What's the difference between volatile and non-volatile memory?

It is telling that a “basic” question like the difference between volatile and non-volatile memory still warrants a feature-length explainer in 2025. That’s because the boundary between the two is the line where a computer’s soul meets its body — where speed gives way to persistence. On one side, DRAM and SRAM fetch instructions at nanoseconds but require constant power; on the other, NAND flash and storage-grade memory retain bits for years at the cost of orders of magnitude slower access. The article efficiently strips away the marketing layer, reminding us that every “instant-on” laptop and every “fast” phone is actually a carefully tuned compromise between these two behaviors. What makes this more than a refresher is how the boundary itself is shifting. The rise of non-volatile memory (NVMe SSDs, but also Intel’s now-discontinued Optane, and new emerging MRAM and ReRAM) is blurring the gap—producing “storage-class memory” that sits closer to the CPU while ensuring data survives power loss. We’re seeing a slow but steady move toward a system where the memory hierarchy flattens, and that has deep consequences for software. If the operating system can treat a volatile state and a non-volatile state with less ceremony (like NVMe’s persistence domains), crash recovery becomes cheaper, transactional systems get less fragile, and the notorious cold-start of an OS may eventually feel as ancient as a magnetic tape boot. The engadget source article does important work here: it frames these technologies not as abstract acronyms but as real trade-offs that define user experience. But I’d argue the deeper story is about how dependent modern AI, cloud, and edge work is on the interplay between the two. Recommendation engines, chat models, and databases all rely on in-memory lookahead and persistent snapshots in canary releases to keep running; losing that distinction would mean dealing with constant re-initialization or a too-slow persistent cache. The next generation of computing — especially AI inference at the edge — lives or dies by that same volatile-vs-non-volatile dance. Still, basic concepts like these matter. At a moment when everyone is talking about terabytes and gigabytes with zero understanding of why RAM is fast and storage is durable, the Engadget breakdown is a necessary conversation starter. It reminds us that the foundation of computing hasn’t changed in half a century, but the materials and the overlays of abstraction are what keep turning.