Testbed and Methods
The graphics card was benchmarked on an open testbed that consisted of the following components:
- Processor: AMD Sempron 3000+ (Palermo), Socket 939
- Mainboard: ASRock 939Dual-SATA2 (ULi1695+ULi1567)
- Cooling: liquid cooling system with three water-blocks, a Eheim 1048 pump, and an Airplex EVO 360 radiator cooled with a 120mm Foxconn fan
- Power supply: be quiet! Dark Power 600W ATX 2.2
- Hard disk drive: 120GB ATA-100 Hitachi (7200rpm)
- Optical drives: TSST SH-D162C DVD-ROM and NEC ND-3520A DVD-RW
- Operating system: Windows XP Professional SP1
The CPU worked at a frequency of 2800MHz (350x8) with 1.55V voltage. The memory worked as DDR-133; its resulting frequency was 233MHz with 2-2-2-5-1T timings and 3.25V voltage.
The testbed ran Windows XP SP1. The card was overclocked with PowerStrip 3.62; the frequencies and temperature were monitored with RivaTuner 2.0 RC16. The GPU was cooled by the water-block. A 120mm fan was additionally installed to blow at the reverse side of the PCB:
I didn’t run a full cycle of tests since we recently benchmarked a GeForce 7600 GT. The Palit card behaves exactly like the EVGA e-GeForce 7600 GT at the same frequencies (for details see this article). So, there is only one benchmark today, 3DMark03 Pro 3.6.0. Its results will show you the effect from overclocking the reviewed graphics card with and without volt-modding.
The benchmark’s settings were left default. The graphics driver settings were selected as to provide an optimal quality/performance ratio.
The performance is high, just like the overclocking gain. By the way, I didn’t squeeze the maximum out of my sample of the card. There was still a considerable safety margin. Besides, it is possible to climb up to an even higher performance level by increasing the frequency of two out of the GPU’s three main subunits (you can select a negative delta for the geometry subunit by modifying the graphics card’s BIOS). I didn’t plan to do that for this review and didn’t check this possibility out in practice.