Your "7,000 MB/s" NVMe drive hits that number once, in a lab, reading one large file. After 50GB of real writes, it drops to 1,200 MB/s and stays there. The terabyte you paid for? You get 931GB. Here is every SSD specification that actually matters — and every trap manufacturers count on you ignoring.
Claim: "7,000 MB/s"The speed lie — sequential vs. random performance
The biggest lie on every SSD box is the sequential read speed. That number comes from a synthetic benchmark reading one large file in perfect, controlled conditions — near-zero real-world applicability for most users. Your operating system doesn't move large files continuously, it moves thousands of small files randomly, simultaneously. This is called random 4K performance, and it's roughly 50 times more important than sequential speed for boot times, app loading, and general responsiveness.
The specification that actually matters: Random 4K Read/Write IOPS (Input/Output Operations Per Second). A "slow" PCIe 3.0 drive doing 25,000 random 4K IOPS will feel faster in daily use than a "fast" PCIe 5.0 drive doing 15,000 IOPS. Sequential throughput matters for one workflow: video editing with large raw files. Everyone else needs random performance, and manufacturers bury that number in fine print because it's less impressive on a shelf tag.
Claim: "Sustained 3,500 MB/s"Cache reality — the performance cliff no one shows you
Every modern consumer SSD has a cache system — the spec that determines whether your "fast" drive stays fast under sustained load, and the one most reviews deliberately avoid testing past the first few seconds. SLC cache works by designating a portion of TLC or QLC NAND cells to operate in Single-Level Cell mode, storing only 1 bit per cell instead of 3 or 4. SLC is 5-10 times faster than the underlying storage. When the cache fills, you hit the write cliff.
Cache comes in two architectures. Static SLC cache is a fixed allocation that never shrinks. Dynamic cache converts main storage cells temporarily, and collapses under heavy write loads as the drive fills up — a 1TB drive at 90% capacity has a dramatically smaller effective cache than the same drive at 50% full. The specification manufacturers actively hide: sustained write speed after cache exhaustion. A "3,500 MB/s" TLC drive can drop to 800 MB/s once the 40GB cache is saturated. Not a defect. The product working as designed. Also not on the box.
For general computing you'll rarely exhaust the cache. For video editors or anyone regularly moving files larger than 20-30GB, find reviews that explicitly test sustained write performance past cache exhaustion — that benchmark is the real speed of the drive.
Claim: "Rated for years of use"NAND types — the endurance math they don't do for you
SSD endurance is measured in TBW: Total Bytes Written — how much cumulative data you can write before NAND cells begin failing. The NAND type determines that number by an order of magnitude.
| NAND Type | Bits Per Cell | Write/Erase Cycles | Typical Use |
|---|---|---|---|
| SLC | 1 | ~100,000 | Enterprise only |
| MLC | 2 | ~10,000 | Prosumer / enterprise |
| TLC | 3 | ~3,000 | Mid-range consumer |
| QLC | 4 | ~1,000 | Budget consumer |
Budget drives use QLC. Performance drives use TLC — a 3:1 endurance ratio most manufacturers don't prominently label. Real-world math on a 1TB QLC drive rated for 150 TBW: write 10-20GB daily (typical gaming/productivity) and you get 20+ years. Write 50GB+ daily (developer, video editor) and you're looking at 8-10 years. QLC doesn't fail gradually — it tends to fail fast once the degradation curve steepens.
Claim: "M.2 = fast"NVMe vs. SATA — the interface gap
SATA SSDs are physically limited to 550 MB/s — a ceiling set by an interface designed for spinning hard drives in 2003. NVMe connects directly to the CPU via PCIe lanes with no legacy bottleneck: PCIe 3.0 tops out around 3,500 MB/s, PCIe 4.0 hits 7,000 MB/s, PCIe 5.0 reaches 14,000 MB/s theoretical maximum.
For general computing, the gap between SATA and NVMe is noticeable but not transformative — boot times differ by seconds, app loading by fractions of seconds. PCIe 5.0 drives cost 40-60% more than PCIe 4.0 for speeds most workloads cannot use, and run significantly hotter, often requiring active heatsinks. Unless you're moving multi-gigabyte raw video files continuously, PCIe 5.0 is a marketing spec, not a performance upgrade.
VerdictWhat to actually buy in 2026
Budget tier (under $80/TB): Samsung 980 or WD Blue SN570, both TLC. Avoid QLC at this tier unless the price gap is substantial. Look for a visible DRAM cache chip on the PCB — DRAM-less drives save $10-15 but sacrifice random performance under mixed loads.
Performance tier ($80-120/TB): Samsung 980 PRO or WD Black SN850X. TLC NAND, large SLC cache, five-year warranties, PCIe 4.0 interface — specs that actually reflect sustained real-world performance, not just burst peaks.
Professional tier ($120+/TB): Samsung 990 PRO for maximum random performance. For write-heavy work, enterprise-grade TLC with DWPD ratings of 1.0 or higher.
The capacity decision: a 2TB drive typically costs 60-70% more than 1TB but provides double the storage, a proportionally larger SLC cache, and 40-50% better sustained write performance. If you're near the price crossover point, 2TB is almost always the correct purchase.
Samsung 990 Pro (performance tier) and WD Black SN850X (value tier). (Affiliate links — full disclosure on our Deals page.)