mmm@cup.portal.com (Mark Robert Thorson) (12/27/89)
When I first began designing with digital logic, the semiconductor memory vendors always had a line of large shift registers. Generally, the shift registers had double or quadruple the number of bits of RAM's made using the same technology. These were often used in frame-buffer graphics displays and high-speed swapping disk emulation. Today, almost nobody (or is that just plain "nobody") makes big shift registers anymore. My question is: does the difference in bits-per-chip between silicon RAM and shift register still exist? Magnetic bubble and CCD bulk memory technologies were entirely based on shift register architectures. Both died when they tried to compete head-on with magnetic disk. I think they died because they were chasing a moving target; their cost-per-bit was always higher than the latest generation of hard disks. After those solid-state technologies died, magnetic disk technology continued to progress to truly incredible heights, and the progress keeps on coming. But I wonder whether improvements in semiconductor processing may have changed the equation. If current generation bubbles or CCD's could be used to make a 20 meg low-cost uncrashable hard disk, that could be a good product. Especially for laptop portables. So here's another question: are magnetic bubbles or CCD's scalable down to the fine geometries used in today's DRAM's?