Overclocking and Optimizing:
Just to get to the rated speed of the Kingston HyperX 2GB PC3-11000 DDR3 Dual Channel Memory Kit required a bit overclocking, but I had to see just how high it would go. In addition, I spent some time seeing if the timings could be improved while running at 1375MHz.
I had read a handful of other reviews on this memory kit trying to gauge the limits I might be able to achieve. While some reviews never even reached the rated speed (due to other system limitations), others were able to get about another 50-100 MHz out of it. I spent several hours trying to work the bus speed up ever higher, but I actually found the ceiling for this test setup after just a few minutes. Despite my best efforts to find a perfect balance between voltage, timings, frequencies, and multipliers, I could not achieve any stable results above 1400MHz. No matter how relaxed the timings were, nothing proved to be any better than running at 1400MHz and 7-7-7-23 timings.
A 24MHz overclock doesn't seem like much, but considering this is effectively a 1066MHz kit custom configured for faster speeds, I was satisfied. Next it was time to see if I could improve upon the timings while running at the rated speed of 1375MHz. While 7-7-7-20 aren't terrible specifications for memory of this speed, but knowing that Kingston offers a similar kit rated at 5-7-5-15 had me interested in seeing how low I could go. After about a half hour of trial and error I resigned myself to the fact that 6-7-7-20 was the best I was going to do. The results were rather binary, as the system either booted or it didn't, and there was no need to pursue any grey area around system stability. While the system passed stress tests with flying colors at 6-7-7-20, I didn't even get the chance to try at anything tighter.
The test system listed on the previous page makes for an excellent test bed for DDR3, as it allows for DDR3 to be compared directly with DDR2 on the same platform. The ASUS P5KC motherboard allows for both memory formats to be used (not at the same time), providing a crystal clear look at whether the performance would justify the cost to upgrade.
While people with fast (think 1066MHz and faster), low latency DDR2 kits may be best suited to wait for DDR3 to really get "up to speed", those with PC2-6400 (800MHz) kits might have more to gain from making the transition now. For this review, a 2GB PC2-6400 kit from OCZ Technology with moderate overclocking capabilities will be tested head-to-head with the PC3-11000 kit from Kingston.
Testing will consist of three benchmarks used to gauge the performance of five different memory configurations. The benchmarks executed include tests from these three packages:
» Everest Ultimate Edition 2006
» Sisoftware Sandra Lite 2008.1.12.34
» Performance Test v6.1
The memory configurations include the Kingston PC3-11000 running at PC3-10666 speed (1333MHz), at roughly its rated speed (1376MHz), and at its maximum stable overclock (1400MHz). The OCZ Technology DDR2 kit was run at its rated speed (800 MHz), and at its maximum stable overclock (984MHz):
» Kingston HyperX DDR3 at 1333MHz, 7-7-7-20, 1.7V
» Kingston HyperX DDR3 at 1376MHz, 7-7-7-20, 1.7V
» Kingston HyperX DDR3 at 1400MHz, 7-7-7-23, 1.8V
» OCZ Technology System Elite DDR2 at 800MHz, 4-4-4-13, 2.0V
» OCZ Technology System Elite DDR2 at 984MHz, 5-5-5-15, 2.1V
From the list of configurations we can see that the DDR2 memory can't run nearly as fast, but can offer much tighter timings. Based on some of the initial things I have seen in similar tests, I was very interested to see how close of a contest this wound up being!