Crucial PC3200 Memory Review - Benchmarking

When it comes to evaluating memory, this can sometimes be difficult since the processor and the memory use a very fast bus, and the aim is to make the memory the bottleneck, not the processor. Because of this, the best way to measure the performance of memory is to measure its bandwidth, or rather, how fast it is in terms of moving data in, out, and around.

Let's meet our test system:

1. Intel Pentium 4 2.4C Hyperthreading CPU
2. Soltek SL-86SPE-L Springdale motherboard
3. ASUS Radeon 9600XT 128MB video card
4. Maxtor DiamondMax 9 80GB PATA HD

Operating systems:

1. Gentoo Linux 1.4, 2.4.24 non-SMP vanilla kernel from kernel.org
2. Windows XP SP1 with all extra updates

Memory:

1. 2x256MB Crucial PC3200 DDR400 RAM sticks
2. 2x512MB Infineon PC3200 DDR400 RAM sticks

I debated heavily about what to do with the timings and the amount of memory I was comparing it to. Keeping in mind that the Crucial RAM is not meant for overclocking, leaving the FSB at its defaults and the timings at SPD seemed to be the most reasonable, since that is the situation that this RAM is most likely to be found in. Nobody buys Kingston HyperX and leaves it at non-overclocked speeds, and by the same token, it does not make sense to review this with my computer overclocked. I will, however, overclock it later.

I also included tests with one stick of Infineon RAM. This is to allay concerns that I gave the Infineon RAM an unfair advantage by making more physical RAM available, although having more RAM available should not significantly affect how the RAM performs. The problem with running only one stick of RAM is I cut off the second channel in the dual channel arrangement, and that will seriously hurt bandwidth tests. The unfair advantage assumption is completely unfounded and you will see why when I show the benchmark results.

Today's benchmark suite includes:

1. SiSoft SANDRA 2004
2. PiFast
3. 3DMark 2001SE
4. PCMark2002
5. PSBench 7
6. ScienceMark 2
7. STREAM
8. OpenSSL

I tried to create a good mix of tests that are a mix of synthetic and non-synthetic benchmarks, as well as performance and bandwidth tests. STREAM is a Linux based test that is slowly growing on me as well, and also alleviates some problems I was having in terms of getting good, reproducible results, especially since I had problems with PSBench 7. OpenSSL is an open-source secure sockets layer library. It is used extensively on Linux systems for any form of network security in applications. I will explain my testing method with STREAM and OpenSSL later on.

All results quoted are the average of three independent runs of the benchmark. This is to help remove the effects of anomalous results.

Next: SiSoft SANDRA 2004 >>
 

More Memory Articles
More By Quantum Skyline