News Blog

Is MRAM better than flash memory? That's a question a new venture business will try to answer.

Freescale MRAM chip

(Credit: Freescale Semiconductor)

Former Motorola chip unit Freescale Semiconductor announced Monday that it has joined with several venture capital firms to form an independent company focused on MRAM (Magnetoresistive Random Access Memory).

The new company, EverSpin Technologies, will "expand its current portfolio of standalone MRAM and related magnetic-based products," the companies said in a statement.

MRAM uses magnetic materials combined with conventional silicon circuitry to deliver a high-performance permanent storage device.

But MRAM must compete with quickly evolving technologies like flash memory-based solid state drives. Flash memory is gaining ground because companies like Samsung, Toshiba, and Intel keep developing faster and higher-capacity devices.

(For more information on MRAM see MRAM-info. For an in-depth explanation of technologies used in MRAM see this explanation of electron spin and so-called spintronics.)

Freescale will transfer the MRAM technology, related intellectual property, and products to EverSpin Technologies and will retain an equity position in the new venture, the companies said. EverSpin is backed by venture firms New Venture Partners, Sigma Partners, Lux Capital, Draper Fisher Jurvetson, and Epic Ventures.

"The decision to form a new company is intended to accelerate the adoption of MRAM," Lisa Su, senior vice president and chief technology officer for Freescale Semiconductor, said in a statement.

"Current Freescale MRAM products have strong traction in the market," Steve Socolof, managing partner of New Venture Partners, said in a statement.

As part of the agreement, EverSpin Technologies will take ownership of the MRAM manufacturing assets and will be based in Chandler, Ariz.

EverSpin will continue to supply products to Freescale's existing standalone MRAM customers. In addition, EverSpin will be a supplier to Freescale of MRAM technology for use in Freescale's embedded products.

Fast silicon is hitting a wall in game PCs, according to Alienware, which is looking for ways to boost game PC performance.

Parent company Dell vowed on Tuesday to pour more resources into the game PC unit and invest in "product development, design, and engineering."

Alienware Area-51 m9750 notebook

(Credit: Alienware)

Alienware's Marc Diana believes optimizing systems for the 64-bit world would allow game PCs to make big strides in performance. In effect, today's 32-bit environments are putting a crimp on PC-based gaming.

"So many people are caught up in this hardware race. Dual-core, quad-core this and that," said Diana, who is Alienware's product marketing manager for desktops. "If these companies--Intel, Microsoft, Nvidia, ATI, and AMD--if they'd just sit down and realize the performance benefit of optimizing their drivers and software for 64-bit."

"I think that would make sense now," Diana said emphatically.

Much of the software in the PC world is still 32-bit, including most copies of Windows XP and Vista. In fact, Diana said Alienware doesn't offer 64-bit operating systems because "we don't feel comfortable shipping a system to a customer with the 64-bit driver support that's out there in the industry."

The most obvious limitation of 32-bit operating systems and applications is a cap--4GB--on how much memory an operating system can use. And some applications can't even use the entire 4GB. "Who cares about DDR3 memory? What about giving me 4GB?" Diana asked.

"They're building (software) for something that is inherently very old technology," he said. "We (need) drivers that are very healthy in the 64-bit space. I'm not saying that 64-bit drivers don't exist. I'm just saying there's not enough software development and support on that end to warrant companies like us to move to 64-bit operating systems."

He also talked about other factors--beyond faster processors and graphics chips--that affect system performance, particularly for consumers who have limited budgets. "If I was looking to invest in one component over another," Diana said, "I would probably invest in a really good motherboard," and after that, a dual-core processor and a midrange graphics card such as Nvidia's 8800GT or ATI's X2 card.

New DDR3 memory is also becoming more of a factor. DDR3 memory is offered in two Alienware platforms. "It is the highest-performing memory now on the market. But I'm not so sure it's quite there yet. The cost is very high," he said. "Six months from now it will start making a lot more sense (economically) than it does right now." Because of this, DDR2 memory is still widely used.

DDR3 memory modules use less power and double the data prefetch buffer to 8 bits from 4 bits per cycle. DDR3 also operates at higher clock rates (1600 MHz), among other improvements.

It was a search for the essence of things that lead to the memristor, says UC Berkeley professor Leon Chua.

This week, HP Labs announced it had made a memristor, or memory resistor, a fundamental circuit element first theorized by Chua decades ago. If they become commercially practical to make, memristors could lead to very dense, energy-efficient memory chips that don't cost much because they don't need much silicon. A memristor has a variable resistance; as a result, memristors can "remember" how much charge was applied to it. (See here for more on HP's memristor.)

