Scientist Uses Nanodots To Create 4Tb Storage Chip

arcticstoat writes "Solid state disks could soon catch up with mechanical hard drives in terms of cost and capacity, thanks to a new data-packed chip developed by a scientist at the University of North Carolina. Using a uniform array of 10nm nanodots, each of which represents a single bit, Dr. Jay Narayan created a data density of 1 terabit per square centimeter. The end result was a 4cm2 chip that holds 4Tb of data (512GB), but the university says that the nanodots could have a diameter of just 6nm, enabling an even greater data density. The university explains that the nanodots are 'made of single, defect-free crystals, creating magnetic sensors that are integrated directly into a silicon electronic chip.' Dr. Narayan says he expects the technology overtaking traditional solid state disk technology within the next five years."

Comments

Wow

[MyLongNickName]

My first PC had 4k of RAM. I should be used to this type of growth by now... but it still makes my heart race a bit when I see ever increasing memory capacity in an ever decreasing form size.

I'll tell my grandkids about my first PC and they will roll their eyes as they leave my retirement home...

Re:Wow

[Lumpy]

Problem is most software developers and OS makers also race to consume that memory. Honestly all the software today is a bloated blob that is horribly unoptimized for speed and efficiency.

It's disgusting how bloated most stuff is because we have 4gig of ram and 2 2.5ghz processors... why make it leand and mean? it compiles, ship it.

Re:Wow

[Idiomatick]

It may be peaking soon though. 6nm is getting close to physical maximums for most techniques due to the casimir effect. Techniques that push chips from 2d into 3d will be the next useful improvement. But after that point we have run out of easy options.

Increasing speed of chips and ram could help relieve that pressure mind you. As programmers can tade off more processing for less drive usage, or count on faster ram and compress everything. This will give us a bit more time. Beyond that we will simply have to get more inventive on how we use computers.

Very very fast internet could become important, if users feel they need access to 10million TB of data personally. That may not be physically feasible on a personal computer. So 'cloud' type services would be important. Having a few duplicates rather than 1million duplicates of any given song is clearly a big improvement. This of course feeding into the idea that when we made the internet we stopped making machines, we just started making components for the one ultimate computer. And when you think of it from that perspective there is tons of room for improvement even if some of the parts are nearing the useful maximums.

Re:Wow

[Thanshin]

why make it leand and mean? it compiles, ship it.

And what's the answer to your question?

If it works, why optimize it? To save in storage space? How much would I be saving? 10$ in storage space for every hour of optimization?

It's not art, it's a business. You could as well ask why we don't replace steel by titanium in cars.

Re:Wow

[Thanshin]

We should get benefits from newer, faster hardware. Instead we get increasingly lazy programmers and zero net benefit in speed, but with all the negative costs of new equipment purchases.

We do get benefits from newer, faster hardware. The possibility of hiring cheaper, less prepared, programmers.

Re:Wow

[Anonymous Coward]

Memory is not really cheaper than CPU once you account for the whole memory hierarchy. The L3 cache only holds a few megs of data, and fetching data from main memory requires hundreds of cycles. So unless you're talking about a loop where data can be prefetched, it's better to pay a CPU penalty in exchange for a smaller memory footprint.

Re:Wow

[Grishnakh]

that is simple - they don't do that or something like that because then your car might last you a long time - and that would cost them money because they wouldn't be able to have you a recurring revenue stream.

What does using titanium instead of steel have to do with cars lasting a long time? As long as you don't let salt corrode them away, steel-bodied cars will last pretty much forever. Here in the southwest, we don't have any problems with corrosion.

Besides, automakers wouldn't bother to apply undercoating if they wanted their customers' cars to rust away.

Sorry but the accelerated use of plastics and cheap alloys isn't an accident or an improvement in cars..

Now this is just plain stupid. Aluminum alloys improve performance in cars greatly by reducing weight, and also by making engines that perform far better. Most plastics are also a giant improvement; again, weight savings.

