GeForce256 and the First T&L-Title

CPU-Scaling

I included all the games we used for benchmarking GeForce into this chart. The question is if GeForce with its T&L-engine is still depending on CPU-performance. We saw yesterday, that GeForce scaled rather similar to TNT2-Ultra in today's games except for NVIDIA's TreeMark. TreeMark was so far the only application that explicitly used the T&L-engine and it was impressive to see that GeForce reached the same scores regardless if running on a Celeron 400 or a Pentium III 550. We were excited to see if we would get the same results with Dagoth Moor Zoological Gardens. The benchmarks were run at 640x480x16 to avoid any impact of the rendering-pipeline or memory bandwidth, then the Celeron 400 scores were compared to the Pentium III 550 scores.

Before you look at the chart I'd like to remind you how it works. It compares the CPU-scaling between Celeron 400 and Pentium III 550 of GeForce w/DDR and TNT2 Ultra. The percentages you see reflect the Celeron 400 scores relative to the Pentium III 550 scores for GeForce and TNT2. A score of 100% means that the Celeron 400 score was a 100% of the Pentium III 550 score with the according 3D-chip, the higher the score the lower the scaling. Thus we expect GeForce to score higher percentages than TNT2-Ultra, especially in games that use its T&L-engine.

GeForce's Celeron 400-score in Dagoth Moor Zoological Gardens is with 79% way ahead of TNT2's 57%, but it still relies a lot on the CPU. We will have to see if this improves with better drivers for GeForce. The difference between the TNT2 and the GeForce-score shows however, that the T&L-engine makes indeed a serious difference in games that are using high polygon counts.