Flash Memory Won't Get Cheaper Any Time Soon 166
jfruh writes "Some melancholy news from the Hot Chips symposium last week: NAND memory, which powers the solid-state drives that have revolutionized storage, has broken the $1 per gigabyte barrier and isn't getting any cheaper. 'They will always be ten times the cost of a hard drive,' says analyst Jim Handy. There are newer technologies in development, but they won't be able to beat NAND on price for years."
No! (Score:4, Funny)
Oh first world problems.
Comment removed (Score:5, Insightful)
But there's a limit to what you need there (Score:4, Interesting)
You say that a high price on flash will hurt development, but when you can fit Wikipedia English into 9GB + 1GB space for the bzreader index file (a good chunk of human knowledge right there), what more do you need?
You need a maybe 1-2GB more for an OS (not Windows) with office suite, browser, some learning tools, dev platforms, etc. Give yourself and the OS some breathing room, and we're only up to $16 of flash. That's a whole lot less than a fixed disk, and you've still got several GBs free.
So I still don't see how this is much of a problem. You could push prices below $1/GB, but it would take a huge sea change (drop to $.25 or less) to make a real difference in the price of the device they are installed in. There's already plenty of storage for a reasonable price, if you're willing to forgo luxuries like porn :D
Re: (Score:2)
Re: (Score:2)
Re:No! (Score:4, Interesting)
Hate to break it to you but people have put crummy old spinning rust hard drives in computers that have been all over the planet, in space and under water for some time. Yes, SSDs are preferred these days, but it's not like ruggedized computers just appeared four years ago.
Hell, I remember field portables with FLOPPY DRIVES. And we liked them.
Re:No! (Score:4, Interesting)
I just bought an old Toughbook about a month ago with only a floppy in it. See, there's this software to program the cabin lights on a 747 that runs only on 95 or earlier and needs to produce a single-sided 3.5" floppy to insert in the aircraft. We have teams that travel the world overseeing cabin upgrades and I got tired of trying to get old Dells to live long enough to last more than one trip.
Re: (Score:2)
Tried that. This software seems to directly address the floppy drive.
Re:No! (Score:5, Insightful)
Oh goodie! It's the "Africa exists therefore you can't be dissatisfied with anything ever" argument.
Re: (Score:3, Insightful)
What exactly are we in the West supposed to do about it really? Play globo-cop? Encourage our governments to meddle in the affairs of other countries?
Really? What's the point of being fixated on other people's business?
Re: (Score:2)
I'll take a stab at this, since I think your point is an earnest one shared by many.
Is it playing globo-cop or meddling to fund and provide a treatment for severe acute malnutrition [wikipedia.org] that can be used by laypeople? And costs only $60 for a two month regimen?
Of course if you think saving people who are starving in a world of plenty is "fixating on other people's business" ... ponder this. Everyone is "other people". Your sainted mother, the wife and kids you love, the guy known to you personally who has been t
Re: (Score:2)
Take a look at the -1 rated parent of the guy you're talking about. That's who he seems to be propping up. He's the one who said the Africans are "too dumb to provide for themselves" and "not his problem".
Re:No! (Score:4, Funny)
BARRIER!? (Score:5, Interesting)
has broken the $1 per gigabyte barrier
It isn't a barrier. $1 is a COMPLETELY arbitrary value. Examples of real barriers are the sound barrier or the clock speed vs. power barrier (region) of silicon. A monetary barrier between low and middle class would be being able to pay for a new car with cash.
There has to be a solid justification to call it such. Otherwise, I could jump up and down SCREAMING that we have just crossed the 98 cent barrier.
A dollar a gig, cool! But no one crossed a real BARRIER.
captcha: barrier
Re: (Score:2)
Go google psychological barrier.
Re: (Score:2)
Really?
You must be using your own definitions of those terms. Either that or you're talking about a pretty cheap car.
Re: (Score:2)
I pay cash for my cars because of three things: 1. I don't buy extravagant cars; the last new ones were between $25K-$30K and the last used ones were half that; 2. As soon as I buy a car I start saving for the next one; 3. A windfall in the 1999-2000 dot-com boom gave me the initial large chunk of cash to start doing this (among other things).
