All DRAM memory designs have the same basic core, and thus the same internal speeds and relatively large latency periods. Various schemes have been developed over the years to optimize or hide these limitations, sometimes improving one while having a negative effect on another. Unfortunately, because of the relatively high cost of SRAM, and the extremely competitive nature of the industry, DRAM is still the only real option for large amounts of system memory and for use in graphics subsystems. SRAM memory is generally used only for cache, either externally (on the motherboard) or internally (small amounts built into the processor or DRAM chip).

Processor speeds have increased at an incredible rate, while memory designers are working to improve DRAM speed without dramatically increasing costs. Assuming that the processor requires more than one clock cycle to execute instructions, the memory subsystem can be two or three times slower than the CPU and still keep pace. With better processor designs, instructions may be executed every clock cycle, or even more than one instruction per cycle. Unfortunately, while processors are currently capable of clock speeds of 500MHz, main memory is limited to 100MHz (perhaps up to 133MHz in some cases). The primary reason for this speed mismatch is because of the inherent slowness of DRAM, which is the type most often used for main memory.

Numerous SRAM and DRAM designs have been proposed and implemented, each attempting to address the issue of speed for a specific situation. Unfortunately, we still do not have an 'ideal' memory architecture that solves all of the problems, because nobody has been able to significantly improve the DRAM speeds with out also adding healthy cost premiums.