While most have accepted Chua's work, it has mostly been considered theoretical. But how did Chua come up with the mathematical formula for proving memristors exist in the first place?

17 memristors in a row, freshly made by HP

(Credit: J.J. Yang, HP Labs)

He started looking at what truly defined different circuit elements, similar to the approach Aristotle took when trying to define substance and essence.

"I asked, 'What is a resistor? What is a capacitor?' No one was really asking that," Chua said. If you asked someone what a resistor was, they'd say, 'It gets hot so let's make an oven out of it.' That was the mentality."

Chua then took four variables: voltage, current, charge, and flux. A resistor was defined by current plus voltage, he said. A capacitor was defined by voltage plus charge. Flux and current made an inductor. That took care of three out of four of the known circuit elements.

There was only one possible combination left, according to Chua. Flux plus charge, which he defined as a memristor.

Why did it take so long to eventually make a memristor? He gave two reasons. One, researchers chalked up evidence pointing to memristors, or effects created by memristors, as anomalies in their own experiments.

Second, material science has made huge strides. The memristors developed by HP measure 5 nanometers across. "That's the length of five sugar molecules," Chua said. "The memory effect dominates."

Leon Chua

(Credit: UC Berkeley)

Chua noted that he actually made a rough prototype back when the paper first came out, but it was impractical and manufacturers weren't interested in developing it.

And, it's not the first time it's taken a while to prove something. He pointed to Aristotle's law of motion. Not familiar with it? That's because it turned out to be wrong. Aristotle said that force should be proportional to velocity. Centuries later, Newton showed that force was actually proportional to acceleration.

"They were looking at the wrong variable," he said. "The same thing happened here."

It's the tale of the lost circuit.

Thirty-seven years ago, Leon Chua, a professor at the University of California at Berkeley, mathematically theorized that scientific symmetry demands that there should be a fourth fundamental circuit element. Engineers were already familiar with resistors (which resist the flow of electricity), capacitors (which store electricity), and inductors (which resist changes to the flow of electrical current), which can be combined to build more complex devices. The fourth circuit, which Chua called a "memristor" for memory resistor, would register how much current had passed.

"He looked at fundamental circuit equations and noticed there was a hole," said Stan Williams, who heads up the Information and Quantum Systems lab at HP Labs, "There should be a device that remembers how much current flowed through a device."

An atomic force microscope image of a circuit with 17 memristors in a row. The memristor consists of two titanium dioxide layers connected to wires. When a current is applied to one, the resistance of the other changes. That change can be registered as data.

(Credit: J.J. Yang, HP Labs)

Williams and other scientists at Hewlett-Packard are publishing a paper in Nature on Wednesday demonstrating that that these things actually exist. HP has a few discrete memristors as well as a silicon chip embedded with memristors. It's a first, according to HP.

If memristors can be commercialized, it could lead to very dense, energy-efficient memory chips. Scientists have made devices that function like memristors, but it took a good number of transistors and several capacitors, Williams said. Memristor chips would function like flash memory and retain data even after a computer is turned off, but require less silicon, consume less energy, and require fewer transistors.

A memristor effectively stores information because the level of its electrical resistance changes when current is applied. A typical resistor provides a stable level of resistance. By contrast, a memristor can have a high level of resistance, which can be interpreted as a computer as a "1" in data terms, and a low level can be interpreted as a "0." Thus, data can be recorded and rewritten by controlling current. In a sense, a memristor is a variable resistor that, through its resistance, reflects its own history, Williams said.

Varying resistance is the same principle at work with phase change memory. The difference in phase change memory, which will come to market later this year, is that changes in resistance are accomplished through a substantial amount of heating. A bit on a CD-like substrate is heated rapidly a few hundred degrees and then cooled. Depending on how rapidly the bit cools, the material becomes crystalline or amorphous. The different states--crystalline and amorphous--exhibit different states of resistance.

"We can get it (resistance changes) with less energy," Williams said. "It is a large amount of resistance change with a small amount of memory."

The secret sauce in HP's memristors is two layers of titanium oxide, a crystalline material consisting of one titanium atom and two oxygens, sandwiched between two metal wires. The bottom layer consists of standard, consistent titanium dioxide. The upper layer is missing a few oxygens--less than 1 percent--which creates voids. When a current is applied (via the wire) to the upper layer, the vacancies are pushed into the lower level of titanium dioxide. That changes the resistance of the lower level. Subsequent bursts of current can then reverse it.

