I'm one of those weird data horders that still buys high capacity drives. I take a lot of RAW photos and videos and I prefer to buy and not rent a license to my music (and I like using sites like Bandcamp to support independent artists). I recently got a 14TB as a new primary, 10TB secondary, along with similar capacity backup drives.
I hope this trend doesn't mean consumer hard drives are going to get more expensive going forward. I keep reading about people predicting SSDs are going to hit the same price/capacity as spinning disks, but I just don't see it with the large >10TB volumes any time soon.
> I keep reading about people predicting SSDs are going to hit the same price/capacity as spinning disks, but I just don't see it with the large >10TB volumes any time soon.
To elaborate on this: $/TB varies widely depending on total drive capacity. The fixed costs of hard drive spindle motors, actuators, etc. are higher than the costs of an SSD controller and PCB. Adding a platter to a hard drive is cheaper on a per-GB basis than adding more NAND flash to an SSD, but there's a crossover point below which SSDs have cheaper price per GB and small hard drives are no longer economical to manufacture. That crossover point has now risen to encompass the majority of consumer use cases, but it is nowhere near 10TB and won't be anytime soon. That crossover point also seems to be moving upward much faster than typical consumer storage needs, so we can expect this segment of the hard drive market to continue to wither.
Pay a little more attention to the hard drives you're looking at. Almost all the product listings for tiny hard drives are for models that are several generations old, usually with branding that the manufacturers have long since retired (WD Caviar, Seagate Momentus). The cheapest listings are often grey market sellers of old stock with no manufacturer's warranty.
Seagate has abandoned all capacities below 500GB, and they offer limited options for form factor and RPM at 500GB. Western Digital still lists one 250GB model and two 320GB models, but none of them are cheaper than a 480GB-class SATA SSD.
Current 500GB hard drives start at $38-40, and 480GB SSDs start around $48-50. These capacities are sufficient for typical consumers.
Most of the consumer use cases for 2+ TB are well served by external drives or a NAS, which aren't part of the market segment that Nidec is predicting to crash this year. Backups and movie/photo libraries don't need to be on local internal drives. Large video game libraries just don't account for very much of the PC userbase. Every PC should have an SSD as primary storage, and with 512GB drives down to $50, very few PCs need any other internal storage.
It's going to be a vicious cycle as well (or virtuous depending on your point of view).
As demand for HDDs weakens, investments in R&D will be scaled back significantly. HDDs are in trouble even if you assume they'll follow the same capacity ramp they once did. If they slow down, they're toast.
(Although SSDs may also hit a wall scaling down; the low hanging fruit may already be picked.)
We're already past the point where hard drive R&D is driven solely by datacenter needs; nobody was working on 14+ TB hard drives with the expectation that consumers would be a non-trivial part of that market segment, except indirectly through cloud storage.
Typically improvements come in the form of the density of the bits you're storing on a platter, which is just as relevant to a single platter or five (or even seven platter helium drives).
SMR is an example of something that's relevant to all datacenter use, but only perhaps half of consumer use. And of course helium itself is pretty much datacenter focused only.
The linked article has the datacenter share last year as 13.6% of all units (but growing shipments). Looking at Seagate's last filing, they had $601m gross profit last quarter, and an R&D spend of $238m had to come out of that.
I don't see how you scale the company down to a fifth of its size and maintain anything like the R&D velocity it once had; and once it slows down, the band of competitiveness of HDD over SSD even for datacenter use is going to continuously narrow.
A complete collapse of the HDD market after a decade is not unthinkable, unless SSD progress hits a wall.
(Of course Seagate does make some SSDs, which I'm not accounting for, but taking the time to separate it isn't going to meaningfully change the picture.)
Nidec divides the hard drive market into 5 segments: PC, External, Consumer Electronics, Datacenter, Enterprise. One of those five segments is crashing. External and Consumer Electronics are contracting slowly but will still be a cash cow for a while because those drives are still selling but don't require much if any R&D. Datacenter and Enterprise are basically the same for R&D purposes, and together are growing slightly.
There's a big difference between Seagate suffering the loss of (half of) one of those five market segments, and scaling the company down to a fifth of its size. Yes, they'll have a harder time funding R&D, but they're quite a ways off from having to drastically slash their R&D budget.
The good news is the mid- to high-end of the hard drive market will be the last to go as SSD can't compete there (yet). The bad news is that in the short term I would expect progress in mechanical drives (both cost and absolute capacity) to slow even more than it already has over the last decade as the market opportunity will warrant less investment by storage companies. By the time the end game plays out, there will be cost effective alternatives. (whether or not they'll be available to consumers is a different question)
Look at it this way: the optical drive/disc market has been 'dead' for years but you can still pick up a new DVD burner for ~$20. Hard drives will likely follow suit: at some point it won't make sense for hard drive manufacturers to invest in developing new drives, but they'll keep churning out the existing stuff for years and years until demand dries up. They've already done the expensive part (developing the drives, establishing supply chains, building the plants and setting up the manufacturing lines), and will keep doing the cheap part (actually making them) as long as there's a market.
