Introducing various new types of memories from shallow to deep

All users are no strangers to the term "memory", because all electronic products must use memory, and usually use more than one type of memory, which is not a "strategic material"! However, for the types, specifications and forms of memory, many people are prone to confusion. For example, DRAM as the "execution" program (data), and the Flash ROM "storing" programs and materials is an example. This album will be from shallow to deep. We introduce the structure and operation mode of various new types of memories.

Memory classification

The electrical memory refers to the memory of the electrical write electrical reading, which is mainly divided into two categories, as shown in Figure 1:

Volatile Memory (VM): The data exists when the power is turned on, and the data is immediately lost when the power is turned off (data is evaporated), such as SRAM, DRAM, SDRAM, DDR-SDRAM, etc.

Non-Volatile Memory (NVM): When the power is turned on, the data can be retained. For example, ROM, PROM, EPROM, EEPROM, Flash ROM, FRAM, MRAM, RRAM, PCRAM, etc. â–² Figure 1: Classification of memory.

Memory unit

The "cell" of a memory is the smallest structure used to access data. If it contains a transistor (Transistor) and a capacitor (Capacitor), it is called "1T1C"; if it contains a transistor (Transistor) and a resistor ( Resistor is called "1T1R"; if it contains a diode (Diode) and a resistor (Resistor) it is called "1D1R".

Each "unit" of memory does not necessarily store only 1 bit of data. Due to the increasing demand for memory capacity, each "unit" can store more and more data according to each "unit". The number of data bits that can be stored is further divided into: single-level cell (SLC), multi-level cell (MLC), triple-level cell (TLC), and four-layer cell ( Quad-Level Cell, QLC), etc.

Memory hierarchy

To understand the various memory configurations of electronic products, you must first introduce the concept of "Memory hierarchy". The memory level refers to how to properly allocate a variety of memories with different storage capacities, different computing speeds, and different unit prices to achieve maximum economic benefits, so that the product has reasonable computing speed, reasonable storage capacity, and reasonable product price.

Figure 2 is a schematic diagram of the memory hierarchy. The top-down sequence is the scratchpad, cache memory, main memory, and auxiliary memory:

Register: In the processor, used to set the function of the processor, mainly the place to "temporarily store" the set value.

Cache memory: A place in the processor that "temporarily stores" programs and data when executing programs, usually in SRAM.

Main memory: Outside the processor, where the program and data are temporarily stored, usually in DRAM, has been improved to SDRAM or DDR.

Assistant memory: A place outside the processor that "stores" programs and data permanently, including: flash memory, disk drives, CD players, tape drives, etc.

Different types of memory have different storage capacity, working speed, and unit price:

Storage Capacity: Secondary Storage (GB) > Main Memory (MB) > Cache Memory (KB) > Scratchpad (B).

Operating speed: auxiliary memory (1ms) < main memory (10ns) < cache memory (1ns) < scratchpad (1ns).

Unit price: Auxiliary memory < main memory < cache memory < scratchpad.

â–² Figure 2: Schematic diagram of the memory hierarchy.

Memory application

All electronic products must use memory, and usually use more than one kind of memory. Due to the variety of memories, users are often confused. We simply explain the difference between different memories. Figure 3 shows the system blocks of the main chip of the mobile phone. A system block diagram includes an application processor, a baseband processor, and a motion controller.

The application processor is mainly an operating system (OS) and an application program (App). The scratchpad and the cache memory are currently built in the processor, and the register is used to set the processor. The function, the program used to set the value of the scratchpad, that is, the software program used to drive the hardware is also called "Firmware"; the cache memory is used to "temporarily store" the program when executing the program. The location of the data, because it is relatively close to the computing unit in the processor, can shorten the time of the program and data back and forth, and speed up the execution of the program is called "Cache".

Since the cache memory costs are high, the capacity is not large. If the program cannot be placed, it can be stepped back in the main memory. However, the SDRAM or DDR used in the main memory is volatile memory, and the data is turned off immediately. Loss, so the data must be stored in non-volatile auxiliary storage after shutdown. Early auxiliary storage uses disk drives, CD players, tape drives, etc. Due to advances in semiconductor manufacturing, most of the current use of flash memory (Flash ROM), or so-called Solid State Disk (SSD), SSD is actually making flash memory.

