On Thu, Dec 05, 2024 at 07:32:03AM -0500, gene heskett wrote:
While I am saying that my results with earlier Samsung have been less than glorious. triple layer nand's turning into half capacity for instance.
There's simply no real value in looking at historic bad models as a guide to future performance (or the opposite). I can remember entire lines of hard drives from reputable manufacturers which were plauged by premature failures to the point that I replaced some multiple times under warranty before pulling them all (e.g., the IBM Deskstar 75GXP). I can also remember SSDs which had problems with repeated file corruption (OCZ Vertex, the only SSDs I ever saw reliably corrupt stored data). Bottom line is that sometimes you'll get a dud, and it doesn't really matter if you had a positive (or negative) experience with a superficially similar product decades ago. The Samsung 980 pros with the bad firmware were a ticking time bomb, but they haven't been sold with that version for years, and they haven't had issues since the fix. Other Samsung SSDs have been fine. The 860s have relatively low write endurance, but that's why they're as cheap as they are. You can either avoid using them in write-intensive settings and get a drive advertised for that role, or you can dramtically underprovision to lower the write cycle of individual cells and create space for caching. That's true for most low-cost drives, which is why they're low-cost, and why high-write-cycle drives are fantastically expensive. The average consumer will never write enough data to matter, but it is possible in pathological cases if something on the system goes nuts and starts sync-writing a really large number of small blocks.