Yes, I recently looked up the BR burner and discs I bought circa 2010 for large backups of photos, etc. Drives, capacities, speeds all about the same, which was mildly surprising.
> I keep reading about people predicting SSDs are going to hit the same price/capacity as spinning disks
Flash doesn't actually have to reach the same price/capacity ratio to wipe out the small end of the HDD market. The performance and solid state nature of flash provide so much value that few will opt for a HDD when they can have a flash device, even if the flash device is smaller and a little more expensive. Also, at some point not having flash IOPS will become intolerable.
Large disks will survive while they have the price/capacity ratio advantage, but the small end of the HDD market is dead. And yes, I do believe this means that the price floor of HDD will climb as the number of lines building HDDs decreases.
The small end is why I'm still buying (or at least was a few years ago) machines with spinning disks. Even now looking at the surface range, the surface go has all the computing power I need but is hobbled by a 64GB SSD and jumping from that at $499AU to the 512GB I'd consider the bare minimum you have to spend $2,879.00AU and upgrade several other components I don't care about.
I think it was last year, after ordering a 2TB SSD, that I realized I've probably already bought the last mechanical hard drive of my life. I could go for one more round of 10TB's or something to max out my Drobo, but by the time those seem small, I do expect SSDs to have overtaken them.
"No way, Grandpa, magnets were never FREE!"
"Well, not free exactly, but we'd harvest them out of our discarded hard drives! Those were a sort of data storage machine, you've probably seen pictures."
"But when you removed the magnets, didn't that mean forfeiting the rare-earth-element deposit?"
Not if you care about stored data. We already had Samsung drives starting to forget contents after few months with a firmware patch forcing periodic rewrite. SSD (using flash technology) will never replace HDD, unless you dont care about retention. Something else past SSD (phase change etc) might eventually reach that goal.
I do some video editing sometimes. I tried to archive one of my projects to Blu-Ray (all of the clips, video files, etc. etc.). Storing 300GB in 6 different BluRays means 6-points of failure.
And more importantly, BluRays have a 20MB/s read/write speed. Modern Hard Drives read/write at roughly 200MB/s.
A 5TB Hard Drive is just far more convenient, faster, and reliable than stacks-and-stacks of BluRays. A BluRay is great for the "50GB of important data" (Tax returns, resume, other important documents).
Having one 50GB "highly reliable" source of storage is great. You can clone that one disk, leave some at your parent's house / some other house you trust, have 3x disks for redundancy at home, etc. etc. Its the easiest way to achieve high-reliability.
Oh, and leave a copy in your 5TB Hard Drives too, for convenience sake. But BluRays are cheap for copies and long-term reliability. But they're otherwise awful to work with in my experience.
LTO-7 tapes in the 6/15TB range are between $60 and $80. LTO-8 tapes in the 12/30TB range are about $199 though seem largely on back-order. Double-capacity (24/60TB) LTO-9 tapes are planned.
Which means you need to store over 30TB to 50TB of information before tape drives are more cost effective. I've always wanted to make a tape-drive computer, but each time I run the costs... its just completely unfathomable that I'd ever store the amount of info needed for it to be cost-effective.
If I were to build a big storage device for my house today, it'd be Nas4Free system with 6TB Hard drives (avoid shingled drives, PMR for cost-effectiveness, Helium is unnecessary if you stay below 8TB). 6x 6TB Hard Drives with RAIDZ2(similar to RAID6) for a total capacity of 24TB of storage / 2x parity drives for resiliency.
Sorry, I wasn't recommending tapes, I was just providing the $/GB as requested by the parent as a point of reference because I had no idea what it was until I looked it up. I figured it'd be interesting to others too.
Plus you need to physically manage the tapes, it is hard work. The alternative being to provision a bunch of hard drives in a NAS and from there it is all software/automated (up to a certain capacity).
Random access times aren't great -- to say the least! However, raw bandwidth isn't bad, LTO-7 gets 750MB/sec (faster than SATA3) and LTO-8 looks to improve on that.
blu-ray dual layer user here -- ~1 hour to burn, needs many to archive some ordinary "Big Enough" data.. yes, use them, and, plurality is the key to longevity.. use multiple, different backups with redundancy
Consumer hard drives are already getting more expensive in some cases. I bought a 8TB shingled magnetic recording[0] hard drive about two years ago, and recently tried to buy another one but I couldn't find any in stock. I bought a more expensive non-shingled drive instead. I assume nobody sells shingled drives to consumers anymore because people are too stupid to understand what it means and think the very slow random write performance is a defect.
