In the CPU and FPU tests of WinBench 99, the L2 cache latency of 14 was consistently the fastest while the cache latency of 13 was the slowest. A pattern's quite apparent in those two tests. The performance of the L2 cache latencies of 1 to 9 were all scattered around the average. The scores began to deteriorate from latency of 10 to 12 and finally culminating with the cache latency of 13 garnering the lowest score. Then, the score suddenly jumped to right to the top with the cache latency of 14. The cache latency of 15 was usually slightly slower than the cache latency of 14 but remained above the average.

In both tests, there are no differences between the performance of the cache latencies of 1 and 5 (the test processor's default setting) because they both scored equally well. The cache latency of 15, which is theoretically the slowest, ended up either just behind the leader (in the CPU test) or on the same level (in the FPU test). So, this goes to show that lower latency does not always correlate with higher performance.

In the 3D WinBench 99 tests, the pattern is quite similar to that of the WinBench 99 tests. The cache latencies of 11, 12 and 13 were once again among the slowest and there's the same big leap in performance from the cache latency of 13 to the latency of 14. Also, like in WinBench 99, the cache latencies of 2 and 8 seem to do poorer than their neighbours though in the 3D Transformation test, they both did well enough to cross the average line.

The cache latency of 13 was once again the slowest, in the 3D Lighting and Transformation test; and the second slowest in the 3D Transformation test. Both cache latencies of 1 and 5 (processor default) performed equally well in the 3D Lighting and Transformation test but the cache latency of 5 surged away and grabbed the top spot in the 3D Transformation test. The theoretical underdog, the cache latency of 15 did quite well in both tests, staying above the average line. Again, the tests show that the lower latency doesn't mean higher performance. Even the processor default latency of 5 was faster than the latency of 1.

The overall 3DMark 99 Max result showed the superiority of the L2 cache latency of 14 over the others. And although the cache latency of 13 wasn't the poorest performer, it was close enough at the second last place. The cache latency of 1 performed poorly here with below average results while the cache latency of 15 was above average overall. The processor default latency of 5 did moderately well by hovering around the middle.

Things were quite different with the 3DMark 99 Max's CPU test. Here, the cache latency of 14 only managed to hang around the mid-pack and the cache latency of 4 was the fastest. The cache latency of 1 was still an average performer although the cache latency of 15 now slipped below average. As usual, the latency of 13 did very badly, now at the third last position. The processor default latency of 5 did quite well here by bagging the fifth place.

In Game 1, the cache latency of 14 was once again the champion with the latency of 13 hanging close to the bottom. This time however, the supposed speed champion, the cache latency of 1 ended up even slower than the latency of 13! The latencies of 5 and 15 were about average although the latency of 15 was slightly faster.

In Game 2, the cache latency of 4 was the fastest with the latency of 14 just behind. The same was true for the latency of 13 which was just the last place at the bottom to the latency of 12. Again, the cache latency of 1 scored badly, hitting the fifth lowest place! The latency of 15 was much better than it and even the cache latency of 5 scored higher.