Slashdot Mirror
Hitachi Predicts 3D Hard Disks by Year's End
daria42 writes "Hitachi has announced that its perpendicular, or 3D, hard disks should be out by the end of 2005." From the article: "Today, hard drives record and store data in a longitudinal fashion, with the read/write heads scanning over a horizontal plane. In perpendicular recording, data bits are aligned vertically, allowing for more data to be squeezed into a finite area. Put another way, data will go from being stored on a two-dimensional XY grid to living in a three-dimensional XYZ space."
Hitachi will actually come out with drives that employ perpendicular-recording techniques toward the end of this year...
So, it looks like it is finally happening for real...
I wouldn't see why not. You might have to put more logic on the drive to so but, hopefully that wouldn't reduce performance at all. Same idea as your system seeing a half dozen IDE drives as a single large drive with an RAID controller card, I suppose.
:)
And if not, it won't matter anyway. By the time these drives are released, the bugs worked out, better versions released, and then price reduced to an affordable range, they'll be making motherboards with whatever new bus interfaces is required.
All I know is, we've come a long way from punch cards or casette tapes.
pressesc (873084) Answered this in his post above. There arn't multiple bits encoded in the same domain, its just that the domains are arranged permendicular to the disk. Its like this, lay three pencils down on the table, end to end. Thats three bits. Which way the points point determines wether each pencil is a 1 or 0... Now, stand all the pencils on end, perpendicular to the table, but still next to each other. The way the pencil points still determins the bit value but you can fit a hell of a lot more pencils on you table if you stand them all on end. I hope this helps someone. opoliges fro teh speeling. Pete
Thats also wrong. See here http://hardware.slashdot.org/comments.pl?sid=14487 1&threshold=1&commentsort=0&tid=198&tid=126&tid=13 7&mode=thread&pid=12132123#12132157
Each bit on the hard disk is represented by a small area of magnetized particles (like lots of little bar magnets).
With the longitudal system, the particles are magnetized so that the North and South are both on the surface of platter (bar magnets lie flat on the surface).
ie. <N-S> <S-N> <S-N> <N-S> <N-S>
With the perpendicular system, the particles are magnetized by a field that is perpendicular to the surface (bar magnets point up or down) ie.
^ ^ ^ ^ ^
N S S N N
| | | | |
S N N S S
v v v v v
Obviously, this has the potential for increasing storage capacity.
Vintage computer adverts: http://www.vintageadbrowser.com/computers-and-software-ads
It has to do with picking up different data as the pickup moves up and down while hovering above the platter.
.5 Moves out a little bit and picks up a wider magnetic feild, like .7 and then it moves father out and picks up 1.2 wide.
envision alternating magnatic feilds of strength.. so if the head is realy close to the platter it picks up a different signal vs if it's farther from the platter.
Like with data recovery that is done on conventional drives.
IF you overright your drive with Zeroes, it is still possible to pick up the old data on the harddrive. This is because each new line of magnatic data that is laid down smears accros the platter after a while.. So say you lay down a line of magnetic polarity and that width that you lay down is 1.0 unit of measurement.. After a few days it would transfer magnatism to the surrounding materal and end up being 1.2 units of measurement wide..
So if you lay down another track, that is 1.0 units of measurement wide.. But with sensitive equipment you can still access the old data that is 0.1 wide on either side of the new data.
So you use that to do something like this:
The head moves very close and only picks up the very center of the magnetic information, like
You see the closer the head gets, the tighter the focus of it's magnetic feild will be. The farther away from the head it the more area it picks up.
Right now drives have a set height at which the head sits. It doesn't move up or down, with these new '3d' drives the heads can pick up information from a veriaty of distances away from the drive.
So you can set it up so that it gets either the same charge, or less, or more of a charge as it moves up and down. Then it's mearly a matter of figuring out how to encode the information..
OK, I found the article and headline a little short on hard facts, so did a quick search for a better explanation. You can find that here:s _5_23/ai_103731260
http://www.findarticles.com/p/articles/mi_m0BRZ/i
The alleged move to 3D is something of a red herring.
It appears that with current longitudinal technology, each bit is encoded by a magnet with a North-South axis that lies in the same plane as the platter itself and occupies some 100 grains of the magnetic material. The novelty here is that in perpendicular recording, the magnet is stood on end with its North-South axis perpendicular to the plane of the platter.
Apparently this theoretically leads to greater areal densities of data exceeding that of the longitudinal technology. This is where the win occurs.
In particular, what initially confused me is that we are not talking about multiple layers of data within one platter. There is still only one layer of data per side per platter, but we have achieved greater areal density of that data. Exactly what that density will be once these drives are in production is anyone's guess.
Any help?
no time, no sig
http://www.xbitlabs.com/news/storage/display/20031 014160044.html
http://www.seagate.com/cda/newsinfo/newsroom/relea ses/article/0%2C1121%2C1555%2C00.html
Toshiba announced "perpendicular recording" technology in 2004 with a scheduled release Q2, not late this year or next year. With a much better description of how "3d" perpendicular recording works.
Actually, you are completely wrong. This article is pretty bad at actually explaining what Perpendicular recording is, so here it goes. Normal drives magnetize a certain area of a very thin layer of magnetic material on the surface of the platter. This means each bit has a certian area. This area has become so small that to make it any smaller would mean it would be too weak to actually be read. So the solution is to magnetize the media in the third dimension, 'into' the platter. This allows the bits to take up less space and still be strong enough to be read. Actually being able to 'stack' bits like you think would REALLY increase storage capacity! And Thers even a technology to come in after Perpindicular recording has ran out of steam called laser assisted recording, where a very weak lazer heats up a tiny spot on the drive, making it much easier to record. But again the spots are so small that they are difficult to read. Personally I hope that the IBM Millipede tech matures to the point where it can replace hard drives. Or maybe a rewritable version of those Holographic discs
The short answer is that it'll work, but the reason is that in the meantime we've taken an agnostic approach to accessing drive contents.
A long time ago, in a galaxy far away, we had MFM and RLL drives which (A) required the controller to have a pretty intimate knowledge of a drive's internal workings, and (B) an access scheme that again was tightly coupled to the drive's geometry. It was in fact an addressing where you had to explicitly state the track, sector and head. So if you moved to some other scheme (e.g., adding a 4'th parameter: depth) it would fall flat on its face.
In the meantime, though, technology got smarter. Both problems got solved as follows:
A) IDE (Integrated Drive Electronics).
The industry basically moved away from having dumb drives and a controller that needs to know the exact internal workings of the drive. It took a lot of hint from SCSI. Nowadays the real controller is on the HDD itself, and the "IDE controller" on the mobo is merely a bridge to the specialized bus to commnicate with the real controller.
That's why nowadays you can have CD-ROMS, DVD-burners, etc, on an ATA ribbon. Or why you can have cache on the drives nowadays. Or why you don't have to buy a new motherboard each time a HDD vendor comes up with a new encoding.
So the short story is that as long as the drive comes with an ATA or SATA compatible controller in it, it will work.
B) LBA (Logical Block Addressing)
The addressing scheme also got more agnostic. We no longer tell the drive the exact track-sector-head coordinates. We just tell it "give me the 1075'th sector" and let the drive figure out for itself where that sector is. (That's another point where IDE comes in handy.)
So the short story is: as long as the sectors can be numbered, any geometry will work. Adding an extra dimension just means you'll have to number the sectors differently. But as long as you can number them, you're all set.
(Of course, this is assuming your drive doesn't end up bigger than 144 PETAbytes, which is the limit for 48 bit LBA with 512 byte sectors. If it's more than that, well, we'll have to switch to using more bits.)
A polar bear is a cartesian bear after a coordinate transform.
So
distance - - - - - - - - >
N S . S N . S N . N S
is now shorter
- - - - - - >
N S S N
S N N S
(Lameness filter encountered. Post aborted!
Reason: Don't use so many caps. It's like YELLING.)
The pick-up head has inertia due to the mass of a hefty ceramic magnet and several hundred turns of copper wire. There's a counterweight balancing it so that there is only a couple of grammes' weight bearing down on the record, but it has a hell of a lot of inertia compared to the steel shank of the stylus, which is attached to a very flexible coupling. So when the groove pulls the needle to the left, the needle moves left but doesn't take the whole pick-up head with it; the magnetic flux lines change and induce a current in the coils. The preamplifier has a relatively high input impedance, so the needle isn't actually doing much work generating electricity. Otherwise it would feel stiffer.
Side-to-side motion is the sum of left and right signals. Up-and-down motion is the difference. By using four coils, not two, and pulling cunning stunts with the wiring, you can create a sum and difference of the sum and difference signals without resorting to op-amps. Which, of course, gives you {more or less} the original signals
Je fume. Tu fumes. Nous fûmes!
The marketing droids turned "perpendicular" into "3D" to increase the hype level.
Yeah, they'll do that..
This advance will probably only give an incremental improvement in density. Sigh.
Well, it's a pretty big increment. It could be as much as 2X.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
You obviously don't do non-destructive video editing. My laptop's hard drive is 80GB, which is enough, and usually has around 15GB free. I have an external FireWire 800 drive I use for video editing. It is 320GB, and almost full. I would welcome being able to get a terabyte or two in the same space.
I am TheRaven on Soylent News
Perhaps instead one ought look toward using linux with a modern filesystem that works to prevent fragmentation in the first place, like ReiserFS.
Akarsz Magyar Gentoo fórumot? Akkor
On top of the applications, I like to RAID just about everything and backup critical data to secondary machines once in a while. I do this because backup technologies (Tape/DVD etc) have not kept pace with hard drives in terms of cost and capacity and hard drives do fail from time to time.
RAID is a good "continuation-of-service" solution (i.e. you stand a good chance of being able to continue using the system during a failure), but IMHO it's absolutely no substitute for a backup. For one, the RAID applies changes immediately - so a slip with the rm command and you've lost your data immediately instead of being able to get it back from last night's backup. Secondly, if the PSU goes boom it could take out all the drives in the RAID (I've seen it happen).
http://blog.nexusuk.org
When the data density of hard drives increases the data transfer speed increases. .006 sec to spin that track past the read/write head (that's 10,000 rpm). This means that the data transfer rate from the track is about 1/6 MB/sec.
Imagine one of the tracks on the platter. Suppose that the track contains 1KB. Further suppose that it takes
Now, double the density of data on the platter. This would make 2KB in the same track, increasing the transfer rate to about 1/3 MB/sec.
(Historically the read/write sensitivity, time required to convert the signal to true binary for the computer, and distance to controller card have been speed bottlenecks. However, I think that the current bottleneck is the data transfer rate from the platter to the read/write head.)
Exam 4/C again. Maybe I'll do better this time.
Normally on a disk the magnetic fields run flat along the top, like what would happen if you lay a series of pencils down on a notebook. Some of the pencils face one way and some the other. The different facing pencils represent a bit, either a one or a zero, depending on if they are facing left or right.
For a "3d" disk, take all of those pencils and stand them on end, so that they are either pointing towards you or down towards the paper. Now you can pack a lot more in there without (theoretically) bleeding over into eachother.
Personally I was hoping for some hypersensitive way to detect and manipulate multiple layers of magnetic media, like the name would imply. But overall this is a nice step forward in increasing density on a 2D platter.
The ______ Agenda
I remember that the Next Big Thing in hard drives was going to be perpendicular recording, back in 1982 when it was seen as the only way to get over 10,000 bits per inch. That was over 20 years ago, and *now* it's the wave of the future? What happened?
The determined Real Programmer can write Fortran programs in any language.
It's nothing nearly as fantastic as what they claim. It's not even like a dual-layer dvd. There are no layers or thickness to the information, it's still stored on the plane of the hard drive.
When they say 3D, the mean the angular direction of the magnetic field. In current drives, the only thing that is measured is the presence or absence of magnetism. With their drives, the direction of the magnetic field also matters.
The limiting factor would be how accurately they can record and read the direction of the magnetic fields.
Sure, there are companies like IBM who put more money in to R&D than the GDP of a small Central American country, and they've been prototyping holographic drives and such for years. Yes, there were press releases, but they never said they were going to be releasing by year end.
Besides, this isn't some pie-in-the-sky technology, it's turning data stored on its side to data stored on its end...if they already having working prototypes in the field that are mass produced, why couldn't they put these on shelves by year end? I mean, it's not a new product, it'll just be the hard drive sizes we've been expecting for a while. Wouldn't surprise me if they started with 650GB in November/December and ramp up over a few years to 1.5-2.0 TB.
Oh, and also note that this isn't some no-name company (i.e. Bit Boys, Infinium) coming out with this release, it's one of the market leaders in hard drive technology (IIRC, Hitachi was the first to produce those CF-form-factor micro drives, even though they were IBM branded).
I don't think believing this makes someone a sucker; I think you're being a bit too cynical. But then again, any sucker would say that, wouldn't they?
Oh what marketing fluff. Headlines mention 2005, the first paragraph in the article says next year, and then the next paragraph says 2007. Which one should we believe? Toshiba is actually the first to bring the perpendicular recording technology to market. We are likely to see the 40GB and 80GB Toshiba drives with perpendicular recording technology in iPods real soon (June?).
Except that using flash for swap is an intensely stupid idea. Flash has a relatively small number of rewrites available before it goes bad. Most flash drives have controllers onboard to try and combat this, but there's only so many rewrites.
The previous sig has been removed due to