Intel looks at leap in handheld memory
Intel is betting that consumers want handheld computers stuffed with 500MB of memory. On Wednesday, the chipmaker described the search for what it calls the "Holy Grail" of mobile memory, with a new technology that will pack hundreds of megabytes of storage into mobile devices at a low cost. A typical handheld, for example, now has 2MB to 64MB of flash memory.
In Intel's vision, the next-generation memory should combine high density--which translates into greater storage capacity--with high performance and the ability to be easily integrated with other types of memory. The new memory must also be non-volatile like a PC hard drive, meaning that data stored inside is not erased once the power on the device is turned off.
So far, Intel has identified three promising technologies: "plastic" memory, or polymer ferroelectric RAM (PFRAM); Ovonics Unified Memory (OUM), which uses the same materials as rewriteable CDs; and magnetic RAM, or MRAM.
Developing new memory technology is vital to the mobile device industry, which seeks to boost the functionality of mobile devices, including cell phones and handhelds, by adding new applications such as video streaming. These applications, however, require larger amounts of memory.
Research efforts are also important to Intel's bottom line. Although revenue from flash memory has been hit hard by the slow economy--especially due to lagging cell phone sales--it represents a large chunk of Intel's revenue. Finding a hit in the next round of memory technologies would ensure that the company would continue to create new sources of revenue.
"The industry has really been searching for what would be considered the Holy Grail" of memory for mobile devices, Stefan Lai, vice president of Intel's Technology and Manufacturing Group, told reporters Wednesday.
SRAM, DRAM and flash
Intel is specifically looking for technology that will combine higher density with the fast read/write speeds of synchronous RAM, the lower cost of dynamic RAM and the non-volatility of flash memory, which can store data when a device is turned off.
DRAM and flash memory are among the main types of storage now used in mobile devices.
Of the three new technologies Intel is researching, OUM is likely the most promising for its combination of low cost and ease of integration with other memory technologies. MRAM and PFRAM each offer faster performance in terms of read/write speeds, though they do so at a higher cost.
Analysts say that handheld makers are definitely interested in a new type of memory that is cheaper than flash.
Intel is "closing in on a real solution" to the issues of price and performance, said Richard Doherty, director of research at The Envisioneering Group.
Doherty sees OUM as the most promising of the possibilities for next-generation memory.
Intel would not place timelines on its next-generation memory. But Doherty said the company could have next-generation products based on OUM in as soon as 30 months or at the end of 2003. "It could be cheaper than flash within a couple of years," he added.
However, PFRAM also has a good chance of becoming popular, even though it can stand up to only a finite number of writes.
"Someone else might find the trick to it," he said. "OUM progress might trigger some new interest in (PFRAM) just to have something to be competitive with Intel."
Simultaneous research
Intel is running simultaneous research programs designed to investigate the three technologies more thoroughly, but it would not disclose plans for producing any of them. Company executives said Intel will offer an update on its memory research efforts later in the year.
According to Intel, OUM could offer good enough performance in terms of price and read/write performance to take over the majority of storage duties for devices such as handhelds.
OUM uses chalcogenide, the same material used for rewriteable CDs. OUM is "really complementary to SRAM," Lai said. "You can use (OUM) as main memory, and it does not require special software. We believe that for portable applications this is a really good application."
Intel could also "conceive of a system where an integrated chip has some SRAM and some OUM," Lai said.
Intel has a working OUM design, which it has used to build test chips. The company is testing their performance, ease of manufacturing, and other parameters that can measure the technology's real-world viability.
"This is something we are looking at developing further...to make it a lower-cost data-storage memory," Lai said.
Still a horse race
Meanwhile, MRAM offers faster read/write speeds--below 10 nanoseconds--and can endure almost unlimited write cycles. However, its cell size, the size of the smallest unit of a memory chip, is larger than OUM, "making it three to four times more expensive than OUM," Lai said.
Smaller cell sizes make for higher densities, which translate into greater storage capacity per chip.
"MRAM is like an SRAM. It has high function but is more expensive. OUM is like DRAM--slower but cheap," Lai said.
Intel is also exploring PFRAM, which uses layers of polymers attached to a base built using complementary metal oxide semiconductor (CMOS), which is standard chip material.
PFRAM is expected to create dense memory chips that will be fairly easy to manufacture, requiring only minor changes in Intel's existing chip plants. These characteristics will make PFRAM relatively inexpensive.
Using PFRAM, Lai said, "we can make a memory that is about one-eighth the cost of what you would have in a basic CMOS type of memory."
However, the technology has a finite number of write cycles. Because of this, it is likely that PFRAM will be used less often, performing jobs such as storing applications, much like ROM, or read-only memory, is used today.
For its part, Intel says there is promise in all three types of memory.
"At this time, we are looking at all the memories," Lai said. "It's a horse race at this point. Anything can happen. There's still a lot of learning we need to do."
