Seagate Exec: Hard disks anything but obsolete...

timwhit

Hairy Aussie
Joined
Jan 23, 2002
Messages
5,278
Location
Chicago, IL
Samsung is set to go straight for the PC hard disk drive market soon with the planned launch of a 16GB solid-state disk drive based on flash chips. The drives will have parallel ATA (Advanced Technology Attachment) interfaces, just like hard disk drives, and come in cases that match 2.5-in. or 1.8-in. drives so they can be used as direct replacements.

Sweet, I could replace my 18GB Cheetah with something a bit faster.

Will probably only cost $600. Time to start buying lotto tickets.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
4,178
Location
Flushing, New York
Look for the price of these SSDs to drop rapidly once they're mass-produced, and for the capacity to increase. I think mechanical and solid-state disks will co-exist for a while, at least until SSDs reach price parity, with the SSDs being used for boot disks and the mechanicals for bulk, slow storage. After the SSDs reach price parity, look for mechanical disks to die a quick death. When will this happen? My guess is by 2010 or thereabouts. Don't forget that magnetic disks are coming up on some very real capacity limits now. I don't think they'll ever get larger than 1 or 2 TB. SSDs on the other hand can be eventually easily pack hundreds of TB into the same space.

Hard disks are not dead, just yet anyway, but they're on life support at this point.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
4,178
Location
Flushing, New York
Thinking about this a bit more, I can up with a good analogy-LEDs.

LEDs have existed commercially for ~35 years but were relegated to niche uses like displays and indicators until the advent of white LEDs about 10 years ago. Between ten and five years ago LEDs made inroads into some small illumination needs like penlights and keychain lights, or for specialized uses. In the last five years, LEDs have gradually displaced incandescents in almost all small lighting needs, and many medium ones like flashlights. Soon, they'll move to things like auto headlights and general home illumination, all but obsoleting incandescents within another five years I'd say.

Right now I'd say SSDs are where LEDs where a couple of years ago. They've made significant inroads into forms of storage which formerly used mechanical disks, all but replacing floppies and being on the verge of replacing things like zip drives. However, they have yet to touch even small bulk storage needs like boot disks. Think of the Samsung's new SSD as the first commercial LED flashlight with output rivaling an incandescent. This happened roughly three or so years ago, depending upon who you talk to. Information technology evolves somewhat faster than LED technology, so in five years SSDs may all but obsolete mechanical storage, taking maybe ten years total from the time it was first used in PCs until it became universal.
 

Dïscfärm

Learning Storage Performance
Joined
Nov 22, 2002
Messages
239
Location
Hïntërländs
jtr1962 said:
...with the SSDs being used for boot disks and the mechanicals for bulk, slow storage.

There's only one problem with this, in that affordable SSDs (i.e. -- based on flash memory) are slower than today's decent SATA hard drives. SSDs based on NAND flash can make a place for themselves today in portable computing and devices, not so much the common desktop computer.

What's a bit more likely to happen with flash technology storage in the coming years is that it will be able to increase in capacity faster than it will increase in I/O throughput. Flash SSDs will become large enough in storage capacity to serve as safe tertiary storage, while hard drives (likely 2.5-inch or 1.8-inch) will serve as storage for the dynamics parts of the operating system and application software where speed is needed during dynamic file operations (i.e. -- lots of reads / writes / changes).

If a lot of storage capacity can be leveraged from relatively-inexpensive NAND flash technology, a high-speed flash memory controller using something on the order of 8:1 or 16:1 interleaving and ECC or Reed-Solomon error correction could be developed to do battle with small cheap perpendicular-recording hard drives.
 

LunarMist

I can't believe I'm a Fixture
Joined
Feb 1, 2003
Messages
16,684
Location
USA
Of course what would one expect Samsung to say, give their puny 2-platter drives and investment into flash memory? I'll likely be long gone before flash memory takes the place of large hard drives. :(
 

sechs

Storage? I am Storage!
Joined
Feb 1, 2003
Messages
4,709
Location
Left Coast
I really expect MRAM to be on the scene before flash is seriously competitive with mechanical drives.
 

Dïscfärm

Learning Storage Performance
Joined
Nov 22, 2002
Messages
239
Location
Hïntërländs
Stereodude said:
SSD's already exist, even in larger capacities too!

http://www.m-sys.com/site/en-US/

Yes. Large SSDs based on Flash, D-RAM, and S-RAM hav been around for a while. None of those are particularly affordable -- especially the ones that use DRAM or Static RAM.

The drives Samsung wants to produce use Flash memories that haven't been commercially produced yet -- flash memories that have much more storage capacity than anything available at this time.




sechs said:
I really expect MRAM to be on the scene before flash is seriously competitive with mechanical drives.

From what I gathered from various sources so far, M-RAM won't be such an inexpensive memory -- think RAMBUS-like pricing. M-RAM will be a natural for Solid State Drives, but M-RAM SSDs will end up being priced like the D-RAM and S-RAM SSDs are priced now. As far as using M-RAM for computer system memory, M-RAM will still be worth it to tweak freaks --even at RAMBUS-like prices -- because it looks to be as fast as conventional Static RAM, at least for data reads. I believe the barriers that the production labs are focusing on is increasing M-RAMs storage density.

 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
4,178
Location
Flushing, New York
Dïscfärm said:
Yes. Large SSDs based on Flash, D-RAM, and S-RAM hav been around for a while. None of those are particularly affordable -- especially the ones that use DRAM or Static RAM.
I'm not sure I even consider a drive based on DRAM or SRAM drive a real SSD. To me an SSD doesn't require battery backup. Admittedly these have their uses, although just buying a system capable of taking a large amount of RAM will accomplish almost the same thing.

From what I gathered from various sources so far, M-RAM won't be such an inexpensive memory -- think RAMBUS-like pricing.
Perhaps if we're lucky within a few years the price will drop dramatically thanks to the economies of mass production, OR something else will come along. I agree that Flash RAM based drives are still too slow in terms of STR for general hard-drive usage, cost concerns notwithstanding. However, as we all know, nothing in this business stays the same for too long.
 

Dïscfärm

Learning Storage Performance
Joined
Nov 22, 2002
Messages
239
Location
Hïntërländs
jtr1962 said:
I'm not sure I even consider a drive based on DRAM or SRAM drive a real SSD.

I'll consider DRAM and Static RAM secondary storage as a reall SSD, just simply an SSD with an Achille's heel -- and not even a particularly bad one at that.


jtr1962 said:
Perhaps if we're lucky within a few years the price will drop dramatically thanks to the economies of mass production...

One of the problems getting RAM companies to get away from DDR is that they are still investing heavily in DDR technology R&D (e.g. -- DDR3 is coming). Before they switch to M-RAM, they will want to get a return on their R&D efforts with DDR technologies, which is going to drag out the introduction of M-RAM. Of course, if some of them feel the need to jump ahead of everyone else, they may just abandon their own DDR efforts earlier rather than later. Then, you might have a race away from DDR to M-RAM. But, I wouldn't count on it.

Whatever happens, you'll first see M-RAM in bleeding edge systems (like super computers, Internet backbone routers), then it will trickle down into the SGI, Sun, HP, IBM world of $100k+ computer systems. Then, after 2 ~ 3 years, it will finally make it to the PC realm when Intel or AMD produce a memory controller for it. By then, the M-RAM factories will have finally tooled up for large scale mass production. Notebook computers might be the first beneficiaries, as you will be able to shut down your notebook computer -- cold turkey -- with programs and data still in memory when you power up the notebook computer later (provided that you don't have a BIOS/EFI setting enabled like "always clear memory when powering up").

 
Top