Intel, Micron sample 20nm NAND flash

Apr 15, 2011
New 64 Gigabit (Gb) NAND flash die from Intel Micron Flash Technologies - Intel and Micron deliver the industry’s smallest, most advanced NAND flash process technology at 20nm. Shown is a 64Gb, or 8 Gigabyte (GB), die measuring just 118mm2. The 64Gb Multi-Level Cell (MLC) NAND device provides high capacity for smartphones, tablets, SSDs and more.

Intel and Micron Technology today introduced a new, finer 20-nanometer (nm) process technology for manufacturing NAND flash memory. The new 20nm process produces an 8-gigabyte (GB) multi-level cell (MLC) NAND flash device, providing a high-capacity, small form factor storage option for saving music, video, books and other data on smartphones, tablets and computing solutions such as solid-state drives (SSDs).

The growth in combined with feature enhancements for tablets and smartphones is creating new demands for NAND , especially greater capacity in smaller designs. The new 20nm 8GB device measures just 118mm2 and enables a 30 to 40 percent reduction in board space (depending on package type) compared to the companies' existing 25nm 8GB NAND device. A reduction in the flash storage layout provides greater system level efficiency as it enables tablet and smartphone manufacturers to use the extra space for end-product improvements such as a bigger battery, larger screen or adding another chip to handle new features.

Manufactured by IM Flash Technologies (IMFT), Intel and Micron's NAND flash joint venture, the new 20nm 8GB device is a breakthrough in NAND process and technology design, further extending the companies' lithography leadership. Shrinking NAND to this technology node is the most cost-effective method for increasing fab output, as it provides approximately 50 percent more gigabyte capacity from these factories when compared to current technology. The new 20nm process maintains similar performance and endurance as the previous generation 25nm NAND technology.

Comparison of two 32Gb 34nm die versus one 64Gb on 25nm and 20nm process from IMFT - This photo shows a comparison of two 32 Gigabit (Gb) Intel Micron Flash Technologies (IMFT) 34nm die versus one 64Gb, or 8 Gigabyte (GB), die on 25nm and new 20nm processes. Shrinking NAND lithography is the most cost-effective method for increasing fab output and reducing die cost. Shrinking from 25nm to 20nm process will provide an approximately 50 percent more gigabyte capacity from IMFT factories when compared to current technology. The new 20nm process maintains similar performance and endurance as the previous generation 25nm NAND technology.

"Close customer collaboration is one of Micron's core values and through these efforts we are constantly uncovering compelling end-product design opportunities for NAND flash storage," said Glen Hawk, vice president of Micron's NAND Solutions Group. "Our innovation and growth opportunities continue with the 20nm NAND process, enabling Micron to deliver cost-effective, value-added solid-state storage solutions for our customers."

"Our goal is to enable instant, affordable access to the world's information," said Tom Rampone, vice president and general manager, Intel Non-Volatile Memory Solutions Group. "Industry-leading gives Intel the ability to provide the highest quality and most cost-effective solutions to our customers, generation after generation. The Intel-Micron joint venture is a model for the manufacturing industry as we continue to lead the industry in process technology and make quick transitions of our entire fab network to smaller and smaller lithographies."

The 20nm, 8GB device is sampling now and expected to enter mass production in the second half of 2011. At that time, and Micron also expect to unveil samples of a 16GB device, creating up to 128GBs of capacity in a single solid-state storage solution that is smaller than a U.S. postage stamp.

Explore further: Ultra-low consumption for the future of electronics

Related Stories

Intel, Micron Introduce 25-Nanometer NAND

Feb 01, 2010

Intel Corporation and Micron Technology, Inc. today announced the world's first 25-nanometer (nm) NAND technology, which provides a more cost-effective path for increasing storage capacity in such popular ...

Micron, Intel try out 50 nm NAND memory

Jul 25, 2006

Semiconductor giants Micron and Intel said Tuesday they were sampling the first NAND flash memory chips built on 50-nanometer processing technology.

Recommended for you

California bans paparazzi drones

1 hour ago

California on Tuesday approved a law which will prevent paparazzi from using drones to take photos of celebrities, among a series of measures aimed at tightening protection of privacy.

Samsung rejects claims of Galaxy Note defect

2 hours ago

Samsung on Wednesday rejected claims that its new oversized smartphone had a defect, after some customers posted pictures of a gap between the frame and display panel.

User comments : 11

Adjust slider to filter visible comments by rank

Display comments: newest first

GSwift7
1 / 5 (1) Apr 15, 2011
and make quick transitions of our entire fab network


That's the part that impresses me the most.
kaasinees
2.3 / 5 (3) Apr 15, 2011
Cycle life goes down as you shrink NAND.
SSD's dont sound so promising anymore.
GSwift7
1 / 5 (1) Apr 15, 2011
Agreed. (with the current technology)

The potential for making a huge pile of money on SSD's is huge though, so I'm sure someone will eventually come up with a good SSD.
Parsec
not rated yet Apr 15, 2011
Guys... SSD's work fine if you configure them correctly. Simply do not put the paging file on the SSD. With a nice pile of RAM (16-32gbyte), the value of paging declines dramatically anyway. Just put the paging file on one of your other Hard drives.