"All we have to do is push around a very small number of vacancies in a crystalline material," Williams said. "We can switch it very fast, faster than we can measure."

Pushing the voids into the consistent layer of titanium dioxide does not change its characteristics otherwise. He likens it to bubbles in beer. "You can have bubbles in it, but it's still beer," he said.

Memristors in green. The wires in this image are 50 nanometers wide, which comes to about 150 atoms.

(Credit: J.J. Yang, HP Labs)

Memory and storage are the new frontier for chip designers. The explosion of data will require new ways to retrieve and store it. Cloud computing? It's a big hard drive, if you think about it. Numonyx, the Intel and STMicroelectronics joint venture, is leading the effort to commercialize phase change memory. IBM is working on ways to store data through magnetic charges on a wire. Seagate Technology, Hitachi, Zettacore, Grandis, and others are working on different memory and storage concepts.

HP has largely exited the chip business, but it has increased efforts to license the intellectual property inside its labs. The company, for instance, will likely try to commercialize the crossbar latch technology, which allows molecular grids to perform calculations. (Williams also works on that.)

While memristors can be made on silicon chips, memristor devices will require engineers to learn a new circuit design discipline.

"The technology is in good shape. The big barrier is not whether you can make it," Williams said. "It is the effort to design new circuits."

Don't make computers seek out data. Make the data move to where it can be used.

That, roughly, is one way to describe the racetrack memory concept, which IBM argues could one day lead to memory that could hold 100 times more data than flash memory does today and cost 100 times less. So that 2GB card you bought for $20 this week would hold 200GB, or more than a lot of notebook hard drives, and cost 20 cents.

In racetrack memory, information is stored in the domain walls, or boundaries, between magnetic regions on a wire. The domain walls are then shuttled up or down the wire via electrical pulses toward another component that can interpret whether the domain wall represents a "1" or a "0."

"We have a series of zeros and ones, and our objective is to shift that information to and fro without upsetting it," said Stuart Parkin, an IBM fellow, in an interview. Parkin is one of the authors of a paper on the subject being published in the April 11 edition of Science. "Unlike a hard drive, we have no moving parts. We have no moving atoms. We just have magnetic moments."

In flash memory and hard drives, data lives in a discrete location and a computer (or hard drive head) finds it. Shuttling the bits on a wire opens up the possibility for making 3D memory, and hence more dense memory, because wires could be stacked on top of each other. The time it takes to record or retrieve data could also be reduced.

Racetrack chips, potentially, could additionally last far longer because they have no moving parts, unlike hard drives, and won't get progressively worn out by successive read-erase cycles like flash memory. Flash chips typically last 100,000 read-write cycles before errors can become a problematic possibility.

The paper in detail describes how they were able to create, move, and interpret domain walls on horizontal permalloy nanowires.

One of the big breakthroughs in IBM's approach, said Parkin, is the fact that the domain walls are moved with electrical current. In the past, scientists tried to move domain walls in this manner with magnetic fields. That created two problems. One, using magnetic fields takes far more energy. Second, the magnetic fields can disturb adjacent magnetic fields, thereby potentially corrupting data.

"We spent about three years together on this. Three or four years ago, people hadn't even demonstrated moving one domain wall with small bursts of current," Parkin said. "It is an understanding of how the magnetic fields work together with building the nanowires in such a way that the domain walls can move smoothly along these wires without getting stuck on small perturbations."

Parkin is a leading figure in magnetic storage research. His work on thin magnetic film structures allowed IBM, among others, to exploit the giant magnetoresistive effect to significantly boost the density of hard drives.

In the next two to four years, IBM hopes to create a complete, working prototype of a racetrack chip with an integrated device that can read the data shuttling across the wire, said Parkin. In 7 to 10 years, chips like this, conceivably, could start coming out of factories. IBM doesn't make memory chips, but is interested in coming up with ideas and inventions in the area it can subsequently license.

It's all about data storage
If the semiconductor market revolved around processors in the 1990s, you can make a good argument that it's going to revolve around data storage in the next decade. The growth of the Internet and digital media has lead to the need for chips, software, and systems that can help store--and then find and retrieve--terabytes and exabytes of data. (An exabyte is a quintillion bytes, or a billion gigabytes.)

"The problems we're looking at aren't computationally driven, per se, but more information management problems," Mark Dean, an IBM fellow and director of the Almaden Research Center, said in an interview in February. "Computation is not the hard part anymore."

