Compilation of public knowledge and experience, originality protection, reproduction prohibited.
专栏 《AutoSARexistence》
<<<< 返回总目录 <<<<
1 Storage system level
Let me ask a question first, why do we need to divide the memory levels?
The faster the speed, the higher the cost, and the smaller the storage space planned from an economic point of view. Therefore, a level-by-level cache strategy must be adopted to save more important data in faster storage, and the storage farther away from the CPU core is larger. Because there are layers, there are also different memory chips.
The storage hierarchy is the order in which the storage system hierarchy is arranged under the computer architecture. Each tier has higher speed, lower latency, and smaller capacity than the next tier (with a few exceptions, such as AMD's early Duron CPUs). Most of today's CPUs are very fast. Most program workloads require memory access. Because cache efficiency and memory transfers are at different levels in the hierarchy, they actually limit the speed of processing, causing the CPU to spend a lot of time waiting for memory I/O to complete its work.
The storage levels in most cars, computers, and chips are as follows:
Register: Probably the fastest access. On a 32-bit processor, each register is 32 bits. x86 processors have a total of 16 registers.
Cache (L1-L3: SRAM, L4: DRAM)
- Level 1 Cache (L1) – Typically accesses only take a few cycles, typically dozens of KB.
- Level 2 Cache (L2) – Approximately 2 to 10 times higher latency than L1, typically a few hundred KB or more.
- Level 3 Cache (L3) – Higher latency than L2, typically several MB in size.
- Level 4 cache (L4) (not necessarily present) – DRAM external to the CPU, but faster than main memory.
Main memory (DRAM): Access requires hundreds of cycles and can be as large as tens of GB.
Disk storage: Flash flash memory, or disk, requires thousands of cycles and has a very large capacity.
There are many types of memory, which can be divided into primary memory and secondary memory based on its purpose. Main memory, also called memory, is known for its fast access speeds, but its capacity is relatively small and its price is relatively high. Auxiliary memory, also called external memory, has larger capacity and lower price, but has slower access speed. Main memory stores programs and data that are needed immediately, while external memory stores programs and data that are not needed temporarily. Information is exchanged frequently between the two.
2 Common memory classifications
Storage ICs are widely used in the automotive market, and DRAM and NAND FLASH occupy the majority of the storage market. Memory chips can be divided into volatile and non-volatile according to whether they can continue to save data after power off. Volatile memory chips can be divided into DRAM and SRAM, and non-volatile memory chips can be divided into NAND FLASH. and NOR FLASH. The current storage market is dominated by DRAM and NAND FLASH.
(1) Volatile memory chips: SRAM and DRAM are the two mainstream types in the current storage market. DRAM has the characteristics of large capacity, low price, and strong scalability. It is more suitable for occasions that require high storage capacity and low requirements for reading and writing speed. On the mobile side, LPDDR ((Low Power Double Data Rate SDRAM)) is more popular due to its low power consumption and small size.
(2) Non-volatile memory chip: NOR flash memory has an efficient reading speed and can directly execute applications in the chip. Therefore, it is widely used in some occasions that require high-speed reading, such as system startup, device firmware, etc. However, in the application scenario of small-capacity memory, the cost performance of NOR is not superior to that of NAND.
NAND flash memory occupies a large share in the storage market due to its high storage capacity, excellent rewriting speed, and more affordable price than NOR. It currently occupies 42% of the market share and has become a strong competitor of NOR.
As technology continues to develop, the storage capacity of flash memory is also constantly improving. Currently, most flash memory on the market uses "2D NAND" technology, which is a planar flash memory chip with a process of 16nm and above. In order to further increase storage density and reduce unit storage costs, the manufacturing process of 2D NAND chips is continuously shrinking to 15/16nm.
In addition, in order to further improve storage density, the industry also uses 3D stacking technology. 3D NAND can further increase storage density by increasing the number of transistors per unit area, while also reducing read latency and power consumption. "3D NAND" chips currently on the market are gradually becoming an emerging market trend.
The distribution diagram of memory in each system in the new electronic and electrical structure is as follows:
Application of memory chips in automobiles
2.1 ROM
For historical reasons, although some types can be read and written, they are collectively called Read-Only Memory (ROM).
PROM: Programmable ROM, programmable ROM, but can only be programmed once. Each memory cell has a fuse that can only be blown once with high current.
EPROM: Erasable Programmable ROM, erasable and programmable ROM, which can be erased and reprogrammed up to 1,000 times. EPROM has a transparent quartz window that allows light to reach the memory cell. When ultraviolet rays shine through the window, the EPROM cell is cleared to 0; programming the EPROM is accomplished by using a special device that writes 1 to the EPROM.
EEPROM: Electrically Erasable PROM, electronically erasable PROM, which does not require a physical independent programming device and can be programmed directly on the circuit board. It can be programmed up to 10^5 times.
2.2 SRAM
SRAM is the abbreviation of Static Random Access Memory (Static Random Access Memory). It is a high-speed cache memory that is used as a cache memory (cache). It can be on the CPU chip or off-chip. SRAM has fast read and write speeds, but is expensive and has small capacity.
Each bit of SRAM is stored in a bistable memory cell, and each cell is implemented with 6 transistors. Such a circuit can remain in one of two different voltage configurations, or states, indefinitely as long as power is available. Even if there is interference, when the interference is eliminated, the circuit will return to a stable value.
In terms of technology substitution trends, MRAM is more likely to replace SRAM due to its durability advantages.
MRAM is the abbreviation of Magnetic Random Access Memory. It has the characteristics of non-volatility, fast reading and writing, low power consumption, and no need to refresh. Due to the persistence advantage of MRAM, it is expected to replace SRAM as cache memory in the future.
