Open up an SSD and an HDD side by side and you’ll understand their price and performance differences immediately. One is mostly empty space with a few chips on a circuit board. The other is a sealed enclosure hiding a spinning platter, a read/write head riding a fraction of a millimeter above it, and a motor that has to spin that platter at a constant 5,400 to 7,200 RPM for years without skipping. The mechanical one is the one more likely to fail, and it’s the one doing more physical work to give you the same gigabyte.
What’s actually inside a hard drive
A hard drive is a tiny factory. Inside the sealed case: one or more aluminum or glass platters coated with magnetic material, an actuator arm with read/write heads on the end, a spindle motor, and a controller board. The heads never touch the platter in normal operation, they float on a cushion of air (or helium, in higher-capacity drives) measured in nanometers. That gap is the whole reason hard drives are fragile. Drop one while it’s running, or even bump a desk hard enough, and the head can contact the platter, which usually means a scratched platter and a dead drive, not a loose connection you can fix.
This is also why HDDs are rated for shock in a way SSDs don’t need to be. A spec sheet for a 7,200 RPM desktop drive will list something like 300G non-operating shock tolerance and far less while spinning. Laptops with spinning drives used to include motion sensors that parked the heads if they detected a fall. None of that matters once there’s no moving arm.
What’s inside an SSD
An SSD is a NAND flash controller, some flash memory chips, and usually a small amount of DRAM for caching the mapping table that tracks where data physically lives. There’s no motor, no arm, no platter. Data is read and written electrically by trapping or releasing charge in flash cells. That’s why SSDs tolerate being dropped, shipped, or jostled in a way a spinning HDD never could, and why they’re silent and draw less power at idle.
The trade-off is wear. Flash cells degrade a little with every write cycle. Consumer TLC NAND is typically rated for somewhere around 300 to 600 total terabytes written (TBW) on a 1TB drive, depending on the model. For a typical home user that’s 10+ years of normal use. For someone constantly writing and deleting large files, like video editors or people running VMs with heavy swap, it’s worth checking the TBW rating before buying rather than assuming all SSDs age the same way.
How each one actually dies
HDD failure is often mechanical and sometimes gradual: clicking noises, bad sectors that grow over time, SMART warnings about reallocated sectors. You frequently get warning signs before total failure, and recovery services exist specifically because the data is still physically there on the platter even if the drive won’t spin up.
SSD failure is usually electrical and can be abrupt. A controller failure can take a drive from fully working to completely unreadable with no warning, and because there’s no physical platter to send to a recovery lab, data recovery from a dead SSD is far more limited and expensive when it’s even possible. Flash wear-out, by contrast, is gradual and most drives will go read-only before they go completely dead, giving you a window to copy data off.
Neither of these facts means one drive type doesn’t need backups. Both die. The practical lesson is that backups matter more with SSDs specifically because recovery is less reliable, not less necessary.
Side by side
| Factor | HDD | SSD |
|---|---|---|
| Moving parts | Yes (platter, motor, head) | None |
| Typical sequential speed | 80–160 MB/s | 500 MB/s (SATA) to 7,000+ MB/s (NVMe) |
| Shock tolerance (non-operating) | ~300G | 1,500G+ |
| Cost per GB (2024) | ~$0.02–0.03 | ~$0.06–0.09 (TLC NVMe) |
| Typical failure warning | Often gradual (SMART errors, noise) | Often sudden (controller) or gradual (wear-out) |
| Best use today | Bulk storage, backups, NAS | OS drive, applications, active projects |
What to actually buy
For a system drive, there’s no real debate left. A SATA SSD is already 5-6x faster than a hard drive for everyday tasks, and NVMe is faster still. If your PC still boots from a spinning disk, moving to an NVMe SSD is the single biggest felt-speed upgrade you can make to an aging machine, often more noticeable than a CPU or GPU upgrade for general use.
Where hard drives still make sense is bulk, cold storage: media libraries, backups, archives you don’t touch daily. Cost per gigabyte on a 4TB or larger internal hard drive is still roughly a third of SSD pricing at those capacities, and spinning platters sitting idle in a drawer or NAS enclosure aren’t under the kind of physical stress that causes most HDD failures. If you’re assembling a two-drive setup, an SSD for the OS and programs plus a hard drive for storage is still the most cost-effective split for most people, and you’ll want a SATA to USB adapter on hand if you’re migrating data off an old drive before retiring it.
Don’t overspend on endurance you won’t use. A budget TLC SSD rated for 300TBW will outlast a typical home PC’s useful life. Save the high-endurance, high-TBW drives for workloads that actually hit those limits, like constant video capture or database work, where the cheaper drive would wear out years before you’d otherwise replace the machine.