This drastically reduces the number of write cycles to the SSD. Most of the time you write a file once when you install a program, and then read it when you use it.

Using SSD's in this way will give you lifetimes that exceed the life cycle of hard disks, because hd's have their own problems due to the number of mechanical components available for breakdown.
kaasinees
1 / 5 (1) Apr 15, 2011
@Parsec

That is the half truth.

SSD's degrade after time and become slow. The smaller the NAND is , the faster it degrades. 50nm has a nice life cycle though. But 20nm? nah.
rgwalther
5 / 5 (1) Apr 15, 2011
Like all tech, SSD have reached their limit and there is no future. It is incredible that anyone can still write this 'end of tech pronouncment' with a straight face.
PinkElephant
not rated yet Apr 15, 2011
SSD's degrade after time and become slow
Not because the NAND degrades, but because of data fragmentation. It's the same reason your HDD becomes slow, until you defragment it. With SSDs, there's now the TRIM mechanism that achieves substantially the same thing as defragmentation of hard drives.
50nm has a nice life cycle though. But 20nm? nah.
If you want long lifetimes, go with SLC rather than MLC NAND. But unless you're an enterprise user, you aren't going to torture your SSD enough to wear it out before the rest of your computer had become obsolete anyway. Plus, with larger SSD capacities, the maximum wear on each individual cell goes down (there's more of an opportunity for the SSD's controller to distribute the load.)

Intel's 320-series consumer-grade SSDs (using 25 nm MLC NAND) have MTBFs on the order of 1 million hours (that's 114 *years*.)
soulman
5 / 5 (1) Apr 16, 2011
Intel's 320-series consumer-grade SSDs (using 25 nm MLC NAND) have MTBFs on the order of 1 million hours (that's 114 *years*.)

Don't fall for the MTBF marketing. It certainly sounds impressive, but what does it really mean?

We would need to know how many SSDs were tested and for how long, and we need to know what is the definition of 'failure' (catastrophic, minor, somewhere in between?).

Here's an example. There is a population of 500,000 25 year old people. During one year, 625 of them die (fail). What is the MTBF of this population?

500k people x 1 year = 500k people years. (500k people years / 625 people) = 800 years.

So, the MTBF for a person is 800 years! How many people do you know that live for 800 years? Even 1/10 of that MTBF figure is optimistic (for people).

So, yes, high numbers are generally better, but don't be fooled by what the figure actually means and how it's measured.

Eikka
5 / 5 (1) Apr 16, 2011

So, yes, high numbers are generally better, but don't be fooled by what the figure actually means and how it's measured.


Indeed. Your method was an example of how to misinterpret MTBF completely.

It can't be longer than what is expected of any single unit separately, so if it is known that a product will allways fail due to some inherent reason after 100 years, its MTBF cannot be higher than that.

There's no inherent reason why an SSD would fail in any number of years other than what damage it accumulates in use.
PinkElephant
not rated yet Apr 16, 2011
500k people x 1 year = 500k people years. (500k people years / 625 people) = 800 years.
That's not how MTBF is computed. To get at MTBF, you take a representative sample of the population, let's say 100 people, and wait until all of them die. Then compute the average of all the recorded lifetimes; that's your MTBF.

Of course, with devices such as SSDs, you don't put them into real-life usage scenarios and wait for them to fail (who has 114 years to spare?) What you do instead, is estimate typical daily load under normal usage, and then torture-test the units until they fail. When they fail, compute realistic lifespans by multiplying torture-test lifespan by total load over the test, divided by estimated normal load.

Or at least, that's the simple model. There are also ways to estimate MTBFs by combining known MTBFs of various components or smaller sub-units of the device (the result is smaller than the smallest MTBF among all components.)
soulman
not rated yet Apr 16, 2011
Indeed. Your method was an example of how to misinterpret MTBF completely.

It can't be longer than what is expected of any single unit separately, so if it is known that a product will allways fail due to some inherent reason after 100 years, its MTBF cannot be higher than that.

Are you really so sure? Seagate's previous generation of enterprise disc drives have a MTBF of 1,600,000 hr. Do you really think ANY disc drive is going to last over 180 years in use?

The example I cited in my previous post was originally written by Wendy Torell and Victor Avelar in the white paper "Mean Time Between Failure: Explanation and standards". I should think they would know a thing or two about how MTBF figures are used and abused.

In fact, MTBF is increasingly being amended by a different measure - Annualized Failure Rate (AFR), for added clarity and precision.