Computer Memory Architecture and Evolution
- 1. COMPUTER Memory Architecture And Evolution Understanding the Building Blocks of Memory Hierarchy Page 01
- 2. Memory Architecture Memory architecture refers to the structure and organization of memory components within a computer system. It dictates how data is stored, accessed, and managed across various types of memory like RAM, cache, and storage devices. Page 02 INTRODUCTION TO Memory architecture is a critical part of computer system design. It directly impacts system performance, processing speed, and efficiency. A well-designed memory system ensures that the processor can quickly access the data it needs, improving overall sys Importance in Computer Systems • Speed – Ensuring fast data access to meet the high-speed demands of modern processors. Cost- • Efficiency – Balancing performance with affordability by using a hierarchy of memory types (e.g., cache, RAM, hard drive). • Capacity – Providing sufficient storage to handle applications and large datasets effectively, especially in high-demand systems. Key Goals of Memory Architecture
- 3. TYPES Of Memory Page 03 Primary Memory • RAM (Volatile) • ROM (Non- Volatile) Secondary Memory • HDDs, • SSDs Tertiary and Off-line Storage • HDDs, • SSDs
- 4. Page 04 Hierarchy Memory Trade-off between speed, size, and cost Register s Cach e RAM Secondary Storage
- 5. Page 05 Hierarchy Memory Trade-off between speed, size, and cost Component Speed Cost Capacity Registers Cache RAM (Main Memory) Secondary Storage Fastest Very fast Fast Slow (relative) Most Expensive Expensive Moderate Cheapest Very Small (bytes) Small (KB– MB) Medium (GBs) Huge (GB– TB+)
- 6. Cache Memory Page 06 Cache Memory is a small, high-speed memory located close to the CPU (or inside it) that stores copies of frequently accessed data and instructions. Its job is to reduce the time it takes for the CPU to access data from the main memory (RAM).
- 7. Inside CPU core Smallest (KBs) Fastest ~1ns Stores most-used instructions/data L1 Inside or near CPU core Larger (KBs–MBs) Very Fast ~3–10ns Backup for L1, more storage L2 Shared by all CPU cores (on chip) Largest (MBs) Slower than L1/L2 ~10–20ns Coordinates and feeds L1/L2 caches L3 Cache Memory Page 07
- 8. Page 08 CPUs were much faster than main memory, causing bottlenecks. First caches were introduced as small, single-level (L1) caches integrated into the CPU. Basic direct-mapped designs. Early Days (1970s–1980s) As CPU speeds increased, L2 caches were added, initially external, then integrated into the CPU. Multi-Level Caching (1990s) Cache Memory EVOLUTION
- 9. Page 09 Introduction of L3 caches, shared among multiple CPU cores. Helped in reducing memory latency in multi-core systems. L3 and Shared Caches (2000s) Modern caches now include: Prefetching (predicting and loading needed data early). Replacement policies (smart algorithms to decide what to keep/remove). Smart and Adaptive Caches Cache Memory EVOLUTION
- 10. Page 10 RAM Random Access Memory RAM is a fast, temporary memory used by the computer to store data and instructions that are actively being used by the CPU. It allows for quick read and write access, making it crucial for system performance. Dynamic RAM Slower Cheaper Main system memory (RAM) DRAM Static RAM Faster More Expensive Cache memory (L1, L2) SRAM
- 11. Vacuum tubes and mercury delay lines used initially. Followed by magnetic core memory — bulky but faster and more reliable. Early RAM (1940s–1950s) Dynamic RAM (DRAM) developed — stores data as charges in capacitors. Smaller, cheaper, and could store more data than core memory. DRAM is Born (1970s) Static RAM (SRAM) introduced — faster than DRAM, but more expensive and used for caches (not main memory). SRAM Development Page 11 EVOLUTION RAM Random Access Memory
- 12. Synchronous DRAM (SDRAM) synced with the CPU clock for better speed. DDR DDR2 DDR3 DDR4 DDR5 → → → → SDRAM and DDR RAM (1990s–2000s) LPDDR (Low Power DDR) for mobile devices — optimized for energy efficiency. Low-Power RAM 3D-stacked DRAM (e.g., HBM – High Bandwidth Memory) Non-volatile RAM technologies aim to combine RAM speed with permanent storage capabilities. Emerging & Future RAM Page 012 RAM Random Access Memory EVOLUTION
- 13. Page 13 ROM Read-Only Memory ROM is a type of non-volatile memory that stores permanent or semi-permanent data—typically system-level software like firmware or boot instructions. Unlike RAM, data in ROM is not lost when the power is turned off.
- 14. Programmable ROM Programmable only once Non-erasable Can be written once by user PROM Erasable Programmable ROM Programmable Erasable with UV light Can be rewritten after erasing EPROM Electrically Erasable PROM Programmable Erasable electronically Can be updated without removing chip EEPROM ROM Read-Only Memory Page 14
- 15. ROM Read-Only Memory Page 14 Data written during manufacturing Permanent – cannot be changed Used in early calculators and computers Cheap in large quantities Mask ROM (1950s - 1960s) Users could program it once only Used a special device to burn the data Useful for small-scale custom applications PROM (1970s) Could be erased with UV light and reprogrammed Transparent window on chip for UV exposure EPROM – (1970s - 1980s)
- 16. ROM Read-Only Memory Page 14 Electrically erasable and reprogrammable Could erase individual bytes Used in BIOS chips, smart cards, etc. EEPROM (1980s - 1990s) A type of EEPROM, but faster and more compact Can be erased/written in blocks Used in USB drives, SSDs, mobile phones Dominant form of non-volatile memory today Flash Memory (1990s - now)
- 17. Page 15 Research paper Referred to Evolution of Memory Architecture