Hi All,
This was sent to me on Friday. Do any long term rmbs holders have any opions on this subject. Looking for feedback. Been a long for three years.
Friday Feature / OUM: Memory Just Right
‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾
George Gilder (2/9/06): Three years ago, GTR Editor Nick Tredennick surveyed the field of candidates for the winning next-generation memory technology. The ideal memory would be non-volatile (holding its state when the power goes off). It would be fast enough to displace ordinary dynamic random access memory (DRAM), which reads and writes at around 50 nanoseconds. And it would be dense enough to compete with Flash memory, now made in 4-gigabit chips.
Among many contenders, including IBM's (IBM) millipede, carbon nanotubes, holographic memories, atomic resolution memories, organic compounds, quantum dots, and polymers, Nick focused on three chief candidates: FRAM (ferroelectric RAM), from Ramtron (RMTR), Fujitsu, Hitachi (HIT), Toshiba, and 15 others; MRAM (magnetic spin), then being entertained by the same Japanese, Motorola (MOT), Cypress (CY), Infineon (IFX), and NVE Corp. (NVEC), among 15 others; and OUM (ovonic unified memory), espoused by the Ovonyx spinout from hapless solar energy innovator Energy Conversion Devices (ENER) with some fun money from Intel Capital and support from a company called Azalea Microelectronics.
Nick reported that MRAM is the fastest, has the most backers, and gets the most press. But it seems hard to build—it doesn't integrate well with CMOS and could be difficult to scale at a pace that matches the improvement in the CMOS process.
FRAM is the only candidate with real products in the market and Fujitsu controllers show that it integrates well with CMOS. But its large cell size may prove a fatal liability competing with the incumbent flash and DRAM technologies, as any major new memory must do.
The winner, wrote Nick, would be OUM. “It integrates well with CMOS. It has low power reads, it operates at low voltages, and its control circuits are simple. It's potentially as dense and cheap as DRAM. As Goldilocks would say (and as Gordon Moore of Intel noted decades ago): “It's just right.”
Now Intel (INTC), Samsung, and STMicro (STM) are all planning to produce the chip in some volumes later this year. All have licences from Ovonyx, which is 41.7% owned by Energy Conversion Devices (ENER). Running Ovonyx is Tyler Lowrey, previously responsible for many of Micron's major breakthroughs. Micron founders Ward and Joe Parkinson have been engaged in Ovonyx from the outset.
GTR Subscriber (2/8/06): I have not seen Saifun’s (SFUN) nitrided read only memory (NROM) technology discussed in any evaluation of next-gen memories. Any thoughts on this one?
George Gilder (2/9/06): Saifun is an evolutionary advance in ordinary flash technology, as MOSIS is an extension of basic static RAM technology, and Rambus (RMBS) is a marginal upgrade of DRAM technology. Phase change from Ovonyx is a radically new and better form of memory that combines the virtues of DRAM, flash, and SRAM.
But I must say that reduction of the likely increment to the bottom line of ENER toward $1B is cautionary. ENER is a horrendous muddle of a company. GTR Analyst Charlie Burger was verging verdant as he ruminated its numbers. But Ovonyx may well be a semiconductor winner comparable to Micron (MU) or more specifically a Rambus with a real invention.
ENER may well have 39% of the IP yield for the world’s highest volume product. It's a quandary. Unlike ENER, Ovonyx has great management. It should spin out and go public or it may well find itself serving as a Trojan horse for amorphous ovoid moonbeams.
Paul McWilliams, Editor, Next Inning Technology (2/9/06): I think the guys at Ovonyx are really pushing the $100B potential a little too hard. What they are counting is the FULL value of every single semiconductor that has a memory cell. Last year MOS memory sales were about $47B. The rest of the $100B includes programmable logic, programmable DSP, programmable micro-controllers and a host of other parts where the value of the part is 99.9% non-memory. Always remember, the life-cycle profitability of a semiconductor is inversely related to the area of the part that is covered by memory.
AI