2.3 DRAM
DRAM is the abbreviation of Dynamic Random Access Memory. It is a highly integrated, large-capacity and cheap memory. DRAM has the disadvantages of large capacity and low price, but slow reading and writing speed and the need to refresh.
Each bit stored in DRAM memory relies on the charge state of a tiny capacitor, typically only about 30 millipofarads in size. Although DRAM has a small capacitance, leakage still occurs, causing the memory to lose charge within 10 to 100 milliseconds. Fortunately, the clock cycle of the CPU is usually at the nanosecond level, so the voltage of the DRAM memory is still relatively stable. In order to keep each bit of DRAM memory charged, the memory system must periodically read and rewrite each location in the memory, which is where the term "dynamic" comes from.
The DDR series is expected to become the mainstream category of DRAM. DDR (double data rate) SDRAM is a high-speed dynamic random access memory. It has the characteristics of high-speed transmission, high bandwidth, low power consumption, and low price. Due to its higher transmission rate and increased bit width of data pre-reading, it is expected to become the mainstream category of DRAM.
LPDDR (Low Power Double Data Rate SDRAM) is a low-power double data rate synchronous dynamic random access memory. It has the characteristics of low power consumption, small size and high performance, and is more suitable for mobile applications.
2.4 MRAM
MRAM is a non-volatile magnetic random access memory. It has the high-speed reading and writing capabilities of SRAM, as well as the high integration of DRAM, and can basically be written repeatedly unlimited times.
MRAM是汽车应用的理想选择存储芯片,MRAM具有快速且不易失的特点。实时监控的传感器数据可以实时写入，而不需要负载均衡或ECC开销。AEC-Q100 1级合格的MRAM将在发动机罩下应用中发现的延长温度（-40℃至125℃）下保留数据20年。意外断电不会影响数据完整性。
2.5 NOR Flash
Nor Flash and NAND Flash both belong to Flash, and Flash memory can be subdivided into many categories, which is beyond the scope of this article. Just post a picture for everyone to see.
NOR Flash is a memory chip with a very wide range of applications. Basically, mainstream electronic products use it. It is even used inside our commonly used mobile phone cameras and screen driver circuit boards. NOR Flash is mainly used to store codes and some relatively small data files. The mainstream interface is SPI NOR. The capacity range is between 1Mbit and 128Mbit. The packaging form is mainly SOP-8 and the size is small.
Structure: NOR Flash memory uses a parallel structure to organize storage units into storage unit arrays. Each storage unit array contains multiple blocks, and each block contains multiple sectors. This structure makes NOR
Flash has faster reading speed and lower access latency.
NOR Flash memory is a storage device using a parallel structure. It organizes memory cells into memory cell arrays, each memory cell array contains multiple blocks, and each block contains multiple sectors. This design enables NOR Flash to have faster read speeds and lower access delays.
The main features of NOR Flash memory are fast read speed and short access latency. It is suitable for applications requiring fast random access. In addition, NOR Flash also has faster erasing and writing operations, and can be erased by byte or sector. However, the storage density of NOR Flash is relatively low and is only suitable for small-capacity codes and programs.
Due to its fast read speed and random access capabilities, NOR Flash memory is widely used in embedded systems to store code, firmware, and startup programs.
The architecture of NOR Flash determines that its capacity cannot be too large, and its reading speed is relatively slow. But the advantage is that it is easy to use and easy to understand. Some data can even be accessed directly using addresses without setting up a file system. This is very popular among Siege Lion friends.
2. 6 NAND Flash
As a storage medium, NAND Flash is widely used in storage products such as solid state drives, UFS, eMMC, SD cards, USB flash drives, etc. Flash memory is a non-volatile memory, and data will not be lost even if the power is turned off.
Structure: NAND Flash memory adopts a parallel structure and organizes storage units into storage unit arrays. Each storage unit array contains multiple pages, each page contains multiple blocks, and each block contains multiple sectors. This structure enables NAND Flash to have higher storage density.
Features:
High storage density: suitable for storing large amounts of data.
Fast reading speed: NAND Flash has fast reading speed
Long erasing and writing time: The erasing and writing operation of NAND Flash takes a long time. Erasing is usually done in blocks.
Suitable for storage applications: Due to high storage density and lower cost, NAND Flash is widely used in large-capacity storage devices, such as solid-state drives (SSD), USB flash drives, and memory cards.
The evolution of automotive NAND Flash memory
Early GPS navigators, driving recorders and other vehicle-mounted equipment were basically equipped with SD cards. Since it only relies on the gold finger of the SD and the metal contact connection in the card slot, it cannot reliably adapt to the changing environment of bumps, high temperature, and high humidity in the car for a long time.
The use of storage devices in the automotive industry has begun to shift from SD cards to eMMC and UFS storage chips, which provide better stability and read and write speeds. Unlike SD cards, which only have a 4-bit bus width, eMMC cards have an 8-bit bus width, which can provide the fastest reading speed of 400MB/S, and the reading speed of UFS memory chips can reach an astonishing 2.9GB per second. . In the future, NVMe protocol automotive-grade SSDs based on the PCIe interface will provide data throughput rates greater than 10GB per second, and massive storage capacity will provide strong support for the next generation of smart vehicle systems. These storage devices also have real-time response and temperature management features that can quickly adapt to complex environmental changes inside the car.
In the next lecture, we will introduce it from the storage subsystem level in order, like peeling an onion, slowly unveiling its veil.
reference
| serial number | Link |
|---|---|
| 1 | Understanding automotive memory chips in one article |
| 2 | SOC processor storage |
| 3 | Introduction to computer storage levels and common storage |
