As we saw previously in the numerous performance comparisons, enabling the Video RAM Cacheable feature improved the performance of the graphics card. Many of the synthetic benchmarks that test the graphics card confirmed that. But they also confirmed that the boost gained was very minimal, usually in the region of 0.2%. So, is this minor boost worth the price?

Nothing could be determined using WinBench 99 or even 3D WinBench 99 because all their processor tests do not involve the graphics card at all. We managed to get a hint of the loss in processor performance using the synthetic benchmark of 3DMark 99 Max. But the true indicator of the performance loss was obtained using the simple yet useful Direct3D benchmark from Microsoft. The loss, as determined by the benchmark was 1.1% - the most significant change recorded in all the benchmark results.

Since individual tests have now shown that the boost in graphics performance is minimal while the loss in processor performance is much more significant, it remains for us to see how the real world performance is really affected. For that, we turn to Winstone 99 and games like Expendable (multi-texturing) and Turok (single-texturing).

In Winstone 99 and the two games, enabling the feature was shown to decrease performance. Although the drop in performance is quite small, it's the fact that the scores recorded were consistently lower when the feature was enabled that sinks that stake in! Enabling the Video RAM Cacheable feature definitely reduces overall system performance.

So, it's my humble opinion that, in the interest of optimal system performance, we should put aside this BIOS feature until the day comes when the bandwidth gap between the video RAM and system memory decreases or when L2 cache size becomes much larger. Until that day comes, video RAM caching will never amount to anything worthwhile.