Re:4Tb of data (512GB)

[biryokumaru]

4 Terabits = 512 Gigabytes.

Somewhat misleading? Yes. Inherently false? No.

Re:Not a "chip", merely a "chip".

[clone53421]

Correct.

“The next step is to develop magnetic packaging that will enable users to take advantage of the chips,” says the university, “using something, such as laser technology, that can effectively interact with the nanodots.”

They have a storage medium with nothing to read or write it... yet.

Although they seem confident that this will come with time, it’s a bit early to be celebrating. Interesting technology, but time will tell whether it’ll ever be usable.

Another five-years-out technology

[alvinrod]

The technology sounds impressive, but then they just give it the kiss of death by announcing that it's five years away. Five years from now it will still be five years away, probably because while it's possible to do, no one has been able to do it in a cost-effective manner. Also if Intel can keep up with their current roadmap, they'll probably be using something close to a 10 nm process. I know that both Global Foundaries and TSMC are working on their 28 nm process (Although they are behind schedule.) so it's not inconceivable that the rest of the industry will already be at that point anyhow.

Single Defect Free Crystal

[kenp2002]

Ok my knee jerk Six-Sigma reflex has just kicked in. On the manfacturing of those defect-free crystals... and about the cost effect and scaling for "overtaking ... in 5 years..."

Ok, here is a tip:

Anytime a politician or scientist taks about 5,8, and 12 year targets there is a reason:

Two 4 year terms = 8 years; when the project falls out they can blame the canidate currently in office.

5 years = A single Term but just a touch beyond to provide an incentive for re-election because if you don't they might cancel the project

12 Year = Two terms for canidiate A and a term for his\her heir... "Don't let the evil Democrats\Republicans kill the project!"

Now last I checked more then a few grants come in at 3,5,8 and 12 year durations... I never hear things coming to fruition in 7 years, or 6 years, or 9 years, or 11 years, or 18 years, 6 months, and 3 days.

There is just something about 5, 8, and 12 they love. Which due to the frequency they cite those values implies there is some weird cosmic alignment which causes innovations to pop at those figures... or I smell 4/5 dentists approve BS.

Another one is the 20 years from now number. What is the maturity on that investment I made...

I would honestly have a lot more respect for senior scientists if they didn't spend every waking hour working on getting grant money leaving the actual work to low-paying interns and students then claiming the work as their own offering nothing more then a second hand "my team and I" comment...

Re:I hate this...

[osgeek]

They don't have any of that information because they don't know any of it. They only have a way IN THE LAB to put a shitload of nanodots onto a medium. They mentioned that they have no packaging (way to read or even really write data into the dots) for an actual product.

It's like Ben Franklin saying, "Okay, I've discovered electricity. Computers should be along in about five years."

Okay, it's not that bad, but I hate that five year timeline that is rarely questioned but is thrown out to lure in investors and grant money.

Slashdot should have an automatic filter that looks for the five year estimate and flags with some "fat chance" special color.

It's 500GB

[xaxa]

4 Terabits = 512 Gigabytes

Except it doesn't.
4 Tb = 4 000 000 000 000 / 8 B = 500 000 000 000 B = 500 GB ~= 466 GiB

Did they mean 4 Tib?
4 398 046 511 104 / 8 B = 549 755 813 888 B = 512 GiB ~= 550 GB.

According to the scientist, it's the former:

"at 10nm per bit, 1cm square stores one terabit."

That would be (1cm / 10nm)^2 b = (1e-2 / 1e-8)^2 b = 1e12 b = 1 Tb.

Re:3 ... 2 ... 1 ...

[Hurricane78]

The problem with SSDs is:
1. cheap
2. big
3. reliable
Choose two!
But even then, you can only be sure of number 3, after some years have passed. For obvious reasons of there not being any test data for years of use, until years of use have passed. ^^