I could have done the same thing even if that windfall had never come, but it would have meant less money into my 401(k).
All this presumed enough income that I actuall
Re: (Score:2)
> Not everyone has that, many live paycheck-to-paycheck,
And most who make good money live close to paycheck-to-paycheck or worse, in debt, meaning they've spent the paycheck before they get it. How many people have a loan, a debt, on a $30,000 car. They could have bought that in cash by starting with a $1,500 car, saving up for a $3,000 car, then a $6,000 car, etc. That would cost them a lot LESS than paying interest to a finance company.
Congrats to you for not putting a down payment on a $50,000 car
Re: (Score:2)
Re: (Score:2)
I also know a lot of people who willingly live paycheck-to-paycheck and blow hundreds at the bar on pay-day, then complain about money, but not everyo
Re: (Score:2)
> Probably because stuff that matters to nerds is often (not always) related to stuff you mostly find in the richest 20% of the world (population).
You mean people are most concerned about what's around them?
STOP THE PRESSES!
Re: (Score:2)
Probably because stuff that matters to nerds is often (not always) related to stuff you mostly find in the richest 20% of the world (population). Probably only 30% of that 20%, in reality.
If only 20% of slashdotters RTFA and approximately 30,000 RTFA (seems to be the common stats recently), that means there are approximately 150,000 active slashdotters, which easily fits within 6% of the world's population.
Oh, and "first world" ceased meaning anything useful when the cold war ended. You want "industrialized nation".
not sure what you're trying to say here, I'm sure you didn't mean that 150,000 is 6% of 7 billion.
Do you mind re-phrasing?
I am sure the "experts" are right... (Score:5, Insightful)
....having a perfect track record and all.
Re:I am sure the "experts" are right... (Score:4, Insightful)
"When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong." - Arthur C Clarke
Re: (Score:2)
Well, if you look at the article, the prediction is that flash prices won't fall much between now and 2 years from now (2015). Not the sort of prediction that you can really counter by exhuming Sci-Fi writers for quotable quotes.
Re: (Score:2)
can't be any other reason (Score:2)
Re: (Score:3)
I remember them saying this about regular hard drives (at $1/MB). I remember 5 of us going together to get my buddy a 512 MB drive for $499 on Black Friday. We beat the experts prediction!
Six months later, you could get one for $399, and by the next Christmas, for $199. So much for that prediction.
I am guessing that this one will end similarly. Somebody will have a drive for .33/GB on Black Friday.
Ya I seem to recall this same story from before (Score:2)
Some "expert" whining that flash can't get any cheaper because of fabs, limitations, etc, etc.
Well, I'm not buying it. Until I start hearing something from the people who actually make the tech, I'm going to say it'll probably keep going. Supply issues are just temporary. Companies can, and are, building new fabs all over. In terms of overall cost that has been getting reduced by both process size (which doesn't seem to be stopping soon) and by advances in how data is stored. Recently we've started to have
Re: (Score:2)
Well, I'm not buying it. Until I start hearing something from the people who actually make the tech, I'm going to say it'll probably keep going.
"This 'telephone' has too many shortcomings to be seriously considered as a means of communication. The device is inherently of no value to us."
-- Western Union internal memo, 1876.
"Radio has no future. Heavier-than-air flying machines are impossible. X-rays will prove to be a hoax."
-- William Thomson, Lord Kelvin, British scientist, 1899.
"There is not the slightest indication that nuclear energy will ever be obtainable. It would mean that the atom would have to be shattered at will."
-- Albert Einstein, 193
Re: (Score:2)
 "Nuclear-powered vacuum cleaners will probably be a reality within ten years."
-- Alex Lewyt, president of Lewyt vacuum company, 1955
Is anyone else here disappointed that one didn't happen bwcause I am.
Re: (Score:2)
Re: (Score:2)
Yeah, and that wire makes them way more boring.
Re: (Score:2)
"having a perfect track record and all."
They were right about CPU clockspeed. We've been stuck below 10ghz for a long time now.
