Comparison of Graphics Cards with NVIDIA's RIVA TNT Chip

Test Setup

To make sure that only the hardware is influencing the benchmark, I was benchmarking all cards with NVIDIA's reference driver 0048. Then I ran each card with its own driver. It was surprising to find out that there was no difference between the reference driver result and the driver provided by the card maker. Some card makers want us to believe they are tweaking their drivers, but my test showed that this is pure hype. The performance of the reference driver was identical to the generic driver in all cases.

The benchmarks were run on an Asus P2B-LS BX-motherboard with Intel Pentium II 400 CPU, 128 MB SDRAM and an IBM DGVS 09U UW-SCSI hard drive. Operating system was Windows 98, the screen resolution for both games was 1024x768, both at 16 bit color depth, the refresh rate was 85 Hz. Each card ran with 90 MHz chip clock and 110 MHz memory clock, the default clock recommended by NVIDIA. The Quake 2 version used was 3.20.

There wasn't much of a performance difference expected between the six cards, only Elsa's Erazor II should perform a bit better than the others, as it's the only card using the more sophisticated SGRAM. The benchmarks showed a different result. When running each card at the default 90 MHz chip and 110 MHz memory clock, the only thing that made a difference was the memory timing. The SGRAM of the Erazor II seems to make it slower than almost all of the others.