With the latest addition of dual channel DDR and CPUs with high FSB, demand for extremely high speed RAM is rising even faster. Or, in order to be more exact, the demand for RAM with extreme overclocking capability is. What I’m going to try to do is get as many memory modules as I can from the high speed market, 2 of each for use in dual channel operation and will measure their maximum stable speed under certain timings and voltage settings.
The memory market is growing strong the latest years. Probably
because the demand from overclockers is rising with incredible speed lately, so
the companies are racing to cover as much of it as they can. This leads
companies to make RAM far beyond JEDEC approved standards, therefore much faster
and/or with the ability to hold much lower timings and/or much higher voltages.
With the latest addition of dual channel DDR and CPUs with high FSB, demand for
extremely high speed RAM is rising even faster. Or, in order to be more exact,
the demand for RAM with extreme overclocking capability is. I will try to cover
as many modules as I can get my hands on and have time to test, plus perhaps try
to update every now and then.
What I’m going to try to do is get as many memory modules as I
can from the high speed market, 2 of each for use in dual channel operation and
will measure their maximum stable speed under certain timings and voltage
settings. Also, I’ll show the abilities of the worst modules out of the ones
I’ll test for each candidate if I have more than 2, trying to minimize the ‘good
luck’ part as much as possible. However, take note that not all modules are as
consistent. The very same modules of even the same production line may overclock
much differently, although not by much most of the time, however I’ve seen even
a 35 MHz difference between two modules. It is draw of luck, like with almost
every other component. Also note that many of these modules were brand new. Most
modules like a small burn-in period to get used at high voltages or speeds. Most
will perform enough better after a week or two of use that will cause them
pressure. Do NOT take these numbers as guaranteed, even if you get the same
modules they might do much better or much worse. Also note the size of the
modules, usually 512mb modules tend to overclock worse than 256mb modules for a
number of reasons, however that doesn’t always hold true.
The testing procedure will include testing at different timing
settings over every voltage setting allowed from the motherboard (2.6v-3.3v in
0.1v increments). Being as exact as I can, the exact multimeter readings from
the mosfet under load are these :
Setting / Real
2.6v = 2.64v
2.7v = 2.75v
2.8v = 2.87v
2.9v = 2.96v
3.0v = 3.08v
3.1v = 3.15v
3.2v = 3.27v
3.3v = 3.38v
Since when any module of ram will perform about the same when it is at the same
speed and same timings if it’s stable, given that compatibility is not an issue,
I will spare comparisons and will only note their overclocking ability.
Stability where tested by running one hour of memtest and then 3 hours of
Prime95 and SuperPi.
Take note that most companies do not approve of overclocking. Overclocking
beyond the manufacturer recommended settings and/or voltages will void your
warranty. So make sure not to push it too much, memory modules are hard to
damage but it’s not something impossible to happen.
Furthermore, ‘burn-in’ is a method that may increase performance slightly on
most modules. However that is possible only if the module is brand new. If the
modules are used even for a few days, it won’t help at all. Burn-in happens when
you get a module and find the maximum speed at the voltage you intend to give it
and then stress the modules (a lot) for a few days at this speed. There is a
chance that the modules will be able to hold higher speeds and/or hold the speed
with less voltage without issues. There is also a chance that it may not help at