In the memory world, several companies are touting approaches for replacing existing technologies. Earlier this month, for instance, Numonyx--a joint venture between STMicroelectronics and Intel--said that it will later this year begin to commercially ship phase change memory (PCM), a type of dense memory that scientists have experimented with in labs for decades.

Start-up Grandis, as well as IBM, meanwhile, are examining spin transfer torque memory (STT RAM), which operates on similar principles as racetrack memory, while Zettacore is trying to store data with designer molecules. (IBM also has a lab project under way in which DNA could be used to organize carbon nanotubes into grids for data storage.)

Traditional approaches, of course, aren't giving up easily. Toshiba discussed a technique for building 3D flash chips. SanDisk acquired a line of 3D read-only memory flash chips when it bought Matrix Semiconductor, and is working on chips that can read, erase, and rewrite data.

In the hard drive world, Seagate will try to increase density on drives with a heating technology, while Hitachi is pursuing patterned media hard drives.

Micron Technology announced that it is sampling 4 gigabyte (GB) memory modules based on high-speed DDR3 technology and said the memory has been validated by Intel to run on its upcoming Centrino 2 mobile processor.

Micron DDR3 memory module

(Credit: Micron)

DDR3 SDRAM or double-data-rate three synchronous dynamic random access memory is expected to be faster than DDR2 SDRAM--now used widely in systems--though this will depend on the speed rating of the DDR3 memory and on what type of DDR2 memory it is tested against.

Micron's DDR3 modules support data rates of up to 1333 megabits per second, enabling better system and graphics performance. DDR3 supply voltage operates at 1.5-volts in comparison to DDR2's 1.8-volts, reducing power consumption by up to 30 percent, Micron said.

Micron said it has received Intel's validation on 512MB, 1GB, and 2GB DDR3 notebook modules for the upcoming Intel Centrino 2 processor technology mobile platform. The 4GB DDR3 notebook modules are still going through the validation process. Centrino 2 processors--formerly known by the code name "Montevina"--are due this summer.

The modules are designed using 2 gigabit (Gb) components, providing high-density DDR3 modules for notebook computers, such as those that would use the Cetrino 2 processor. High-density memory modules with large capacities are becoming increasingly important for notebook computers as graphic-intensive operating systems and other content heavy applications continue to make their way onto the market, Micron said.

DDR3 memory products that support Intel's high-performance desktop, workstation, server, and mobile platforms in 2008 are also being developed.

Micron's 512MB, 1GB and 2GB modules are in mass production now, with its 2Gb-based DDR3 4GB modules expected to be in mass production in Q2 2008.

It's been a long haul for phase change memory, but the goal is in sight.

Numonyx, the memory joint venture between STMicroelectronics and Intel, is already shipping samples of phase change memory (PCM) chips to customers and will start shipping PCM chips commercially later this year, CEO Brian Harrison said at a press conference Monday.

"We expect to bring it to market this year and generate some revenue," Harrison said. "It is one to two years before it becomes widely commercially available."

Hearing a CEO talk about existing samples and near-term commercial shipments is a big deal for PCM. The technology has been stuck in the proverbial "a few years away" phase for a long time.

"It could be cheaper than flash within a couple of years," analyst Richard Doherty in said in 2001, predicting the technology might hit the market in 2003.

"We are making good progress," Stefan Lai, one of Intel's flash memory scientists, said in 2002.

Gordon Moore, co-founder of Intel and the man for whom Moore's Law was named, had an article in the September 28, 1970 issue of Electronics predicting that Ovonics Unified Memory, another name for the same type of memory, could hit the market by the end of that decade. (The same issue of Electronics also included this article: "The Big Gamble in Home Video Recorders.")

The delays have largely stemmed from two sources. First, it's not an easy technology to master. In phase change memory chips, a microscopic bit on a substrate gets heated up to between 150 degrees and 600 degrees Celsius. The substrate is made of the same stuff as CD disks. The heat melts the bit, which when cooled solidifies into one of two crystalline structures, depending on how fast the cooling takes place. The two different crystalline structures exhibit different levels of resistance to electrical current, and those levels of resistance in turn are then as ones or zeros by a computer. Data is born.

Both Intel and ST made a significant amount of progress in controlling the material in the past few years, Harrison said.

Size matters
Second, the makers of flash memory have continued to improve their technology. Back in 2001, some believed that flash would hit a wall at the 65-nanometer level of chip design. Then that got moved to 45 nanometers. Today, manufacturers mass-produce flash at 65 nanometers and have samples at 45 nanometers. Numonyx has samples of traditional NOR flash at 32 nanometers. Why switch when the existing technology continues to work?

