From RDRAM to Embedded DRAM Michael has it covered, and from the information that he presented in this interview, it's clear that the whole RAM segment of the PC market is undergoing radical changes…

SHARKY EXTREME
Give the readers a short background bio on yourself Michael, and update them as to what your current responsibilities at LG Semicon include.

MICHAEL SPORER
I'm the Director of Technical Marketing for Memory Products at LG Semicon America. I've worked for eight years in the areas of memory technology, specifically: system architectural implications of memory design, design compatibility, test and performance characterization, and high performance memory packaging. At LG Semicon I'm currently working on the development and implementation of next generation memory product solutions, including Rambus DRAM products.

SHARKY EXTREME
Consumers have had to endure system dram upgrades everytime they've upgraded their PCs during the last five years. In 1999, again consumers are going to be forced to upgrade from PC-66 or PC-100 SDRAM to PC-133 SDRAM and again to 200+MHz RDRAM sometime towards the beginning of next year. Why have newer CPUs required a more aggressive specification from DRAM manufacturers more quickly than in the past, and do you see this trend continuing?

MICHAEL SPORER
Years ago the performance of RAM met the requirements of the CPU. CPU performance has increased at a much greater rate than that of the memory. To accommodate these divergent trends there have been modifications to system architecture so that the memory is not the bottleneck of the system. The most significant change has been the incorporation of caches.

Although caches significantly improve the system performance the traditional RAM increasingly lags behind CPU performance. As long as the CPU is not idle while waiting for data everyone is happy. But CPUs will soon surpass the gigahertz clock rate and significant improvements in RAM are needed.

The need for 'new RAM with a new box' will actually become more acute in the future because the architectural system improvements are reaching their point of diminishing returns. The next step is to improve RAM performance.

On the one hand this seems unfortunate, but I have some old systems with of 8MB and 16MB DIMMs. I wouldn't bother to put these old modules in a new system because the best I could achieve would be a 48MB system memory and at the same time I'd render the old system which is still running useless.