http://www.businessweek.com/technology/content/jan2004/tc20040121_9640_tc139.htm
While RAM has widely been perceived as a major bottleneck in current PC architecture, it has been the 'only game in town'. The fact that RAM is slow is compounded by its need to be refreshed and its volatile nature. The article suggests alternatives which are in development and which have the potential to completely change PC architectures in the next 5 - 10 years.
A current of related evidence supports this estimate -- in the last couple of years we have moved to the 1GB level on a well-stocked desktop [and indeed, 256MB is getting close to the acceptable minimum on even low-end systems], and servers routinely press against the 4 GB limit of current 32-bit chips. Size, heat, and cost considerations all suggest that we won't simply go on adding DIMM chips to make this memory possible [though the recent announcement of 2GB modules certainly increases memory density to the point that few slots are needed to support 4GB RAM], making a new architecture more, rather than less, likely.
Add to this the overall need for system speed, particularly in graphically intense applications, and a new architeture looms ever more probable.
Posted by jho at January 28, 2004 12:20 PM