Test results
I could go on at length here in an attempt to add to the suspense, but you can already see that everything boils down to a single graph, so let's just get started.

First of all, I should point out the obvious: Don't take the results above as any sort of statement about the quality of the memory tested. The only measurement here is how well each type of memory reacts to (in some cases silly amounts of) overvolting.

Starting at the bottom, the Kingston RAM didn't seem to like either the relaxed timings or the higher voltage at all. The speed here is listed at 199, but all that's really signifying is that the HyperX wasn't stable even at stock DDR400 speeds when its voltage was raised. I suspect that the HyperX is simply designed with really tight timings at stock voltage in mind, and this sort of thing just isn't its bag, baby.

Next up is the Mushkin. It doesn't benefit very much from additional voltage, but at least it wasn't hampered by it. Pushing the voltage all the way up to 3.3V did result in a significant increase in top speed.

The Corsair RAM started out with a reasonably good overclock and responded pretty well to the additional voltage, as well, especially at 3.3V, where it topped the 2.8V speed by 23MHz.

The Crucial Ballistix did very well for itself here, hitting a top memory clock of 244MHz at stock voltage and extending that all the way to 261MHz at 3.3V.

OCZ said that their Gold Edition RAM responded well to overvolting, and they weren't kidding, but it also managed to nab second place for the stock voltage tests. It also posted the largest gains with increased voltage, including a first place finish (by a nose, or a MHz anyway) in the 3.3V category.

Since the OCZ RAM was the top finisher, it got the, umm.... honor of being subjected to 3.7V. The added voltage allowed it to up its high score by 5MHz, but I suspect there's more to the story than that. Throughout my stability testing, all of the RAM failed in a very consistent way, either crashing Sphinx or causing it to throw an error message, such as a divide by zero error. However, when I was testing the upper limits of the OCZ RAM at 3.7V, the failures were much different, always resulting in a hard lockup.

Based on these differences in behavior, I suspect the limiting factor here may be the motherboard and not the memory. I tried everything I could think of to make the motherboard more stable, from lowering HyperTransport multipliers to bumping up north bridge and CPU voltages, but the result was always the same. Perhaps with a different motherboard, the OCZ Gold Edition-DDR Booster combination could hit even greater heights.

On a more general note, you'll notice that in no case did the 2.8V DDR Booster test configuration allow for higher stable speeds than the 2.8V motherboard configuration. Either the Gigabyte motherboard used in this tests provides cleaner power than other boards, or the advantage of the DDR Booster is more its upper voltage range than the "cleanliness" of the power it delivers.