Better Price, Performance, and Value? OCZ Titanium CL3 DDR2-800 Review- Page 3/8
December 14, 2007
By
Thomas Soderstrom
Timings and Overclocking
We retested all DDR2 modules using OCZ's "warranty limit" voltage of 2.30V, reaching the following "best" timings for the 800MHz, 1066MHz and 1200MHz speeds (if possible), before finding the "Maximum Overclock" for each set at standard 5-5-5-15 timings.
Best Stable Data Rate and Timings
Device
800MHz
1066MHz
1200MHz
Max OC
OCZ Titanium CL3 PC2-6400
3-4-4-12
4-5-5-15
Failed
1130MHz
Kingston HyperX PC2-6400 UL
3-3-3-6
4-4-4-8
Failed
1094MHz
Crucial Ballistix PC2-6400
3-3-3-6
4-4-4-8
5-5-5-15
1252MHz
The Crucial Ballistix PC2-6400 is the real spoiler here. It's unfortunate for most memory brands that these low-cost DIMM's have reached our lab because otherwise the competition would be much closer. We even replaced the original set with a second pair - which should rule out any "ringers" the company might have sent - and yet again these hit the same impressive results achieved by the first set.
On the other hand, local availability may rule out the use of a particular brand, and we still need to consider the actual performance impact these timings bring. Further, the CAS Latency, RAS to CAS Delay, Row Precharge, and RAS represented in the four familiar number don't represent all of the timings a module can set, so performance testing might not always favor the modules that appear to be "quickest".
Our "Max Overclock" setting required overclocking the CPU FSB, so we use it to represent the higher speeds overclockers could achieve if memory speed was the sole factor limiting their system's overclocking capability. To simulate this overclocking limitation issue on a chipset that doesn't support so high an FSB, we used the 5:4 (DRAM:FSB) ratio rather than the board's lowest 1:1 ratio.
Let's consider what the SiSoft Sandra memory benchmark reveals for the tested speeds listed in the chart above.
Sandra shows the "quickest" modules with a slight lead in raw memory performance, but let's see what effect this has on other benchmarks