Since the cache memory (SRAM) and the main memory (SDRAM, DDR) are places where the execution program is used to "temporarily store" programs and data, the bus unit (Bus) is directly connected to the arithmetic unit in the processor, and is generally used. "bit" to calculate the capacity; and auxiliary storage is the place to "permanently store" programs and data. Since a Byte can store a half-word, it is generally a "Byte" (Byte). To calculate the capacity.

â–² Figure 3: System block diagram of the main chip of the mobile phone.

Static Random Access Memory (SRAM: Static RAM)

It is called "Static" by storing one bit (1 bit) of data in six transistors (MOS) and "requires" to periodically replenish the power supply to keep the contents of the memory.

The structure of SRAM is more complicated (6 transistors store 1 bit of data), the access speed is faster without using capacitors, but the cost is also higher, so generally the paired capacity requirements are lower but the speed is higher. The memory, for example, the central processing unit (CPU) has built-in 256KB, 512KB, 1MB "Cache memory", which is generally using SRAM.

Dynamic random access memory (DRAM: Dynamic RAM)

A transistor (MOS) plus a capacitor (Capacitor) is used to store a bit (1 bit) of data, and when needed, "need" to periodically replenish the power supply to maintain the contents of the memory, so called "Dynamic" .

The DRAM structure is relatively simple (one transistor plus one capacitor). Because the capacitor charging and discharging takes a long time to make the access speed slower, but the cost is lower, the general requirements for the paired capacity are higher but the speed requirements are higher. Low memory, for example, a PC motherboard that typically uses more than 1GB of DDR-SDRAM is a type of DRAM. Due to the speed of the processor, the speed of the traditional DRAM has been unable to meet the requirements, so it has been improved to two types of SDRAM or DDR-SDRAM.

Synchronous Dynamic Random Access Memory (SDRAM: Synchronous DRAM)

The central processing unit (CPU) is the same as the "operation clock" when the main memory (SDRAM) on the motherboard is accessed, so it is called "synchronous". Since the CPU does not need "Wait" when accessing data, so it is more efficient, SDRAM access speed is faster than DRAM, so early computer motherboards use SDRAM to replace traditional DRAM, but there are only a few industries. The computer still uses SDRAM.

DRAM uses a transistor (MOS) and a capacitor to store a bit of data (a 0 or a 1), as shown in Figure 4 (a), when the transistor (MOS) is not conducting, no electrons flow, the capacitor There is no charge, the data representing this bit is 0, as shown in Figure 4 (b); when the transistor (MOS) is turned on (positive voltage is applied to the gate), the electron will flow from the source to the drain, and the capacitor has The charge, the data representing this bit is 1. In order to "storage" these flowing charges, a tiny capacitor must be used. As shown in Figure 4(c), DRAM is because the capacitor needs time to charge, so The speed is slower than SRAM. â–² Figure 4: Schematic diagram of the structure and working principle of dynamic random access memory (DRAM).

Disadvantages of dynamic random access memory

Dynamic Random Access Memory (DRAM) uses a transistor plus a capacitor to store a bit (1 bit) of data. Since the capacitance of a conventional DRAM is made of "yttria" as an insulator, the dielectric constant of yttrium oxide is not enough. Large (K value is not large enough), so it is not easy to attract (storage) electrons and holes, which makes it necessary to constantly replenish electrons and holes, so it is called "dynamic", as long as the power of the computer is turned off, the electrons stored in the capacitor are The hole will be lost and the data stored in the DRAM will be lost.

To solve this problem, the easiest thing to do is to replace the "yttria" as an insulator with a material with a large dielectric constant (K is large enough) so that electrons and holes can be stored in the capacitor without loss. At present, the industry uses "PZT" or "SBT", a "ferroelectric material" with a large dielectric constant (large K value) to replace yttrium oxide. It is possible to store electrons and holes without escaping, so that the original "volatile" dynamic random access memory (DRAM) becomes a "non-volatile" memory called "ferroelectric RAM" (FRAM). ).