Again, in the past few years, Intel and ST have made progress and figured out a way to produce PCM chips on the manufacturing lines developed for standard chips. That has eroded the barriers to bringing PCM out.

Although Philips, IBM, and others have made progress in PCM, only Samsung is close to coming out with chips commercially, Harrison said.

Why will the world want PCM? Performance, says Numonyx CTO Ed Doller. PCM chips can survive tens of millions of read-write cycles, he said, or far more than flash. Reading data to PCM chips takes 70 to 100 nanoseconds, or as fast as NOR flash. Data can be written to the chips at a rate of 1 megabyte a second, or equivalent of NAND flash. There is also no erase cycle, making it similar to DRAM.

In other words, you have the best attributes of three different types of memory--plus, PCM will potentially use far less power.

The cost premium is also coming down fast. By next year, Numonyx hopes to make PCM chips, using 45-nanometer processes, that can hold two bits of data per cell. If that's possible, those chips would compete in price with single-bit-per-cell NAND flash, the memory that's being put into solid-state drives today, said Doller.

But the most important thing is that scientists believe they will be able to increase the density of these chips comparatively easily. In the future, standard flash chips will need additional circuitry for error correction and other functions. Not so with PCM. The smaller the bits get, the less heat that will be required to flip them, Doller added.

"The most important thing is that it is scalable," Doller said.

Do you want to know the best thing about a notebook with a flash memory drive, rather than a conventional hard drive?

It's the silence.

The notebook I'm testing--a Dell Latitude D830 with a 64GB flash hard drive from Samsung--hasn't emitted a sound in three days. Flash drives, which store data in NAND flash memory, don't require motors or spinning platters. Thus, there are no whirring mechanical noises.

A Dell Latitude with a flash drive. You can definitely tell a difference in performance, but is it worth the $900 premium?

(Credit: Michael Kanellos/CNET News.com)

Compare that with my T42 ThinkPad. It sounds like a guinea pig got trapped inside, particularly during the start-up phase. Vzoooot. Cronk, cronk, cronk. Zip, zip. (Pause.) Gurlagurlagurla...zweeee.

The lack of a mechanical hard drive also means lower power consumption and less heat. In turn that means the fan rarely, if ever, needs to kick into action. As I type, for instance, the notebook is running eight video streams-- two from CNN, two from CNET, two from MSN, a video on new bands on Crackle, and a pirated Led Zeppelin video on YouTube--and the fan won't trip over. The computer is running on battery power and the videos, with a few minor gulps, are all running smoothly.

If it did have a conventional hard drive, the fan would have flipped on, sapping battery power, and cranking out some white noise. I know that because I got the fans on my ThinkPad (as well as home notebook from Hewlett-Packard) to start in similar circumstances.

Is the quiet and extra battery life worth nearly a $900 premium? In a word, no, but you've got to look at the future. Although in the price stratosphere now, flash drives will start to compete more directly with drives over the next four years. Flash memory density continues to increase at a rapid pace, doubling almost every year, and large manufacturers like Samsung, Toshiba, SanDisk and Intel have or are opening factories geared at churning out flash. Taken together, this will lead to an easy availability of chips, better capabilities, and recurring price wars.

Flash prices dropped 50 percent in 2006. Prices rose a bit in 2007, but then dropped 50 percent from the fourth quarter of 2007 to the first quarter of 2008, says Jim Handy of Objective Analysis. Hardware manufacturers can now buy 1GB of flash for $3, he added.

When the premium becomes more acceptable, say $100, the category could take off. The lack of noise isn't one of the benefits I expected, but it was tangible. Listening to the drive on my IBM always prompts two thoughts. One, turning on a PC takes more time than it should and, two, this thing could collapse at any moment. To be honest, the ThinkPad has never imploded because of a hard drive problem, but the internal clanking makes it sound like it could. Silence gets rid of a minor aggravation.

Flash drives also boost performance, although less than I expected. The Dell with the flash drive takes anywhere from 1 to 6 seconds to come out of standby mode, depending on what types of applications were left on. Occasionally, the video that was playing when the computer was put into standby mode starts again. The ThinkPad takes at least 12 seconds.

Starting the Dell after a complete shutdown takes 19 seconds. It takes the ThinkPad 45 seconds to get to the part where I can enter a password. After the password is entered, it takes another 55 seconds before the computer is operational. Some of the slower times on the ThinkPad can be attributed to a slower processor and a more ornate start-up cycle. Even if you don't take that into account, the flash advantage only comes to seconds.

