Skip to main content

Parallel Processing, Part 2: RAM and HDD

Single Or Dual Channel RAM; Single, Dual And Quad Hard Drives

In the first part of our parallel processing article series, we discussed the performance differences among single, dual and quad core processors, because an increased number of processing units delivers much better performance gains than any clock speed increase could. The performance benefits of processing workloads in parallel, or at least deploying multiple devices, is found not only in processors, though. Dual and even quad graphics setups have been around for years, and two more components have benefited from this trend too: RAM and hard drives.

Memory controllers were, in fact, the first component that went from a single unit to a parallel layout, utilizing two memory channels in an effort to increase bandwidth. In early 2003, Intel introduced the 865 and 875 chipsets, which transitioned Pentium 4 platforms on Socket 478 from a single DDR333 memory controller to dual-channel DDR400 logic. At that time, memory bandwidth was more than doubled on paper, but the performance impact was also experienced in real life. But what’s the situation today? Many notebooks run with only a single memory module (channel), and no one seems to notice an impact on performance. In addition, the current Intel Core 2 Duo processors come with a huge and sophisticated 2 MB or 4 MB second level cache Compare Prices on Core 2 Duo Processors, which reduces the impact of either poor or super-fast memory. Is dual channel memory really necessary?

Since we felt that the result of a test between single and twin channel memory configurations was somewhat predictable, we decided to go further. We added some more benchmarks to furnish this article, comparing the performance difference between a single hard drive and RAID 0 setups with two and four drives using popular benchmarks. While we already did lots of analytical articles on RAID, this time the focus is on mainstream desktop benchmarks, instead of storage benchmarks.

Join our discussion on this article!