Update 12/14/2004:  This article (and the information contained within it) accurately describes the problems we encountered but mis-identifies their cause.  The actual cause of our test difficulties is discussed here; the information in that article supercedes this one.  The MSI K8N Neo2 does not falsify its CPU multipliers; this article has been left online for archival purposes only. 

Original Article Follows:

This isn?t what I intended to write or publish tonight. Originally, tonight was supposed to be a review of Mushkin?s DDR500 on both AMD and Intel platforms.  Instead, a simple RAM review turned into an in-depth exploration into why MSI?s K8N Neo2 Platinum performs the way it does.

Of RAM and RAM Scaling:

I started my Mushkin review with the P4 system running on 875 Canterwood.  I plugged in the DDR500, ran my tests at DDR400 speeds (to establish a baseline), adjusted the multiplier and bus speed, and then re-ran at DDR500. 

Here?s the shape of the line I got.  Its exactly what I?d expect?a 25% increase in potential bandwidth from DDR400 to DDR500 yields a 22.4% increase in tested bandwidth.  Overhead means we never quite get the theoretical performance boost; but Sandra (being a linear test) delivers almost all of it.

On the P4 platform, I ran a full suite of tests at DDR500, including Super Pi, which beats heavily on both memory controller and main memory when you do a full 32 million iteration.   The Mushkin RAM passed a Super Pi run flawlessly at DDR500 on the P4 platform. That bold is important, so pay attention.

next >

---

• AMD + DDR500: The Puzzle Begins. (1 of 5)
• Scaling Errata Explored in-depth. (2 of 5)
• Everest, Registers, and Hardware Identification: Anatomy of a Lie. (3 of 5)
• The False Half-Multipliers: Accident or Deliberate Mis-Representation? (4 of 5)
• MSI?s Decision to Fake Multipliers Sabotage Other Manufacturer?s Products (5 of 5)