Korean researchers report creation of faster, more resilient ReRam (10nS!)

http://www.physorg.com ^ | 07-20-2011 | Staff

[Red Badger]

WOW!

INSTANTANEOUS ACCESS!

Can a ultra-super-duper fast solid state 'hard drive' be far away?.......................

[HiTech RedNeck]

Tanta-lizing.

[Red Badger]

10nS memory chips ping!.................

[Red Badger]

I’m drooling on my keyboard!.....................more than usual..........

[ShadowAce]

[battousai]

withstanding a trillion read/write cycles

Can a ultra-super-duper fast solid state 'hard drive' be far away?.......................

My thoughts exactly... I always feel guilty when I write any data to my SSD, but with a trillion write cycles I could actually use an SSD for more than just the operating system :P

[HiTech RedNeck]

This new memory is expected be able to perform a trillion operations before conking out. That’s a lot, but I still wouldn’t want to see it used for something terribly busy like swap space, without wear-out indicators. It would work smashingly good — for maybe about a year.

[Obadiah]

The question is... will this go the way of rambus (RDRAM)?

[battousai]

...but I assume these will end up in high-end (big $$$) server class drives first before trickling down to the consumer level :(

[Red Badger]

This is just mind bogglingly fast!
Imagine boot-up seemingly instantaneously!
Imagine 3-D video in real time!
Imagine televideo online in real time!
Imagine a super heavy graphics web page loading instantly!
Imagine whirled peas!................

[Red Badger]

After a year your computer will be obsolete, anyway!..................

[Red Badger]

It has always been thus, Grasshopper........................

[Red Badger]

Only time will tell...........the Market will decide.................

[KarlInOhio]

It's not all that much if they count read cycles too, especially if you spend a lot of time in an idle loop pounding on the same section checking to see if there is anything else to do. TI had a demo of their new Ferric RAM processors and if you ran a program all the time instead of putting the processor to sleep, it wouldn't take long to use the memory past its guaranteed spec.

all with a switching time of just 10ns (about a million times faster than current flash chips)

So it takes about 0.01 seconds to read or write flash memory? Sounds like some author went into journalism because he didn't have a chance at a math degree.

[Calvin Locke]

Tantalum oxides can also be used in coating telescope mirrors.

[dayglored]

> So it takes about 0.01 seconds to read or write flash memory? Sounds like some author went into journalism because he didn't have a chance at a math degree.

Yeah, it's hard to reconcile that with the current SSDs which can write at >100MB/s or better (let's see, that's.... hmmmm.... 10nsec/byte. Even with interleaving, and block read/write, there ain't a factor of a million in there).

Somewhere, some apples and oranges are being compared...

[Red Badger]

Does that mean you can see only a trillion miles?.................

[Red Badger]

Math and science are not required courses in Journalism Skool.............

[antiRepublicrat]

I just did a quick calculation. Let’s say a particular part of storage is rewritten 100 times a second 24/7/365. That means ten billion seconds, with a trillion ops max we end up with 317 years of writes.

We’re so used to hearing trillion all the time, we often forget it’s a really big number.

[HiTech RedNeck]

I’d think that reads would be gentler on the devices than writes, but without technical details from the horse’s mouth who knows? Even if they are, it would require careful programming for routines intended to run in the new memory. “register” really needs to mean something. One stack variable in the memory and you could (i=0; i<N; i++) the hell out of the device before you knew it.