Inventor of GMR Bids To Shake Up Storage, Again

Nrbelex writes "Stuart S. P. Parkin, an I.B.M. research fellow largely unknown outside a small fraternity of physicists, thinks he is poised to bring about a breakthrough that could increase the amount of data stored on a chip or a hard drive by a factor of a hundred. This is the man who pioneered exploiting the giant magnetoresistance effect in the 90s, causing disk storage to jump ahead of the Moore's Law curve. If he proves successful in developing 'racetrack memory,' he will create a universal computer memory, one that can potentially replace DRAM and flash memory chips, and make a 'disk drive on a chip' possible. It could begin to replace flash memory in three to five years, scientists say."

Comments

Storage leaps

[CopaceticOpus]

Imagine the social implications of the $50 5TB thumb drive...

"I was over at Jimmy's house yesterday and asked him to put some good stuff on my thumb drive. He gave me HD copies of the top 80 movies released in the past two years, plus 2000 of his favorite albums.

Meanwhile, a second thumb drive I keep clipped to my belt has been keeping an audio/video recording of the last 17 months of my life, nonstop."

Re:Anybody bought a hard drive in the last 10 year

[doombringerltx]

Singling him out does take in account the attitude of the times. Anti-antisemitism was much more common then, but even in those days The International Jew was a step beyond. They earned him the title of the only American mentioned in Mien Kampf.

Re:Sounds like bubble memory...

[JoelKatz]

He should be able to move his magnetic domains about once per nanosecond. If you put 128 bits on a ring, that would mean a single ring could clock out 128 bits in as many nanoseconds. There's a bit of write overhead involved in bringing the correct domain underneath the write head, kind of like seek time. Again, for 128 bits per ring, assuming random access, average seek delay would in the neighborhood of 64 nanoseconds.

You would likely rig things so that writes are buffered and sequential addresses take sequential magnetic domains. So write speed would be quite high unless you had out-of-order writes to the same domain, in which case you'd stall an average of 64nS.

You can imagine a pathological write pattern, kind of like trying to alternately write to disk sectors on complete opposite sides of the disk. Even with such a pathological write pattern, you could still do one write every 200nS or so.

Guesstimated throughput would be, assuming 128-bit operations, 1nS rotation time, 128 bits per racetrack, and reasonable buffering and multiplexing:

Random read: 250MB/s
Sequential read: 5GB/s
Random write: 500MB/s
Sequential write: 5GB/s

Re:Oh, no! Not the dreaded

[rolfwind]

Blu-Ray burners are currently retailing for 500$ and up; HD-DVD for 600$ and up, and the media is between 10$ and 25$ per disk. At those price points, I'd say we're still having issues getting to BR/HD-DVD, let alone beyond it. In five years these burners will be under 100$ and the media will be under 5$; at which point the technology in the article will hopefully be hitting the wild at a price point similar to where Blu Ray and HD-DVD are now - prohibitively expensive for consumers, but available for those who actually need it.

I'll wager with you that it will be less than 3 years, and that the media will be less than $3 a blank.

What is also different now is that people have backup hard drives - in the 90s almost no one had a spare hard drive, perhaps an extra one in the machine. But just a spare hanging around for backups? No. Now, they are in every store.

The humble CD-R,DVD-R has competition - those harddrives and flash drives it did not have 7 years back. And those devices are faster and more competitive in key areas. Especially cost - a 500 GB harddrives cost less than $90 on Pricewatch. That is about 106 4.7GB DVDs in capacity at only twice the cost. Figuring time involved burning - that is less expensive.

I don't even think Blu-ray or HD-DVD is that exciting from a capacity standpoint for back-ups. DVD-Rs are only convenient for me right now when I give data to friends - that's it.

Perhaps HVD will come out soon and that is the only thing I am looking forward to in the area of backing up on optical disks.

Re:Whats really interesting..

[Just Some Guy]

Imagine not needing elaborate schemes like virtual memory paging, harddrive caching, file systems, or even needing to compress things as often.

Virtual memory is A Good Thing (tm), and once you have a working VM system, paging can be a relatively easy add-on. Stripping out paging but keeping VM won't be all that much simpler, I don't think.

HD caching, I'll give you.

File systems? You'll still need to find all that data and mete out access to it. Since I don't know what latencies would be like in this storage, it's possible that current filesystems that are optimized for high latency and high throughput might still be reasonable.

Finally, at some point you're going to want to transfer that data from storage to the CPU or to another machine. As busses and network links have finite bandwidth, compression will still rule the day.

Having said that, I still welcome our terabyte-thumbdrive overlords.