The first runs set a baseline with all the timings and voltages set to Auto. This set the memory speed to 1066MHz but gave a small advantage to the OCZ DDR3-1333 over the Kingston as the OCZ memory had lower latency and increased bandwidth. This suggests that the BIOS was running the OCZ on tighter timings than the Kingston modules but the difference in PCMark05 was tiny. We've found that the Overall score in PCMark05 has a margin of error of plus or minus 100 marks so the two results of 7,424 and 7,603 are relatively identical.
Both sets of memory were happy to run at their maximum rated speed of 1333MHz on the standard 1.5V memory voltage and this time we forced the latency timings to the same 7-7-7-20 with the result that memory performance was identical while the Overall score in PCMark05 had some variance.
The next step was to feed some extra voltage to the memory and Northbridge/Memory Controller Hub. We started with 1.7V to the memory which is the rated voltage however we couldn't tighten the Command Rate to 1T which is the figure shown in the CPU-Z screen grab that Kingston had supplied so we used their suggested figure of 1.9V and the memory responded admirably. The problem is that the difference in performance was absolutely miniscule and there is no way that we'd recommend you use a voltage setting that is higher than Kingston's warranted figure to get an extra one percent out of your memory.
We left the voltages at 1.9V for the memory and 1.7V for the Northbridge and raised the front side bus from 333MHz to 440MHz to give a CPU clock speed of 3.52GHz. Both the Kingston and OCZ were happy to run at 1400MHz on a memory multiplier of 1.6x. The performance of the system shot up as a result of the increased processor speed and once again there was nothing to choose between the two types of memory. For the final test run we tightened the latency timings from 9-9-9-25 to 7-7-7-20 which gave a theoretical increase in memory bandwidth but you'd be hard-pressed to tell it from the performance tests.
Actually that wasn't quite the final test as we gave the OCZ Titanium DDR3-1600 a run on the same BIOS settings with a memory multiplier of 2.0x to give a colossal clock speed of 1760MHz. This was faster than we managed during our original review of the OCZ Titanium and it duly returned the highest test results of the day.
There are a number of morals to this tale:
It has clearly taken Asus a few BIOS revisions to get to grips with the P35 chipset and DDR3 memory and the result is better overclocking for both the CPU and system memory. We cannot separate the Kingston DDR3-1333 and OCZ Platinum DDR3-1333, which is exactly how it ought to be. Both types of memory have the same specification and the Asus P5K3 Deluxe works equally well with both of them.
Perhaps the most important point, though, is there are limited gains to be made from super-high DDR3 clock speeds compared to much cheaper DDR2. The only compelling reason to go for the more expensive stuff right now is the old 'future-proofing' argument because eventually all CPUs and motherboards will be using this new type.
Kingston's KHX11000 D3LL more than makes up for its troublesome performance first time round and delivers the goods with aplomb. It's still expensive compared to more readily available DDR2 memory but, being it's £20 cheaper than the OCZ Platinum that we used as a comparison, we would still tentatively recommend it.