Review: Mushkin EMS PC3500 Memory

by Tarinder Sandhu on 10 April 2003, 00:00 4.0

Tags: Mushkin

Quick Link: HEXUS.net/qaqw

Add to My Vault: x

Overclocking with comparisons

Given that there are a number of high-speed modules available from a number of different manufacturers, I wanted to see just how good the Mushkin module was in direct comparison. Three settings have been chosen. Firstly, the super-harsh setting of 2-5-2-2, then the usual performance setting of 2-7-2-2, and lastly a relaxed setting of 2.5-7-3-3.

All the modules were given 2.75v load voltage and had to pass an entire loop of 3DMark 2001SE in order to be ratified at a certain speed. That's not total stability, but more of a guess at the probable MHz ceiling.

Firstly, 2-5-2-2. I'd not recommend a tRAS rating of 5 cycles, at high speeds, simply because it gives the RAM so little time to close down a bank for re-use. If the bank is activated again without being closed down in time beforehand (i.e too short a tRAS rating), data corruption is likely to occur, both on the memory and hard drive level.

The Mushkin module becomes the first to break the DDR-400 barrier with these settings. It seems as if module is tuned to work well with harsh timings.

Now 2-7-2-2

The other modules gain ground on the Mushkin PC3500 stick. It seems as if it just loves really low latency timings.

Now 2.5-7-3-3

The relatively poor performance here lends weight to the assertion that the Mushkin module is most at home with extremely low latency DDR-400+ settings. That's goes against the 2.5-7-4-4 timings prescribed at 217MHz. I'd take low latency over slightly higher MHz any day.

This is what all good, performance-orientated memory should do, ideally. That's at just above default voltage.