ScienceMark2.0 & Pifast
Starting off, as always, with a look at bandwidth and latency. They're usually a decent indicator of probable performance.
The important two results to consider are the highest and lowest. Both are run with no PAT enhancement. Memory bandwidth results tally with increases in RAM MHz; a ~25% increase for 25% increase in RAM speed, even though latencies are much tighter for DDR400. Note how enabling PAT technology adds almost 10% to bandwidth. That, Intel will tell you, is the main performance difference between Canterwood and Springdale chipsets.
Latency results are also interesting and intriguing. PAT enabling shaves off around 8ns of latency at DDR400 speeds. Crucial Ballistix's higher latencies at default speeds make it look poor in comparison.
Translating latencies and bandwidth into a real-world application, we see that, when PAT is not enabled, higher bandwidth and latency is around 2.5% faster than the DDR400 2-2-2-5 setup. This is exactly why enthusiasts prefer to jack up their boards' FSB and memory speeds. Having PAT enabled at DDR400 provides another performance boost, one that outstrips the advantages of PC4000 memory and high FSBs.
