AI summary by TwinMind

How does this SSD store 8TB of Data? || Inside the Engineering of Solid-State Drive Architecture

Branch Education

9m 37s

SSD Data Storage Capacity and Analogy

  • 8TB SSD can store 64 trillion bits of data
  • Penny analogy: each bit represented by penny (heads/tails = 1/0)
  • 64 trillion coins laid flat would cover entire state of Massachusetts
  • SSD can find 5x5 meter area of pennies anywhere on this surface
  • Can erase/change state of all 64 trillion coins up to 10 times daily
  • Can read state of all pennies ~70 times daily

Physical SSD Components

  • 18 NAND flash microchips store all 8TB data
  • SSD controller microchip (brain) interfaces between chips and computer
  • 8 DRAM chips serve as memory buffer for lookup/translation tables
  • Memory channels: 18 channels (one per NAND chip), each with 16 wires/traces
  • Additional traces for sending read/write commands

NAND Flash Chip Internal Structure

  • Single chip contains stack of 8 die
  • Each die has 2 planes
  • Each plane: vast array of vertically stacked charge trap flash memory cells
  • Organization: thousands of blocks → hundreds of pages → tens of thousands memory cells
  • Each charge trap flash memory cell holds 3 bits of information
  • Total: ~10 million pages across entire drive

Data Flow and Operations

  • Computer requests data (e.g., picture)
  • SSD controller sends request to specific location: chip 9, die 4, plane 1, block 351, page 45
  • 3D NAND flash chip accesses location, sends contents via 16-trace memory channel
  • Controller routes information back to computer
  • Reverse process for writing data to storage

Translation System

  • Computer sees SSD as massive set of 512-byte sectors
  • 8TB drive = 15 billion sectors
  • SSD controller maintains massive spreadsheet (lookup/translation table)
  • Logical to physical translation: computer sees sectors, SSD knows physical locations
  • Translation table continuously updated during use
  • Stored in DRAM chips when active, saved to NAND flash when inactive

Super Pages and Parallel Processing

  • Individual pages can be written/read, but super pages more common
  • Super page = all pages across 18 chips sharing same die/plane/block/page designation
  • Large videos spread across all 18 chips in super page(s)
  • All 18 chips run in parallel for faster read/write times
  • Super blocks also utilized (common block address among all 18 chips)

Additional Controller Functions

  • Wear leveling
  • Garbage collection
  • Error correction code execution

Sponsorship Information

  • Video sponsored by Kioxia
  • Featured SSD is Kioxia data center class SSD
  • Designed for consistent performance/latency in hyperscale applications
  • Kioxia also provides enterprise SSDs for top performance and reliability

TwinMind

Get summaries like this for your own meetings

TwinMind takes notes in the background during meetings, lectures, and calls, then writes the summary for you.

Free plan to get started

No bot joins your calls

Phone, desktop & Chrome apps

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome

Your next meeting, summarized like this

TwinMind listens in the background and turns meetings, lectures, and calls into searchable notes you can chat with. Start free.

Available on phone, desktop & Chrome