Testing Method & Test Setup
Knowing about D9PFJ’s capabilities from recent Ballistix reviews, we armed ourselves with an Ivy Bridge testing platform that should allow our memory to show every last bit of its overclocking potential.
To make sure that our figures represent the sort of stability safe to use
ever day, we are going to run each setting until we get a 150% pass of eight
375MB instances of HCI Memtest that is considered one of the toughest memory
||ASUS Maximus V Gene (BIOS 0086)
||Intel Core i7-3770K
||XFX 8600 GT
||Kingston HyperX Lovo KHX1600C9D3LK2/4GX
||Samsung 40 GB
||Windows 7, 64 bit SP1
But immediately, we encountered a couple of issues.
First, we found out that Kingston’s SPD sets the subtimings too tight, making
frequencies in excess of 800MHz unreachable without manual adjustments. On the
screenshot below we demonstrate which values we had to loosen when testing the
overclocking capabilities of our kit.
thing, we’ve noticed is that unlike all D9PFJ we tested previously, our LoVos
fell flat on their face using even primary timings. Raising tRCD by one value
above tCL and tRP has noticeably improved the situation.
As you can see in the results chart, maximum stable frequency is a function of voltage only up to 1.65-1.70V, past which more juice only makes things worse. Knowing that Micron ICs are able to digest a lot more voltage, we flashed them with an SPD from 2GB Ballistix sticks to eliminate the sub-sub-settings as a culprit for such phenomena. Unfortunately, SPD didn’t make any difference, so we reckon that limited voltage scaling is caused by imperfect PCB design.
Either way, even from results leading to 1.65V we see that D9PFJ samples used on our LoVo are nowhere near the quality of higher-end offerings, which indicates that there is serious binning going behind the curtain. For example, in order to operate our kit as DDR3-2133 one must use a very loose timing set of 11-12-11-30 that will render the frequency gain useless. Therefore, it’s safe to say that the LoVo’s are not a well-clocking memory even despite the ICs and low rated voltage that initially promised a lot more than we got.
Discuss this article in the