I know it's early to say, but from the preview X38 is way underwhelming, performance-wise. Also it made me very skeptical of this higher bandwidth/higer latency tactic that Intel's been pulling out. Other than extra features added (lots of them are marketing-driven), what Intel has been doing (975X -> P965 -> P35) is increasing bandwidth at the cost of latency. Out of curiosity, I dug up some of the old articles here @AT.
955X Article (You could get an idea of 925X)
975X Article (Shows 955X data also)
Finally X38 Preview (No comparisons vs 975X but one can get an idea)
Note how performance has changed from 925X -> 955X -> 975X -> (X38). The only meaningful boost looks to have happened from 955X -> 975X, but then again the 975X platform in that review employs a ridiculously aggressive memory configuration (DDR2-667/3-2-2-6) using FatBodies. Without the exact specification given WRT the 955X testbed in that review, it's hard to judge just how much 975X improved upon 955X.
I would love to see a 925X/955X/975X/X38 comparison with the same CPU (something like Pentium Extreme Edition 965?) and same memory configuration. I bet the memory controller performance hasn't changed much for the past 3~4 years. What's improved is the obvious manufacturing tech which lets the revised chipsets clock higher, and memory itself which has little to do with chipsets.
At least this time around the higher bandwidth might actually benefit quad-core CPUs
when overclocking, but if one were to use a dual-core, or a quad-core at its stock speed, I think you might as well get a decent quality 975X board (can be had for like $100~150 if you look around) and be done with it. And of course if you already own any Intel chipset board that works without trouble, there is zero reason to upgrade. (I am strictly speaking in terms of performance here) I just don't see anything X38 has going for it other than more modern feature set. Add the price and power consumption factors, and the situation gets even worse.
P.S. Oh I almost forgot one thing: At least I should give Intel some credit for finally fixing the broken dividers with P35 memory controller. I myself don't have much experience with P35, but according to various reports the memory dividers are working a lot better than previous Intel chipsets.