Slashdot Mirror
Why Aren't SSD Prices Going Down?
Lucas123 writes "NAND flash memory makers took an economic beating from 2007 through the first quarter of 2009 due to supply outstripping demand. During that time, solid state drives dropped in price 60% year over year. But after the economic meltdown, fabricators pulled back on production and investment in new facilities and the price of SSDs have remained flat or increased over the past year, and that is not expected to change until 2011. Until that time, SSDs remain 10x more expensive than hard disk drives. SSD vendors, however, are using a few tricks to get sales up, including selling lower-capacity boot drives that hit a sweet spot in the techie/gamer market."
It's going to be really really hard to convince me that Asian electronics manufacturers aren't engaged in price fixing en masse against the rest of the world whenever a technology cost remains unnaturally high. Hell, after realizing how many times I was the victim of it with LCDs I pretty much expect it.
I mean, really, I feel like a moron for ever knowing that they allowed price fixing -- even promoted it -- inside their borders and then believing that stopped at the rim of the continent. Right now the only question is how many markets is this happening in? They're obviously very good about it, little chance the regulators in other countries will catch it let alone the easily bribed authorities isntalled there.
My work here is dung.
SSD cost is limited by the cost to refine and turn Silicon into Flash Memory.
The price will only go down as the process size goes down, currently at 32nm with Intel's Latest drives. Once it reaches 8nm or the like then the cost will truly be comparable to Hard Drives. Until then, don't expect a miracle.
The same 60GB drive I paid $230 for 6 months ago is now $130 after rebates and $160 before.
Thanks to everyone shutting down factories during the recession, there is a severe shortage of analog parts, electrolytic capacitors, and some FET's. It is typical to see a 4-8 week lead time on an order of 20k. A 16 week lead time makes you Very uncomfortable and you start looking for second sources or redesigns.
Some analog/digital companies are shipping at 16-24 week lead times.
Some electrolytic capacitors are at a 40 week lead time.
And at least one major company stopped accepting new orders.
In the mean time, some distributors are starting bidding wars on parts that they do have.
Right now, demand is far greater than the supply. It is going to be at least another year before prices start to come down.
While the question of whether prices aren't being competed further down because of collusion, or just because of inescapable production costs is an interesting one(and hopefully somebody has their forensic accountants on it, just to be sure), it seems reasonably obvious why SSDs have settled into the niche that they have, and why the manufacturers are making the size/price decisions that they are.
Now that the initial round of epically bad JMicron controllers are mostly gone, and the boring Samsung reference ones are confined mostly to build-to-order options on corporate laptops, all but the ghastliest SSDs are embarrassingly superior to HDDs for the sort of random mixed read/write that makes such a difference for desktop responsiveness. At the same time, though, nothing short of alien nanotech is going to allow them to touch HDDs in price/GB. That being so, you would expect to see SSD capacities largely cluster around "enough for a Windows boot volume, with a few key applications on it; but not much more". Anything less is largely useless to the target market(or, more accurately, anything less is aimed at the embedded devices market, and probably uses entirely different connectors and isn't sold at retail) and anything more gets very expensive very fast. This is, also, the reason why a lot of the high capacity (512GB to 1TB+) SSDs that you see are actually 2 or 4 of the vendor's lower capacity boards stuck together behind a cheap RAID chip. The market for the super high capacity ones just isn't all that big, at least among systems that use SATA as a storage connection bus, so the high capacity drives being sold are practically low-volume engineering samples, just polished enough to be sold for the usual early-adopter premium.
The only real forces supporting the existence of SSDs larger than that are high-end laptops(if you only have one drive slot, you can't adopt the mixed SSD/HDD strategy), a few loony enthusiasts(if you are the sort of person who buys every highest-end video card on release day, you can probably be convinced to go for a couple of 512GB SSDs, in RAID of course, for your gaming machine) and some truly titanic databases run by the deep-pocketed(though it isn't clear how much of that is SATA connected, and how much is the directly PCIe attached stuff, which is even faster).
Whether he is right or wrong really depends on how large a disk you want.
Barring truly revolutionary advances in silicon device fabrication, and(I'm not sufficiently up on my physics to know for sure) possibly a change in physics, sputtering a thin metallic film with the appropriate magnetic properties will always be cheaper, per square centimeter, than fabbing a complex integrated circuit. Further, it is quite likely that the smallest possible magnetic domains will continue to be smaller than the smallest possible flash cells, so you get more bits per square centimeter, and you pay less per square centimeter with the magnetic stuff.
However, as you note, Flash is pretty much ready-to-go. Virtually all the cost is the silicon. Packaging and soldering are relatively cheap(and, since every HDD also has a controller board, both technologies pay the "assemble a PCB" cost). With magnetic storage, though, whether HDD or tape, you have to build a fairly complex and expensive machine to enclose the cheap magnetic medium, and read/write, and keep dust off, and so forth.
If you adopt the naive strategy of comparing each technology's "sweet spot cost"(ie. the cost/GB of the device with the lowest cost/GB of each tech), I suspect that Mr. Unsworth is correct, if not forever, at least for a long time. However, a great many applications don't actually care about that metric. If you know how many GB you need, you don't care about "what is the lowest cost/GB?" you care about "how can I most cheaply get X GB?"
In the case of HDDs, the "sweet spot price" is somewhere between 5 and 10 cents a GB. However, the sweet spot is measured in 1-2 TB devices. If you, say, only needed 20GB, you would be unable to find anybody to sell you a 20GB drive for $1-$2. A quick look a newegg suggests that the cheapest retail HDDs are around $30-$35. You do get 80GB to 160GB for your $35; but you basically can't spend any less. The cost of a machined housing, hiqh quality spindle motor, packing, shipping, etc. just make that impossible. For the same $30-$35, retail, you are looking at around 16GB of flash(less if you want AES encryption and stuff, a little more if it is on sale). Thus, for any application that needs 16GB or less, SSDs are, in absolute $/GB terms, actually cheaper than HDDs(in addition to their other virtues: quiet, low power, shock resistant, small size, etc.)
I suspect that, over time, HDDs will be cheaper than SSDs in "sweet spot price" more or less forever; but the capacity(currently around 16GB, was more like 8GB the last time I wrote something like this) below which the absolute cost advantage lies with SSDs will continue to creep up. If it manages to creep up faster than software bloats, we may reach the dramatic tipping point where an SSD is cheaper, as well as better, than an HDD for the boot volume of a "normal computer", as opposed to just embedded systems, the occasional netbook, and space/power constrained devices.
Someone is buying massive quantities of them. That's the only explanation. The enterprise is only now catching up to the benefits of low latency SSD access. You can use fewer servers and serve data with much better latency and throughput for IO bound tasks. Anyone who needs low latency random access to data (ads, search, data warehousing, OLAP, content distribution networks, hotspots in map data serving, etc, etc) are switching to SSDs right now as quickly as their budget allows.
I think your facts are out of date. The latest SSDs beat hard drives in every category, sequential reads/writes and random reads/writes. SSD random write performance has been superior to the fastest hard drives for quite a while now. Performance is even better with TRIM and 4K alignment in Windows 7. It is sequential write performance that has typically been weak, but even that is no longer the case.