"If it takes one and a half minutes versus two minutes to boot up, are you going to care?" asked Handy.

Weirdly, shutting down both computers takes about the same amount of time. (Flash drives can take minutes off the launch of Outlook, but I couldn't test it because of network problems.)

The drive itself. Samsung puts its flash into a bay that would ordinarily accommodate a much larger 2.5-inch drive. As the market takes off, Samsung will chop down the size of this module.

(Credit: Michael Kanellos/CNET News.com)

Battery power is tough to compare. The Dell has a larger battery pack than the ThinkPad. The ThinkPad is also much older. Still, the Dell with the flash drive seems to last longer than notebooks with standard drives. Fully charged, the battery says it will go five and a half hours, and the time remaining on the battery seems to follow the clock, i.e. an hour of battery time nearly comes to 60 minutes when few applications are on. With eight video streams, the five hours drops to two, but then kicks back up as windows are closed. Handy noted that a flash drive might consume a watt of power while a fast drive might consume 12 watts.

The drawback is the price. The same Latitude with an 80GB standard hard drive currently sells for $869 on Dell's site. Swapping the drive for a 64GB flash hard drive adds $899 to the price. The upgrade more than doubles the price of the notebook to $1,768 and slightly eliminates storage. That's down from the $920 price for the flash drive a few months ago, but out of reach of most buyers. (And it's worse at other vendors. Apple, which started offering flash drives after other PC makers, sells its 64GB flash drive upgrade for $999.)

Photos and high-definition video, among other applications, is also boosting the need for storage, which can favor hard drive makers. Samsung, among others, believes that corporate buyers only need around 64GB of storage, which will be economical to provide in flash in a few years. Consumer laptops, however, come with 160GB to 500GB of storage; 500GB of flash may not be reasonably affordable until 2012, and then consumers might need terabytes.

But if you can offload files onto a backup hard drive, flash could work for you.

Samsung is touting the reliability of solid-state drives, while citing an explosive market for the devices in server computers.

SSDs are based on flash memory chip technology and have no moving parts. Hard-disk drives (HDDs), in contrast, use read-write heads that hover over spinning platters to access and record data. With no moving parts, SSDs avoid both the risk of mechanical failure and the mechanical delays of HDDs. Therefore, SSDs are generally faster and more reliable. The catch is the cost: SSDs are currently much more expensive than HDDs.

Samsung 1.8-inch SSD

(Credit: Samsung)

There are also concerns about wear. That is, flash has the potential to wear out after tens (or hundreds) of thousands of write cycles.

This characterization, however, is too simplistic, according to Michael Yang, flash marketing manager at Samsung. A flash device that is rated at 100,000 write cycles, for example, can write 100,000 times "to every single (memory) cell within the device," Yang said. In other words, the device doesn't write to the same cell over and over again but spreads out the writes over many different cells. This is achieved through "wear leveling," which is carried out by the SSD's controller, he said.

This would make it virtually impossible to wear out a flash chip. Yang said a pattern could be perpetually repeated in which a 64GB SSD is completely filled with data, erased, filled again, then erased again every hour of every day for years, and the user still wouldn't reach the theoretical write limit. He added that if a failure ever does occur, it will not occur in the flash chip itself but in the controller.

On another topic, Yang cited explosive demand in the enterprise server market that caught his company by surprise. "At first it just sounded like an interesting idea," he said. But then demand took off. As Yang explained, companies like Citibank and American Express peg server performance on IOPS or input/output operations per second. "HDDs do 120 to 150 IOPS. SSDs 10,000 to 30,000 IOPS." Because of this overwhelming speed advantage many large corporate customers are opting for SSDs, despite the significant price premium SSDs command compared with HDDs.

Regarding cost, Yang expects to see a 35 percent to 45 percent year-to-year drop in SSD prices. This will be a welcome relief since 64GB SSDs currently can add as much as $900 to the price of a notebook PC.

In the third quarter, Samsung is slated to bring out a 128GB SSD based on MLC (multi-level cell) technology--which uses multiple levels per cell to allow more bits to be stored. But the company sees even larger-capacity SSDs, ranging all the way up to 250GB, possibly before the end of the year.

The company is also working with notebook PC makers to design ultrathin notebooks with SSDs that can fit into potentially even thinner designs than the 0.76-inch thick MacBook Air, which uses SSD.