With the advent of medium-scale and large-scale integrated circuits, all-electronic solid-state memories have become a reality. Semiconductor memories are smaller than equivalent core memories; they can be faster and consume less power; and in the long run, semiconductor memories will probably be substantially cheaper.

Bipolar-transistor static memorics, which are essentially multiple flip-flop registers plus read/write/selection circuits, are directly compatible with transistor/transistors or emitter-coupled logic and are very fast: read and write times below 50 nsec are readily possible. No rewriting after reading is needed (nondestructive readout, NDRO). Bipolar memories are fairly complex integrated circuits and are still expensive. They are, therefore, used mostly in small "scratchpad" memories.

随着中规模和大规模集成电路的出现，全电子固态存储器已成为现实。半导体存储器比等效的磁芯存储器体积小，速度更快，消耗的功率更少，而且，其价格最终大概会便宜得多。

双极晶体管静态存储器本质上是由多触发寄存器加上读、写、选择电路构成的存储器，直接与晶体管/晶体管逻辑电路或射极耦合电路兼容，速度又非常快：读、写时间很容易达到50毫微秒以下。读出后无需重写（非破坏性读出，NDRO)。双极存储器是相当复杂的集成电路，现在的价格仍然很贵。因此，它们主要用于小型的“哲时"存储器。

MOSFET (metal-oxide-silicon field-effect transistor) semiconductor memories involve simpler integrated-circuit patterns and are cheaper than bipolar memories. While older MOSFET circuits needed level-changing amplifiers to supply large logic-voltage swings, some newer MOSFET memories are TTL-compatible. MOSFET memories also come as static (flip-flop-register) memories but usually as dynamic memories. In a dynamic memory, each bit is stored in a shift register whose output is fed back to the input through a clock-gated MOSFET refresh amplifier, so each stored bit is recirculated and regenerated, say, 1000 times/sec. The refresh amplifier can be time-shared among 16 to 32 memory cells. Simple silicon-substrate MOSFET memories are slower than bipolar memories (and slower than some core memories). Typical access times are between 300u sec and 2u sec with nondestructive readout. Different types of MOSFET circuits (complementary MOSFETs, sapphire and garnet substrates) are under active development and can be expected to lead to substantially faster MOSFET memories. Compared to core memories, semiconductor memories have the advantage of nondestructive read-out. On the other hand, semiconductor memories are volatile; i.e., memory contents are destroyed when computer power is turned off. In sufficiently critical applications, one must provide an emergency power source, such as a trickle-charged battery which, when a power failure is sensed, can take over memory operation for a time sufficient to transfer the entire contents of the memory onto an auxiliary magnetic storage medium (disk or tape).

MOSFET（金属氧化物硅场效应晶体管）半导体存储器的集成电路结构比较简单，价格比双极存储器便宜。虽然老式MOSFET电路曾需要变电平放大器来供给大的逻辑电压变动，但是一些新的MOSFET存储器却是与晶体管-晶体管逻辑电路兼容的。MOSFET存储器也可作为静态（触发寄存器）存储器，但通常是作为动态存储器的。在动态存储器中，每一个位都存储在移位寄存器中，移位寄存器的输出再通过具有时钟脉冲门的MOSFET更新放大器被反馈给输入。因此，每一个存储起来的位都以，譬如说每秒一千次的速度反复循环和反复产生。更新放大器能在16~32个存储单元中间进行分时。简单的硅衬底MOSFET存储器比双极存储器速度慢（也比某些磁芯存储器的速度慢）。典型的存取时间在300微秒到2微秒之间，而且读出是非破坏性的。不同类型的MOSFET电路（互补型MOSFET，蓝宝石和石榴石衬底），都在积极研制之中，并可望发展成速度快得多的MOSFET存储器。同磁芯存储器相比，半导体存储器具有非破坏性读出的优点。另一方面，半导体存储器是易失的，即如果计算机的电源被切断的话，存储器里的信息就会被破坏掉。因此，在特别重要的应用场合，必须配备紧急电源，例如连续充电池，一旦检测到电源故障，该电池即可接替存储操作，使存储器的全部信息有足够时间转移到辅助磁性存储媒体（磁盘或磁带）上去。

Plated-wire memories are magnetic memories which utilize small zones of magnetizable thin films plated onto wires, rather than magnetic cores, for bit storage. Plated-wire memories permit fast access (access times as low as a few hundred nanoseconds) with nondestructive readout, and are nonvolatile. But, although batch-production methods have been developed, quality control is not simple. As a result, plated-wire memories are not cheap and have been applied mostly in higher-priced digital computers (especially in aerospace-vehicle computers); MOSFET memories seem to have overtaken plated-wire circuits in the low-cost minicomputer field. This situation may or may not be changed by future improvements in plated-wire-memory fabrication.

镀线存储器是用镀在线上的小区段磁化蒋膜而不用磁芯来进行位存储的磁性存储器。镀线存储器可实现快速存取（存取时间才几百毫微秒），其读出又是非破坏性的，而且是非易失的。但是，虽然已经研究出了成批生产的方法，质量却不易控制。因此，镀线存储器的价格并不便宜，只用于高价的数字计算机（特别是用于宇航飞行器的计算机）；而MOSFET存储器在成本低的小型计算机领域内，似乎已经赶上了镀线电路。将来在镀线存储器制作技术上的改进，也许会改变这种情况，但也许不会。

责任编辑：admin