semiconductor physics and Devices

Second Semester Lecture Course

Sheng Yun Wu

Week 11: Semiconductor Memories – RAM, Flash, and Emerging Memory Technologies

Lecture Topics:

Introduction to Semiconductor Memories
- Semiconductor memories are devices that store data in digital electronics.
- Types of memories:
  - Volatile memory: Loses data when power is turned off (e.g., RAM).
  - Non-volatile memory: Retains data even when power is off (e.g., Flash memory).
- Memory hierarchy in computers: Ranges from fast but expensive memories (e.g., CPU cache) to slower but larger memories (e.g., hard drives, SSDs).
Random Access Memory (RAM)
- Dynamic RAM (DRAM):
  - Structure: Each bit of data is stored as a charge in a capacitor, with access controlled by a transistor.
  - Operation: Data must be periodically refreshed because capacitors lose charge over time.
  - Speed: Fast read and write operations but requires refreshing, leading to power consumption.
  - Applications: Used as the main memory in computers and mobile devices.
- Static RAM (SRAM):
  - Structure: Each bit is stored using a flip-flop made of transistors, without needing a capacitor.
  - Operation: Does not require refreshing, so it is faster and more power-efficient than DRAM.
  - Applications: Used in CPU caches and small memory sections where speed is critical.
- Comparison between DRAM and SRAM:
  - DRAM: Higher density, cheaper, but slower and requires refreshing.
  - SRAM: Faster and more power-efficient but lower density and more expensive.
Flash Memory
- Flash Memory Structure:
  - NAND Flash: Data is stored in an array of memory cells made from floating-gate transistors.
  - NOR Flash: Similar structure but provides faster read access, typically used in embedded systems.
- Operation:
  - Programming: Electrons are stored on the floating gate to represent data.
  - Erasing: A high voltage is used to remove electrons from the floating gate.
- Characteristics:
  - Non-volatile: Retains data without power.
  - Endurance: Flash memory cells wear out after many write-erase cycles, typically between 10,000 to 100,000 cycles.
- Applications:
  - NAND Flash: Used in solid-state drives (SSDs), USB flash drives, and memory cards.
  - NOR Flash: Used in firmware storage for embedded systems (e.g., BIOS in computers).
- SSD (Solid-State Drives): NAND Flash is the backbone of SSDs, which provide faster access speeds and lower power consumption than traditional hard drives.
Emerging Memory Technologies
- Magnetoresistive RAM (MRAM):
  - Operation: Stores data using magnetic states rather than electrical charge, offering non-volatility with the speed of RAM.
  - Advantages: Fast read/write speeds, high endurance, and non-volatility.
  - Applications: Potential replacement for both SRAM and DRAM in high-speed applications.
- Resistive RAM (RRAM):
  - Operation: Data is stored by changing the resistance of a material, switching between high and low resistance states.
  - Advantages: Non-volatile, faster switching times than Flash memory, and lower power consumption.
  - Applications: Emerging technology for use in IoT devices and AI accelerators.
- Phase-Change Memory (PCM):
  - Operation: Uses materials that can switch between amorphous and crystalline phases, representing data in different states.
  - Advantages: Non-volatile, faster than Flash memory, and scalable.
  - Applications: PCM is used in specialized applications requiring fast non-volatile memory, such as AI systems.
- 3D XPoint (Optane):
  - Developed by Intel and Micron, this memory offers performance between DRAM and Flash memory, combining high speed and endurance with non-volatility.
  - Applications: Used in data centers and high-performance computing systems.
Memory Hierarchy and Speed vs. Capacity Trade-offs
- Memory hierarchy:
  - Registers (fastest, smallest) → CPU Cache (L1, L2, L3) → RAM → SSD/Hard drive (slowest, largest).
- Trade-offs between speed, cost, and capacity:
  - Faster memories like SRAM are more expensive and used in smaller quantities, while slower memories like Flash and hard drives are cheaper but used for mass storage.
- Latency vs. Throughput:
  - Latency: Time taken to access a single memory location.
  - Throughput: Rate at which data can be read from or written to memory.
Memory Applications in AI and IoT
- AI accelerators: Emerging memory technologies like MRAM and RRAM are critical in AI applications where speed and endurance are crucial.
- IoT devices: Low-power, non-volatile memory technologies are essential for battery-operated IoT devices.
- Embedded systems: Flash memory is widely used in embedded systems where firmware must be stored and retained without power.

Examples:

Calculate the memory access time for a DRAM and comparison with SRAM.
Design of a simple memory hierarchy for a computing system, balancing speed and capacity requirements.
Explanation of the programming and erasing process in NAND Flash memory and how it affects endurance.
Comparison of emerging memory technologies' read and write speeds (MRAM, PCM, RRAM) with traditional Flash memory.

Homework/Exercises:

Explain the differences between DRAM and SRAM regarding structure, operation, and applications. Why is SRAM faster but less dense than DRAM?
Compare NAND and NOR Flash regarding speed, endurance, and applications.
Describe the operation of MRAM and its potential advantages over DRAM in terms of endurance and power consumption.
Design a memory hierarchy for a high-performance computing system, specifying where different types of memory (e.g., SRAM, DRAM, Flash) would be used.