Splinter Cell: Chaos Theory with resolution set to 1920x1200 and 4x AA and 8x AF.The same has been done on a reference ATI CrossFire Xpress 3200 AM2 motherboard. The two NVIDIA boards in our lineup, nForce 590 SLI I.E. and nForce 590 SLI AMD, are both tested with similar methodology. This time, GeForce 7900 GTX 512MiB card(s) is/are used.
What this kind of testing will show is how much performance benefit Intel Core 2 Extreme X6800 provides over AMD Athlon 64 FX-62 AM2, with all other factors largely equal. Secondly, it also shows GeForce 7900 GTX (and SLI) and Radeon X1900 XTX performance on both CPUs, with both single and dual cards taken into account.
You may be wondering why our NVIDIA test setups don't use EPP memory. The answer lies with the inability of our Intel Edition board to run a 2GiB pack of Corsair EPP-supporting memory. When set to SLI-Ready memory and, say, CPU OC 0% the board would fail to POST and overclocking protection would kick in. We note that the board was received at the very last moment and there simply wasn't enough time for NVIDIA to ship out another reference model, one that, hopefully, would be more partial to EPP lovin'. That's why, in a roundabout way, all motherboards were tested with 2GiB system RAM set to DDR2-800MHz with 4-4-4-12 timings. We'll be looking again at nForce 590 SLI I.E. EPP support with retail examples that will follow in a few weeks.
The reference board, it must be noted, also failed to complete our regular motherboard torture/stability test. It would hang/crash after an hour of sustained GPU/CPU load. It was stable enough to deliver a clean set of benchmarks that were consistently repeatable, though. Let's head on over to them now.