Scaling is not working for them like it is for Logic circuits.
Every body expects that making the transistors smaller and smaller will drop the prices....but there is a very nasty problem.
They keep reworking the controlers and wear intervals,...and spare rewrite area....etc....but it is not easy.
I don't claim what I just wrote is totally coherent....but it is a nasty issur with the physics.
I have been reading the technicals on the Flash memory and they haave got a problem....
Scaling is not working for them like it is for Logic circuits.
Yes, with regular NAND Flash which has many elements per flash “cell”.... I was hoping they could have figured out the MRAM, Phase CHange RAM, or Memristor tech for solid state storage which promises fewer elements per “bit” and would allow many fold densities over conventional memory tech.
But they seem to be a lot like Nuclear Fusion, always X number of years away....
“I have been reading the technicals on the Flash memory and they haave got a problem....
.
.
I don’t claim what I just wrote is totally coherent....but it is a nasty issur with the physics.”
Note that I said “solid state”, not “flash”. There are non-volatile memory technologies on the way that make flash look slow, unreliable and low-density.
That said, flash is absolutely there today as a workable improvement over magnetic hard drives. For most users (and by that I mean those not running enterprise databases with large write loads) current SSDs are a fine solution. I recommend the Intel 3/500 series, which include a five year warranty:
Almost no home users will even come close to the write limit. Five years is plenty of lifetime, and there’s nothing to say that they won’t keep on ticking for long after that.
You need backups regardless of the local storage technology you’re using. Both SSDs and magnetic drives can fail, usually at the least convenient time.
Personally, my plan for my next system I’ll get this year is to use Raid 1 (mirroring) with multiple 240 GB SSDs. That also provides a ~2x read throughput performance improvement. There is no write penalty as with spinning disks, since there’s no rotational latency. As with traditional drives, I’ll buy my five SSDs (four for use, one as a hot-swap replacement) from different lots.
Backups will still be needed, but system availability should be 100%.
and (dangerdoc):
“I’ve had much better luck with spinning discs with most of them lasting longer than their usefullness. Solid state has a limitted number of read/writes and if used frequently can have a significanly shorter lifespan than a spinning disc.”
Flash (not solid state in general) has a limited number of writes, unlimited reads. Since most computers do far more reading than writing, this generally works out well. For me, five years is a long horizon on computer equipment. Most often by then performance has improved so much the old system is moved to some secondary role, or given away. I expect even with my fairly heavy system use that SSDs will easily last five years worth of write cycles.
It’s unfortunate that even the enterprise class SSDs don’t include modular controller electronics. The manufacturers don’t want to spend the extra $1 or so on a connector (and take a very small hit on reliability). If they had it, you could order a $10 replacement card that would very likely restore access to your data.
SSDs don’t have the mechanical failure modes that plague rotating drives, the most common being head crashes, drive motor failure and head servo failure. It’s true that you can spend thousands and sometimes recover some data, but if you’re doing proper backups you can save all that money, time and hassle. Plus, in the case of a head crash, some data will simply not be recoverable. Back up your data! ;-)