Related Articles

Today we are looking at Corsair's XMS2-5400UL DDR2 667 memory modules, and putting them through their paces. These are high performance DDR2 memory modules rated up to 675MHz and targetted to those who need a performance part. For us we immediately thought about pushing these to the limits so that's what we did for our review.

The transition to DDR2 is driven by the need for ever-higher data transfer rates, and it is easier to increase the clock speed of DDR2 memory than that of DDR memory. Unfortunately DDR2 generally requires slower timing for control funtions (RAS, CAS etc) and therefore only really shines at high clock rates.

The Corsair site tells us the modules are rated for 3-3-2-8-1T at up to 675Mhz, with a conservative 4-4-4-12 SPD default timing. Since these are supposed to be performance memory modules, we used very aggressive memory timings - a bit more aggressive than they are rated for actually. For most of our testing, we ran them at 3-2-2-7-1T - we want the best performance possible, and in this case the improved timing is far more important than an extra 8MHz that we could get by running at 3-3-2-8-1T. The only real problem that we encountered running at 3-2-2-7-1T is that the modules were not reliable above 667Mhz.

We are using an Intel nForce 4 board for the testing - the current s939 Athlon processors do not support DDR2 modules at this time - however we've heard rumours that next year will bring us DDR2 Athlon 64's.

Benchmark Setup

Our benchmark system (and competitors) consisted of the following:

• Intel Pentium 4 560 (Prescott 3.6 GHz, 800 MHz FSB, LGA775)
• 2*512 MB Corsair XMS2 PC5400UL @ 3-2-2-7-1T 667 MHz
• 2*512 MB OCZ PC2 5400 4-4-4-12 modles
• Asus P5ND2-SLI
• Western Digital WD1200JD SATA hard drive
• NVIDIA Reference 6600 GT 128 MB PCI-Express

The tighter memory timings were achieved without adding any voltage to the RAM. As we do not have other nForce DDR2 benchmark results in our database, for this first memory review we will be comparing different results achieved with the same module on the same motherboard - so in a way, consider this article to be a combination review and memory tweaking guide.

We used Sandra and RightMark for our memory benchmarks - both give useful results in measuring memory performance.

Benchmarks & Overclocking

In order to get a good idea of the performance differences between stock and tweaked memory and processor settings, we are putting our test rig through quite a few tests!

Please note: in order to make our test results easy to read, we labeled the bars on the chart with the processor and memory settings - for example:

18x800/666
3-3-2-8

Means multiplier of 18, 800MHz CPU FSB, 666MHz memory bus, 3-3-2-8 memory timing. All memory timings at 1T.