Re: (Score:2)
Re: (Score:2)
While that may be true, it still cant' compete with the boost in frequency that performance got in the past. Technically a 3570K is 3 generations better then a core 2 duo, but it's barely 2x as fast. Usually you get an almost doubling of performance every generation. The performance gained from i920 to i2500 and to 3570 has been abysmal. Less then 40% increase from i920 to i2500, less then 20% going from i2500 to 3570. That is a huge deal.
Depends on your definition of "soon" (Score:5, Informative)
What the article actually says in the last paragraph is that there's currently a capacity shortage, that's expected to be resolved by 2015. The article also says manufacturers think they can go down another process node, and then do another 3 after that using 3D stacking. Then he says new technologies "with the speed of DRAM and the storage capacity of NAND" might make their way out of the lab next year.
Overall, the article's contents don't really seem to support the notion that it's game over for SSD capacity improvements.
Re: (Score:2)
That notion is off topic. The summary says that NAND "has broken the $1 per gigabyte barrier and isn't getting any cheaper." Since the article says prices will be flat through 2015, the article at least in the short term supports the summary.
The article actually says "Don't expect SSDs to ever get much cheaper." It suggets there may be two more process shrinks (16nm and unspecified).
Re: (Score:2)
'Out of the lab' and 'affordable' are several years apart.
This is where the $1/GB barrier comes in. It's not an option to come to market with a new SSD tech that costs more than $1/GB, unless there is another significant benefit.
NO: whoever is in the lab working on it, will be rushing to get out a tech that they can corner the market with for $0.50/GB.
The storage market is huge, and there is plenty of financial incentive to push the research and development along expeditiously.
Exactly (Score:2)
Spin 'em if you got 'em.
Re: (Score:2)
Why would you *not* compare them?
Re:Exactly (Score:5, Funny)
Re: (Score:2)
Why?
10X my white and flabby ass (Score:3)
Re:10X my white and flabby ass (Score:5, Funny)
Yeah, but Newegg will probably have it for $27K.
Re: (Score:2, Informative)
Speaking of newegg you can get a samsung 1TB for $635. That's quite a ways below the $1 per GB price point.
Re: (Score:1)
Google says you're wrong:
http://www.sabrepc.com/p-2521-fusion-io-fs6-802-640-cs-0001-512tb-iodrive-octal-multi-level-cell.aspx?gclid=CNfJ5vOVybkCFcU5QgodwkYAMg [sabrepc.com]
Ok, its $10K, but that's 5TB which you can't even buy right now. :)
Still, at the enterprise class, you're looking at ~$500, so its more like ~20x.
But its still apples to oranges as a PCI device will get you huge bandwidth. Different tools for different problems folks!
Re: (Score:2)
You might want to reexamine where the comma is placed in the price.
Re: (Score:1)
Who the hell needs 4TB in a single SSD? You can buy a multiple of smaller drives which total 4TB for a damn sight less than $29K. I don't think people making the price comparison are worrying about the extreme cases. For 'typical' sized drives of each type SSD is getting near $0.50/GB, and HDD is around $0.05/GB. That's close enough to 10x in my book.
Re: (Score:2)
> Who the hell needs 4TB in a single SSD?
Anyone that doesn't want to mess with an array to handle a simple use case of having a lot of stuff.
It's not 1988 anymore. There's ton of multi-media content out there. You can buy it or you can create it yourself. As tech and formats continue to improve and the "problem" only gets bigger.
Not everyone is a passive couch potato content with an anemic iPad.
OTOH, the price difference makes even the less extreme cases of a 1TB or 500G drive problematic.
Re: (Score:2)
Re: (Score:2)
Anyone that doesn't want to mess with an array to handle a simple use case of having a lot of stuff.
JBOD works fine with modern media players. For example, either Plex or XBMC (or even XBMC as a Plex client) works fine with multiple disks, even if they are spread across multiple servers, which might provide nfs, smb...
Re: (Score:2)
So in other words, stupid people. Right?
"Mess with" an array? How about hooking up 4 SATA and power cables to 4 drives, and engaging RAID software. Hint: it's a trivial amount of work.
Re: (Score:2)
So in other words, stupid people. Right
Or people who don't have a lot of time to mess around with unfamiliar concepts.
How about hooking up 4 SATA and power cables to 4 drives, and engaging RAID software. Hint: it's a trivial amount of work.
