Crucial PC2-5300 DDR2-667 Ballistix Memory Review :: Overclocking

08-08-2005 · Category: Hardware - Memory

By Tulatin

Overclocking Memory was always an interesting art, especially when locked processors came into play. Without the ability to drop or raise a processor's multiplier, you are essentially at its mercy when overclocking, that is, unless you chose to utilize dividers. However, the issue to utilizing dividers is that no matter how fast the memory goes, its bandwidth will be forever restrained by the lacking FSB, whose latency comes from being forced to carry data between processor and chipset.


Crucial PC2-5300 DDR2-667 Ballistix Memory Review

Thus, don't be surprised when during performance tests, the overclocked memory modules yield nearly identical results to their stock counterparts. With this in mind, the only real point of running Overclocking tests on memory such as this is to show people who choose to push their Intel boxes to the very limits just how far the ram can go before a divider must be utilized. Quibbles aside, it's time to examine our Overclocking methodology. While most reviewers out there are stuck using dividers and fiddling with voltages for this and that we have a secret weapon. Our NVIDIA Reference motherboard allows for asynchronous memory and FSB clocks, through the use of a system of clever dynamic dividers. While it is true that the system is not flawless, and on occasion will be incapable of providing certain dividers, it is certainly convenient to be able to set DRAM clocks down to the MHz. Thus, with this ability in the bios, we simply tap the Ram's clocks up in wee, 5 MHz increments, utilizing 3 loops of Memtest x86+'s test 8 in between to test for stability. When instability is reached, we begin by attempting to fiddle with the FSB a little, trying to boost or drop it in 3MHz or so increments, so to rule out the possibility of an errant divider. Failing that, memory voltage is boosted, and if a gain cannot be purchased by these means, we resort to lowering timings.

While in past with DDR1 it was fruitful to lower them one at a time (and considering that DDR1 is terribly stringent on latencies), we've decided to simply run the gamut with DDR2 memory. With that in mind, CAS, RAS to CAS, RAS Precharge are each lowered by one, while tRAS is lowered by a total of five points. Again, we repeat the upward baby steps, utilizing Memtest x86+'s test 8 to find stability, finally settling on a windows session of S&M to get a final ideal of stability. Without any further Adieu, let us move onto the results of our experimentation. Beginning with the standard clock of 667MHz and timings of 4-4-4-10, as per specified by Crucial, we began by simply hopping the FSB up in the miniscule 5 MHz hops, reaching a barrier at 684 MHz. At this point, we attempted to raise V-DIMM, and with a .1V boost, we were awarded with a grand 16MHz jump - essentially nothing of interest. Now, one would expect the relaxing of latencies to have a positive effect on memory Overclocking, yet in our case it seemed to afford us no benefit - boosting them up to 5-5-5-15 (as per Crucial's conservative on module SPD rating), yielded absolutely no benefit whatsoever, resulting in our steadfast (and rock solid I may add) maximum of 700MHz. While it's nothing astronomical, it still manages to show that the memory has a bit of headroom with it, and should allow for a fair amount of FSB raising without the need to begin utilizing dividers. As to the images, remember that DDR2, much like Intel's FSB speeds is measured by multiplying the frequency at hand by 4, thus turning our 175 into 700.


Crucial PC2-5300 DDR2-667 Ballistix Memory Review
Crucial PC2-5300 DDR2-667 Ballistix Memory Review