It should be noted that the descriptions of the modules and their overclockability tests carry enough information for you to make your choice, but we want to complement these data with the results of performance tests of systems equipped with that memory at the settings recommended by the manufacturers. The tests were performed on the same system that we used for our overclocking experiments.
The memory frequency is always set in sync with the clock generator. That’s why we again had to play with the CPU frequency multiplier:
- The CPU was clocked normally, i.e. as 10 x 200MHz, for testing the PC3200 (DDR400) memory
- The CPU was clocked as 9 x 222MHz for testing the PC3500 (DDR438) memory
- The CPU was clocked as 8 x 250MHz for testing the PC4000 (DDR500) memory
So, all the memory kits were tested under identical conditions; the performance of the systems depended only on the frequency and timings of the memory subsystem. These tests will help us choose the highest-performing kit among the reviewed ones, considering that they all have different frequencies and latencies.
First, we want to run some synthetic tests that are specifically designed to measure the memory subsystem performance.
Quite expectedly, the real bandwidth provided by different memory modules depends directly on the memory frequency. So, the results of the tests aren’t surprising at all.
As for the measured latency, it’s not all so obvious. Latency depends both on the timings and the frequency. That’s why the OCZ PC4000 EB Dual Channel Platinum Edition kit wins the latency test as this memory features rather aggressive timings of 3-3-2-8 and works at 500MHz. The second and third places go to the other DDR500 SDRAM modules, from Corsair and G.Skill, that work with 3-4-4-8 timings.