It's a trivial amount of work if you already know what RAID is and which RAID software to use and how to set it up and what the gotchas are and what not to do.
For example: if Joe Newbie wants to take one quarter of his files with him on a trip, does he need to bring all 4 SATA drives or can he just bring one of them? The answer is, of course, is "it depends", but the point is that if Joe makes a bad decision there he could end
for now, SSD is no faster for large files, though (Score:2)
For that multimedia you speak of, a rotating drive will be about as fast. A 10K rpm platter drive makes a lot more sense for video, which is sequential access.
When SSDs get faster for sequential access, then I'll be interested in larger. I don't see any need for many TBs of tiny files, and SSD is only impressive with small files. Very large databases are about the only use case I can think of for large SSDs, and maybe media laptops. Even with 40TB of data, I only want a 128GB PCIe SSD for caching.
Re: (Score:2)
Re: (Score:2)
That's a dumb statement. To an enormous extent, SSDs and HDDs are used for different but overlapping purposes precisely because they have a very significant cost difference. In many cases -- almost certainly most cases -- the cost determines which you get (either directly or indirectly), so comparing the cost makes complete sense.
Re: (Score:3)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re:10X my white and flabby ass (Score:4, Informative)
When you buy a 4 TB SSD, you're not paying $29k for the NAND. You're paying for someone to go through the trouble of amassing 4 TB of flash, design an arrangement with controllers which can address that huge amount, and produce it in bulk. Very few people are demanding that much capacity in an SSD, so the cost of that engineering and tooling work gets amortized over fewer customers. About $2.5k for the NAND, about $26.5k for the engineering and tooling.
With the lower capacity SSDs, those production costs are amortized over much larger volumes, and a much greater fraction of the drive cost is the NAND. A 128 GB Crucial M4 drive contains $80 worth of NAND (actually probably a bit more since there's some overprovisioning to substitute for cells which die early), and sells for $100.
Re: (Score:2)
Very few people are demanding that much capacity in an SSD
And particularly in a single SSD, for HDDs the price/GB goes down with size while with SSDs my impression is that you need to fill the channels on the controllers but after that it's just double the capacity for double the price. If you need a bigger SSD just get many and RAID-0 them.
Re: (Score:2)
Not only is the spot price for NAND $5 per 8GB ($640/TB), but by remarkable coincidence retailers are selling complete SSDs - damn good ones - at $635.99/TB [amazon.com]. The quoted price of $29,000 for a 4TB SSD is for a ludicrously overpriced high end enterprise drive. The only problem with Samsung selling a $2560 4TB SSD is that it way overshoots the psychological barrier of $1000 for one drive. They could clearly make one if they wanted to in a 3.5" form factor for a negligible investment, but not enough people woul
Re: (Score:2)
That's not true. IBM's retail on their 1u SAN module 26TB is $33k. You'll never pay more than $26k, and they might go as low as $21k if they love you.
Re: (Score:2)
A 4TB hdd can be had for roughly USD$200, or less. A 4TB SDD is USD$29k.
Apples to oranges. You're comparing a cheap consumer-grade HDD to a low-volume enterprise-grade SSD.
The 1TB Samsung 840 EVO SSD is currently going for $635.99 at Newegg. So 4GB of SSD storage would cost $2543.96 – less than 10% of the figure you quoted. So it's about 13x what the magnetic HDD would cost, not 10x – close enough.
The fact that you can't get more than 1TB in a SSD unit without paying insane enterprise prices i
Re: (Score:2)
>> A 4TB SDD is USD$29k.
>
> Not if your 4TB SSD is an array of 512GB SSDs.
For some reason I am reminded of the very first IBM hard drive.
Re: (Score:2)
I think most people have figured out when you when you mention "Crucial M4 512 TB drives" you really mean 512 GB, and similarly for "$750/GB" (really $750/TB). I'm not going to criticize the lapse. I've done it too many times myself.
Agreed 100% on the $40/TB for hard drives.
And SSDs are already at least down to $635.99/TB [amazon.com], and probably less if we look harder. In fact the Samsung 840 EVO is a pretty damn high end consumer SSD.
