Synthetic benchmarks
Synthetic CPU tests tell us relatively little about the real-world performance of a processor. Even so, the near perfect scaling in both ALU and FPU performance across Tigerton, Clovertown and Kentsfield processors does indicate that the three almost certainly share the same under-lying cores, the only difference apparently being in how they are packaged.
Testing of memory bandwidth produced highly inconsistent results on the Tigerton system (not to mention took ages) because, we think, of the huge amount of memory installed. Because of this, we weren't able to produce reliable figures and have chosen to exclude the test from our analysis.
However, SiSoftware Sandra's Memory Latency tests did appear to be running correctly and showed the FB-DIMM-equipped systems to be at a significant disadvantage to the Kentsfield with its DDR2-833.
The Clarksboro 7300 chipset shows even higher latency than that of the 5000X, presumably due to the extra complexity in the chipset from having to manage four independent front-side busses.