Recently I've joined the home NAS crowd. I never thought I would but it just happened naturally. I was rewatching the GoT BluRays to prepare for the new season and it was very frustrating having to change discs every 2 episodes, taking forever to load a disc, printing the same stupid copyright warning every disc. So instead I went and ripped them (funny that the copyright message saying not to copy them was a factor in convincing me to actually do so). But then I started running out of disk space so here I am a few weeks later using RAID5 and multiple drives exposed through the LAN and having ripped my entire BluRay library.
It's actually quite cool in that I now get both the benefits of streaming (can play any ripped title on demand whenever I feel like it) and none of the drawbacks: doesn't need Internet nor does it use my monthly traffic limit, it's at much higher quality than any Internet stream (GoT BluRay episodes average at around 35mbit/sec while an Internet stream doesn't go above 5mbit) and there are no "concurrent streams" limit (well, other than the one inherent to the speed of my LAN/NAS but I've yet to hit that one).
From another thread: http://serverbuilds.net and associated forum/chat is an active community for affordable DIY NAS with used enterprise hardware and FreeNAS, Unraid, ESXi, etc.
My guess is that demand for hard drives less than 3TB+ is going to dry up and production will end. There isn't much point in small hard drives any more; at the same price people would rather cope with less flash storage than more HDD.
Agree. On the other hand there should be a long term appetite from large capacity hard drive from datacentres. The flow of data being captured isn’t reducing.
Hard drives are very nearly dead for hot storage. I expect them to by wholly dead by the end of 2020 because the TCO of HDDs is going to outstrip that of SSDs because of the fundamentally lower reliability.
My personal, offhand, experience is that SSDs tend to fail about 2 or 3 times sooner than a hard disk.
My anecdotal experience matches yours. In addition, SSDs often fail catastrophically. Total loss of data, not just some bad sectors. They're worse than hard disks in that regard.
The failure rates for SSDs I've seen when I've seen them compared to HDDs are lower. So long as you aren't using QLC in certain servers. Although there is indeed less warning of impending failure.
Use case also seems important as HDDs are relatively adversely affected by mobile usage.
>as NAND tends to lose data integrity as data is repeatedly read out
This doesn't tend to be an issue because the controller can automatically rewrite the cells as needed. What is an issue is if you leave the SSD unpowered for a few months and the cells leak enough charge to be unreadable.
I suspect most SSD firmware actually does automatic old-age scrubbing of somekind within the firmware. This theory seems likely confirmed with your offline decay theory.
Makes sense. SSDs are bigger and cheaper than ever. Not as much need for spinning metal drives. Not that the need is gone, but most people can probably get by with a single large SSD where before they need an SSD and HDD to store media.
For my pro-home solution (audio production), Dropbox solves all my backup needs. All my work folders are in Dropbox, and if my hard drive explodes, at most I'd lose a dozen of minutes of work.
For this reason I don't consider sync services to be a "backup". Because both sides always want to "sync" with each other, it's way too easy for a fuckup on your local machine to wipe the remote copy, or the remote server to wipe your local copy.
It's simply not needed. A majority of PCs have battery backup, and everyone uses journalling filesystems. The risk of large-scale data loss from a power failure is much smaller than the typical consumer cares about, and power loss protection capacitors make a noticeable difference in BOM.
And that's all that consumers need to have protected. Preventing catastrophic data loss is adequate for consumers. Consumers don't need to ensure complete transactional consistency for writes that are in-flight at the time of an unexpected power loss; they are conditioned to expect minor loss or corruption of data that is actively being modified when power is lost. Having at most a few MB of data at risk during the rare slices of time where a consumer PC is actively writing to the disk is an acceptable risk for these use cases.
This is interesting, but is it really as simple as "laptops are now using SSDs over HDDs"? I'd like to see some more detailed analysis of this prediction.
As recently as 2 years ago, Lenovo tried to sell me a laptop with a mechanical HDD. They actually succeeded -- their price for an SSD was insane -- but I ripped it out and replaced it immediately.
Today a mechanical HDD isn't even an option. The new generation of the same machine is offered only with a PCIe SSD.
That's just one point of anecdata, but I think it's a pretty stark one.
I think another factor is that HDD that are included in computers are larger by default. So instead of shipping with a 256GB HDD and forcing the customer to buy another one later, it ships with a 2TB HDD.
I hope this trend doesn't mean consumer hard drives are going to get more expensive going forward. I keep reading about people predicting SSDs are going to hit the same price/capacity as spinning disks, but I just don't see it with the large >10TB volumes any time soon.