Inconsistent (Score:1)
"[NAND memory] isn't getting any cheaper" combined with "they will always be ten times the cost of a hard drive" could mean either:(a) both SSD:s and spinning drives will suddenly stop getting cheaper for no apparent reason or (b) Whoever wrote TFA and TFS are morons who doesn't realize that the first statement doesn't follow from the second.
I'm guessing (b).
Re: (Score:2)
For SSDs it's hardly "no reason". There is a limit in physics to the dramatic process shrinks we have relied on for a long time, and as we get asymptotically closer to that limit, price falls are going to slow. Or do you think we can exceed or readily even approach one transistor per molecule?
Already the raw cell write endurance has been falling precipitously as we have resorted to first two-level cells and then three-level cells. There is a limit to how much can be regained through overprovisioning. OK, ma
If these analysts actually hit the mark.... (Score:1)
We would all be stuck using 1.2ghz CPUs requiring exotic liquid cooling because we've hit the limit on die shrinks.
We would all be stuck with 500GB hard drives because there is no way to increase areal density of HDD platters
We would all be stuck on 1.5mbps DSL lines because there is no cost effective way to push data quickly over consumer grade circuits
We would all be on Windows Phone because MS was going to out innovate Apple and Google.
The doomsday soothsayers have been around forever and usually have ze
Re: (Score:2)
Few months later... 400gb links! We can't sustain that growth!
Few months later.... 800gb over a single fiber!
We can't sustain that growth!
Few months later.... you can now purchase 8tb/s over a single fiber and we have a working version running from Stockholm to Frankfurt with no repeaters using standard fiber.........
I give up
Re: (Score:2)
> We would all be stuck on 1.5mbps DSL lines because there is no cost effective way to push data quickly over consumer grade circuits
To a large degree, that's true. You can have the greatest digital signal processing on earth, but if you have a copper pair is of a length and quality that's right at the edge of what could do 1.5mbps ADSL back in 1999, the harsh truth is that you aren't going to do a whole lot better with a copper pair of equal length and quality today.
What changed was:
* the length of the
your first example is true (Score:2)
I had a 1 GHz CPU around 10 years ago. Right now I'm using a 1.2 GHz. Before that, CPU speeds would double every few years.
Okay, I cheated because my current 1.2 GHz fits in my pocket. I do have two machines with five year old CPUs that run 3-3.5 GHz, the same speed as a new machine five years later. So there ARE some real physical limits. That's why phones are dual core and servers have eight cores - because they couldn't make faster processors they had to join together more processors running at the
Crossbar (Score:3)
Re:Crossbar (Score:5, Interesting)
HP's memristor/ReRAM hasn't been mentioned in a while. That technology looks promising, and like the parent states, Crossbar has 1TB chips in testing. Does that mean there will be a USB flash drive with this technology? I'd not hold my breath, especially remembering how holographic storage was always just around the corner, from back in 1992 with a company called Tamarak to a few years ago with InPhase (well, their stuff is now owned by the state of Colorado, so who knows what state their IP is in...)
However, SSD isn't the be-all and end-all in storage. One can always make an array using battery backed up DRAM if needed and had the cash.
Slashvertisement (Score:1)
for YEARS (Score:2)
Oh, thank goodness (Score:5, Funny)
I was worried that Flash might stay expensive for a while, but now that an analyst is predicting it I know it won't actually happen. So, expect a massive crashing in prices pretty much immediately.
And the reason is... (Score:2)
... collusion and profiteering, maybe racketeering. I think the only reason Samsung produced TLC was to use it as a buffer to justify continuing to keep the prices of MLC artificially high. Hopefully other manufacturers, since they don't (yet) also produce TLC to compete directly with Samsung, will instead finally reduce their MLC prices to compete with TLC. There might be some sort of gentlemens' agreement preventing that, though, since Samsung's TLC can buffer the MLC prices for the entire industry, no
Re:And the reason is... not your reason... (Score:2)
And the reason is... not your reason...
The actual reason is that as costs go down, capacity goes up as a moving maxima on a bell curve.
It'd be cheap to buy the capacities of flash storage we have today in the future, but they won't be manufactured; instead we'll have much higher capacities at about the current price point.
