Design Article
Universal memory, round two
Paul Boldt and Don Scansen
11/8/2009 12:01 AM EST
The success of any technology should be measured years after its first adoption, not when it is fresh out of the development lab. Many technologically advanced products fail, while many simpler ones generate high revenues over many years. Predicting the future for any emerging technology requires speculation based on a relatively small amount of information available on the new devices and a purely historical perspective of similar products that cannot hope to anticipate future events and human behaviors. Only time will tell.
Discussion in the labs and in the press around the so-called universal memory has been under way for perhaps 10 years. While that may not sound like a long time, it is an eternity in the semiconductor world. Consider that in 1999 256-Mbyte DRAM was considered high-end, and 250-nanometer logic processes were on the forefront, if not quite the bleeding edge, of process technology. In consumer products, Apple had just deleted the 1.44-Mbyte floppy drive on its iMac line of computers, while Iomega's 100-Mbyte Zip Drive was wildly popular. USB drives were in their infancy, with densities in the 8-Mbyte range and the typically high prices associated with a new technology. (Contrast that to the present, when 4-Gbyte USB drives are included, along with the more traditional pencils and rulers, as required elements on a sixth grader's school supply list.)
|
| Structure of Freescale MRAM cell. Source: Chipworks |
This trip down memory lane is more than mere nostalgia. It was in the heyday of the Zip Drive that three "first round" contenders--PCRAM, MRAM and FeRAM-- were identified as candidates to become the universal memory. Today, we continue to debate the candidates.
Phase-change RAM (PCRAM, or simply PCM, as Numonyx now describes it) has finally been seen in the wild.1 A Chipworks reverse-engineering blog post dated May 11, 2009, begins with the words, "The long-awaited phase-change memory (PCM)," a phrase that pretty much sums up PCRAM's journey to market--although, considering the aggressive marketing and associated press coverage of PCM in recent years, the blogger might have added "much anticipated."
If its other aliases are not forgotten, phase-change memory, aka Ovonic Unified Memory (OUM), harks back to 1970, when a 256-bit version appeared on the cover of Electronics.2 Nowadays, you might also see references to PRAM, intended to suggest that PCRAM is "perfect" RAM.
The Chipworks reverse-engineering analysis of a Numonyx PCRAM provided proof that products were, at long last, available. A cross-section of the memory array identified a PCRAM cell that appeared to comprise at least an "upper" silicide contact and a layer of phase-change material encapsulated within a barrier layer. The structure rests on top of a tungsten plug.
Next: PCRAM




Semiconductor Analyst
11/10/2009 8:32 AM EST
I would not bet against NAND for the foreseeable future. So while lithographic scaling will be a problem sub 32 nm, that problem is faced by everyone. And yes it will be hard to scale the interpoly dielectric and tunnel oxides. But HfO or AlO are being used in DRAMs and microprocessors. And Samsung and Toshiba have been qute clever at migrating to 2, 3, and 4 bits per cell. That and multichip packaging (I think one vendor has shown 9 dies in a stack) would suggest that flash has a way to go before bust.
I don't see any of the other memory technologies as having the kind of infrastructure that flash and DRAM have. The reports of flash/DRAM deaths are quite premature.
Sign in to Reply