Dynamic RAM (DRAM) stores each bit by using a single transistor and capacitor.
Capacitors are the dynamic element. However, because they lose their charge quickly,
they require a fresh infusion of power thousands of times per second, so DRAM chips
include circuitry that performs refresh operations automatically. Each refresh operation is
called a refresh cycle. Unfortunately, a DRAM chip can t perform a refresh operation at the
same time it performs a read or write operation.
Because DRAM circuitry is simpler, more memory cells can be packed into each chip.
(In other words, DRAM has higher density.) In contrast, fewer SRAM bits can be implemented
in each chip, effectively making SRAM more expensive than DRAM. Despite its simpler
circuitry, DRAM is slower than SRAM because of its required refresh cycles and less efficient
circuitry for accessing bits. With current fabrication technology, typical access times
are 10 to 20 ns for DRAM and 1 to 5 ns for SRAM. Improvements in fabrication technology
can decrease both access times but can t change the performance difference.
Note that neither RAM type can match current microprocessor clock rates, which
range from 3 to 4 GHz at this writing. For zero wait states in memory accesses, these clock
rates require the following memory access speeds: