When are areal densities going to reach a wall?

[ElFenix]

if you've been paying attention you'd probably know that areal densities have been increasing faster than transistor densities. 1997 we had about 10 million transistors in a pentium 2, today we have 110 million on the r300. thats 11x over about 5.5 years. (heh, thats faster than moore's law, right?) granted, a lot of that is cache so its easy to pack on there, but still.

in 1997 we had 6.4 gig hard drives. these were 5 platter units (IBM) so about 1.3GB/platter. the latest from WD is 66GB/platter. thats a 52x improvement.

the question is, whats the area of the data storage "bit" on a hard drive? and how long before this catches the size of a transistor?

 

[glugglug]

Platter radius is slightly smaller than 3.5/2 inches = 44.45mm

So platter area is a bit below 6206mm^2 (6.206 x 10^-3 m^2)

6.206x 10^-3 m^2 per 66 GB = 1.176 x 10^-14 m^2 per bit (a square with an edge length of 108nm per bit).


So data storage density is already greater than transistor density.

 

[glugglug]

BTW actual density is higher than that listed above because those calcs dont take into account a) unused space in the center of the drive, or b) the platter doesn't occupy the ENTIRE 3.5" available drive bay space.

P.S. There have already been demos of data storage densities 5x higher than this. Terabyte drives will probably be in PCs in about 2-3 years.

 

[Mday]

i have a feeling terabyte PLATTERS will be in PCs in the coming years...

 

[zephyrprime]

It's not necessary to do any of these calculations because you can just look up the correct figures at the web pages of hard drive companies.

So you can see here that the max areal density of a IBM 120gxp is 29.7 gigabits/square inch. I expect this figure is similiar to that of all current 40GB/platter drives.

Also, calculations like those done before are flawed because they don't account for ecc data, sector headers, sector gaps, and spare sectors and are also problematic because they don't take into account that platter areal density is a function of the track location.

Well ElFenix, you may have already heard of the super paramagnetic limit and how it's an ultimate limitation of hard drive areal density. But the thing is, you can skirt around this by selecting materials with lower magnetic susceptability.

But given the size of an iron atom, you can only have about 4x10^16 iron atoms per square inch (assuming a square crystal which is probably wrong but close enough for calculations like this). If you could actually read and write to each atom individually, that would yield 41,000 terabits per square inch. Not considering variable areal density, this would yield a hard drive of ~55,000 terabytes per platter.

With such large platters, an Ultra ATA/133 interface would only be able to address a drive with two platters. So they better get cracking on serial ATA!

But of course, such technology is pie in the sky at the current time.

 

[tart666]

The aerial density will stop climbing at this dizzyfying rate very soon. And the reason is market related. There is already a massive glut is storage supply, the commoditization of the drive business is evident, less and less money is spent on research each year.

Hell, the inventors of MR and GMR heads got out of the drive business.

so: the hard drives will face the same fate as ram : very slow progress due to extremely low margins (anyone notice how much slower the ram is compared to the CPU?)

my 2 ¢

Toni

 

[Mday]

to be honest, the age of the winchester HDDs will prolly reach a brick wall sooner than later. but given the capacity, price and reliability of the technology, without a good replacement, they will stay even if the capacity reaches some plateau.

IC die size is also reaching a limit as quantum effects become more prominent

 

[tart666]

Mday:

exactly those words have been uttered for the last 25 years (ie: progress will run into a wall in 5 years tops)

Technologically, there is room (and evidence) for progress for the next 10 years for both IC and hard drive industries. The issue at hand is the market forces of customer's willingness to pay the premium for advanced technology. While the users are still willing to open their wallets for a faster CPU, the hard drive business is dangerously close to a complete customer apathy. Anyone run out of hard drive space lately? My guess is not in the last 2 years. While habitual spending will continue, the extra space will not fill up for the next few years, even with the divx/dvd rip/TiVo upon us. As for me, I am extremely stretched to use the 0.5TB of storage in my house.

Without the high premiums, the research in the storage area will grind down to a trickle. Anyone looked at stocks of HD equipment manufacturers? (veeco, MRC, etc) their financial statements continually indicate lack of demand from HD makers, meaning severe cuts in research/investment budgets. Combined with commoditization of the business, this means slow progress.

The progress will stay slow until the next killer app is found. perhaps HDTV TiVO / DVD2 ripping will be it. perhaps the capacity will outrun the demand even then. make your stock purchase decisions wisely.

some more ¢

Toni

 

[Evadman]

IBM was researching a drive built on what they called "pixie dust" if I remember correctly. It was a atom (or a small amount of atoms) on a small pyrimid of sorts that would keep the magnetic properties from affecting the other adjacent particles. I remember a quote of a terabite platter being about the limit of the tech, but not sure. I will see if I can find the article.

 

[zephyrprime]

IBM was researching a drive built on what they called "pixie dust" if I remember correctly

An article about pixie dust can be found on IBM's page. Pixie dust is being used in some of the newest IBM drives. The "pixie dust" works by giving magnetic domains more magenetic moment. That is, it takes more energy to flip the bits.

 

[ElFenix]

thanks for the discussion!

i sorta figured we'd have market reasons for not going ahead before a physics reason, since i talk to so many customers that don't use more than a handful of gigs...

 

[rimshaker]

The ultimate limit is when a magnetic field change can no longer be detected reliably anymore. There is an ultimate limit of course... we're talking about the spin orientation of a single electron. This what is going to bring terabyte platters in the near future.

 

[Howard]

Hopefully, non-volatile memory chips will be cheap enough in the future to be affordable for the home user, and they would definitely be faster than disk-based systems today.

 

[SKORPI0]

Interesting article here. Scroll to Seagate article dated Sept 6,2002, at buttom of page.

A standard 95mm magnetic disk has a useful surface area per side of about 6.5 square inches. This means that a single platter 3.5-inch disk drive could have a storage capacity of 50x6.5x2.0 (two-sides) Tbits, or 650Tb or 81TBytes! Seagate, in its publicity release, makes the point that a single 3.5-inch disk drive of this type could accommodate the entire print-contents of the Library of Congress in Washington DC!

Probably available before end of this decade. :frown:

 

[ProviaFan]

Originally posted by: SKORPI0
Interesting article here. Scroll to Seagate article dated Sept 6,2002, at buttom of page.

A standard 95mm magnetic disk has a useful surface area per side of about 6.5 square inches. This means that a single platter 3.5-inch disk drive could have a storage capacity of 50x6.5x2.0 (two-sides) Tbits, or 650Tb or 81TBytes! Seagate, in its publicity release, makes the point that a single 3.5-inch disk drive of this type could accommodate the entire print-contents of the Library of Congress in Washington DC!

Probably available before end of this decade. :frown:

Woah, with a couple of those in RAID you'd be able to store the entire contents of the Kazaa network on one system. :Q

I bet these drives won't be happening unless the RIAA gets their grubby little fingers in there to "secure" the drives first.