Hardware — RAM and Storage

Week of 2026-10-20 · Download .docx

Objectives

Compare storage interfaces and their throughput
Explain RAID levels and their redundancy characteristics
Identify RAM types and ECC use cases

Key terms

NVMe: Non-Volatile Memory Express — PCIe SSD protocol delivering 3,000–7,000+ MB/s read speeds.
SATA III: Serial ATA revision 3 — maximum 600 MB/s; used for SSDs and HDDs.
M.2: Small internal slot supporting SATA or NVMe; key notch type determines supported protocol.
RAID 0: Striping — data split across drives for performance; zero redundancy.
RAID 1: Mirroring — identical data written to two drives; full redundancy, 50% capacity.
RAID 5: Striping with distributed parity — requires 3+ drives; tolerates one drive failure.
RAID 10: Stripe of mirrors — combines RAID 0 speed with RAID 1 redundancy; needs 4+ drives.
S.M.A.R.T.: Self-Monitoring, Analysis and Reporting Technology — drive firmware health monitoring.
DDR5: Current DDR generation; physically incompatible with DDR4 slots.
ECC RAM: Error-Correcting Code RAM — detects and corrects single-bit errors; primarily for servers.
Secure Erase: SSD firmware command overwriting all NAND cells to render data unrecoverable.
Click of death: Repetitive clicking from HDD indicating imminent mechanical failure.

The concept

STORAGE INTERFACE HIERARCHY

Not all storage is equal in speed. A traditional SATA III hard drive or SSD tops out at 600 MB/s due to the protocol's bandwidth ceiling. NVMe SSDs connect through PCIe lanes — PCIe 3.0 x4 delivers about 3,500 MB/s; PCIe 4.0 x4 reaches 7,000+ MB/s. Both SATA and NVMe drives can use the M.2 physical slot — the key notch type and motherboard wiring determine which protocols are available.

RAID LEVELS

RAID 0 (striping) splits data evenly across drives so simultaneous I/O boosts performance, but there is zero redundancy — one drive failure destroys the entire array. RAID 1 (mirroring) writes identical data to two drives simultaneously; losing one drive causes no data loss. RAID 5 (distributed parity) requires a minimum of three drives — data and parity are striped across all drives, so any single drive can fail and be rebuilt. RAID 10 combines RAID 1 mirrors with RAID 0 stripes for both performance and redundancy but requires at least four drives.

RAM GENERATIONS AND COMPATIBILITY

DDR4 and DDR5 RAM use different key notch positions and are physically incompatible — a DDR5 module cannot fit in a DDR4 slot. A BIOS update cannot enable DDR5 on a DDR4 motherboard because the memory controller and PCB traces also differ. ECC RAM detects and corrects single-bit memory errors using additional check bits. It is standard in servers where undetected memory corruption could cause data loss or crashes.

STORAGE HEALTH AND DISPOSAL

S.M.A.R.T. tracks internal drive parameters like reallocated sectors and spin-up time. When values exceed thresholds, the drive is warning of imminent failure — back up immediately. A clicking hard drive (the 'click of death') means the read/write head is failing to locate its target track repeatedly. SSD Secure Erase triggers the drive's firmware to overwrite all NAND flash cells, making data unrecoverable without leaving media intact for analysis.

Worked examples

Example 1: A NAS device with four drives fails one disk. Investigation reveals it uses RAID 5. Impact: the array continues operating in degraded mode using the parity data from the remaining three drives. Action: replace the failed drive immediately; the RAID controller will rebuild the array. RAID 5 can survive exactly one drive failure — a second failure before rebuild completes destroys all data.

Example 2: A user reports their PC's Event Viewer shows multiple disk error events and the drive is making occasional clicking noises. Action: run a S.M.A.R.T. diagnostic tool (CrystalDiskInfo or Speccy). If critical attributes are above threshold, back up all data immediately and plan drive replacement. Never delay — clicking indicates physical failure in progress.

Common mistakes

Assuming RAID 1 is a backup — RAID 1 protects against drive failure but not accidental deletion or ransomware.
Thinking any M.2 drive works in any M.2 slot — check key type and whether the slot is wired for NVMe or SATA.
Using Quick Format to prepare a drive for disposal — it does not erase data; use Secure Erase or physical destruction.
Installing DDR5 RAM in a DDR4 motherboard — physically incompatible; the key notch prevents insertion.
Ignoring S.M.A.R.T. warnings — these are predictive alerts; act before the drive fails completely.

Self-check

Try each question before reading the answer. Answers at the bottom of this page.

1. Which storage interface is fastest?

NVMe PCIe
SATA III
USB 3.2 Gen 2
Thunderbolt 3

2. RAID 5 requires a minimum of how many drives?

3. RAID 0 provides:

Improved performance with NO redundancy
Full redundancy by mirroring
Both performance and redundancy
Parity-based fault tolerance

4. ECC RAM is used primarily in:

Servers
High-end gaming PCs
Low-power laptops
Any Windows Pro system

5. A clicking hard drive sound indicates:

Imminent HDD mechanical failure
Normal disk seek activity
Fan bearing failure
PSU noise

Self-check answers

1. A — NVMe PCIe 4.0 reaches 7,000+ MB/s — far beyond SATA III's 600 MB/s ceiling.
2. A — RAID 5 needs at least 3 drives to distribute parity and tolerate one drive failure.
3. A — RAID 0 stripes data for speed but has zero redundancy — one drive failure loses all data.
4. A — ECC corrects single-bit memory errors; critical in servers where data integrity is essential.
5. A — The 'click of death' is the read/write head repeatedly failing — back up immediately.

Canvas is the official record. This companion enhances the PGCC curriculum; it does not replace it. Last name and class year only. Students with a 504 plan or IEP: your accommodations apply.

↑ Back to top