Abstract
Modern computing architectures employ a multi-tiered Memory and Storage Hierarchy to balance execution speed, persistent capacity, and financial cost. While volatile DRAM provides sub-microsecond access times for active execution, non-volatile media (NVM, SSDs, and HDDs) retain data across system power cycles at varying latency and cost profiles.
- Category: System Memory & Storage Hardware
- Key Trade-off: Access Latency vs. Cost per Byte
- Volatility Boundary: DRAM (Volatile) vs. NVM / SSD / HDD (Non-Volatile Persistence)
The Storage & Memory Hierarchy Spectrum
Hardware storage technologies range from ultra-low-latency volatile CPU registers down to high-capacity, low-cost mechanical disks:
Technical Parameter Comparison
| Technology Tier | Volatility | Access Latency | Throughput Bandwidth | Typical Capacity | Relative Cost |
|---|---|---|---|---|---|
| DRAM | Volatile | Tens of GB / module | Highest (\$$$$$$) | ||
| Non-Volatile Memory (NVM) | Non-Volatile | A few | Tens to hundreds of GB / module | High (\$$$$) | |
| Solid State Disks (SSD) | Non-Volatile | A few | Moderate (\$$$) | ||
| Hard Disk Drives (HDD) | Non-Volatile | Lowest (\$$) |
System Integration & Design Implications
- DRAM vs. Secondary Storage: Because CPU execution speeds require sub-microsecond memory responses, operating systems treat DRAM as temporary scratchpad memory, utilizing virtual memory mechanisms like Demand Paging to swap idle frames out to SSD or HDD backing stores.
- Emerging NVM Architecture: Persistent Non-Volatile Memory bridges the gap between DRAM and flash SSDs by offering near-DRAM read/write speeds while maintaining persistent data across system reboot cycles without block-erase restrictions.