The saddle spot where you get the-best-bang-for-the-buck will stay at about the same price point going forward as capacities increase. This is the same thing that happened with hard drive
The article is wrong. (Score:2)
Industry Disagrees (Score:2)
Ignorant statement - NAND is going 3D NOW (Score:2)
NAND is going to be 3d stacked, and it's going to at the very least provide another 10 years of life to NAND before resistive RAM or another technology finally takes over.
Even 1 single process tick (whether it be reducing size below 20nm, or stacking a layer of NAND with a 3D process) will bring the cost below the so called "$1 barrier".
"Samsung has big plans for future iterations of the V-NAND tech, including 3D chips with up to 24 layers, all connected by using "special etching technology" to drill down t
Re:What a scam (Score:5, Insightful)
Made by machines in $10B fab plants that need to be payed off before they are obsolete.
Re: (Score:2)
Made by machines in $10B fab plants that need to be payed off before they are obsolete.
Shhh ... Marx was pretty sure that all the means of production we'll ever need had been already developed and built.
Re: (Score:3)
Patents are a part of it, but they're minuscule compared to the capital requirements. Semiconductor manufacture isn't a basement hobbyist game; it's the absolute cutting edge of technology, and the people who make the machines that make the chips are creating custom, precision hardware for a very small customer base. Commercial-scale semiconductor plants run about $1 billion minimum, for a 10k-30k wafers per month "minifab" and can run up to $8-10 billion for a "gigafab" churning out 80k-100k wafers per m
Re: (Score:2)
Re: (Score:2)
And I'm sure all of the materials, and power are provided for free, too. And the design work. Not to mention the shipping costs.
Re: (Score:2)
If the margins are so low, why are the owners so rich?
Did you fail basic arithmetic in 4th grade???
Re: (Score:2)
You realize the companies you're talking about are almost all public? Of which the balance sheets and P&Ls are available in the 10Q/Ks filed with the SEC every quarter, fully viewable just a few clicks away?
If you want to make a direct difference, you can. Fire up Etrade and buy a few shares of Micron or any other tech company you feel is gouging or not fairly considering "human effort". Then you'll (literally) be one of the "rich owners" and can voice your concerns at their annual meeting.
Having sai
Re: (Score:2)
http://www.justice.gov/atr/public/guidelines/211578.htm [justice.gov]
There's even a contact address at the bottom of the page you can use to report the signs of collusion you've witnessed: antitrust.complaints@usdoj.gov.
Anyway, that's how it works in the grownup world.
Re: (Score:2)
Your original premise is that things should be almost free. When you reach the ripe old age of 18 and enter the real world, you'll slowly realize it's actually good that things aren't free.
And perhaps realize that society is composed of individuals, most of which are not all that bad. Then proceed on to actually enjoying life, and take the tribulations in stride.
Good night.
Re: (Score:2)
Except that charging more for semiconductors probably wouldn't land anybody in jail. So, your premise is very confusing. If somebody wanted to start a factory for NAND and charge more that what anybody else does, then they are perfectly entitled. It's not a very good business model, and they probably won't make any money because very few people would want to buy th
Re: (Score:2)
Re: (Score:3)
That's because of the efficent use of bits by the kernel. All non-random stuff fed in gets further sorted to zeroes and ones, and the ones inverted to zeros. Those zeros are then fed out of /dev/zero, while the random stuff goes out /dev/urandom
Re: (Score:2)
And it's web scale too!
Re: (Score:3)
Re: (Score:2)
2: Flash will always be 10x cost of harddrives. In other words, Flash won't overtake harddrives on price.
That's assuming that hard drives keep getting bigger and cheaper. The amount of R&D money required for each generation of improvement (in most technology) goes up, but the spending for HDDs has gone down as manufacturers see that they're hitting diminishing returns. The number of people who will pay for 4TB disks is lower than the number that will pay for 2TB, which is lower than the number that will pay for 1TB disks and so on. For a lot of users, even 500GB is more than they will need for the lifet
Re: (Score:2)
Most home users won't be writing nearly that much data to a 256GB drive before the purchase a new computer. New techniques will be solving the solve write-c