PCM might be viable longer than charge trapping. It's golden era might come after floating gate. Charge trapping might be the way in between (I'm not sure how well it scales below 20nm as Spansion hasn't commented on that yet).
Numonyx has no choice as it can't promote Spansion's technology. It has PCM from Intel and Elpida/Intermolecular to promote it with them on the one hand (for NOR), and charge trapping on the other via Hynix (for NAND). It's clear to me that PCM is more viable as a candidate to compete with NOR than with NAND because of the density.
Others (like Samsung, Hynix, etc...) are basically trying to extent floating gate as long as possible with increasingly expensive techniques, hoping the holy grail is found before long, and in some ways PCM could become this holy grail (read comparable to NOR, write comparible to NAND, reliability second to none). In the mean time you can be sure they spend good dollars on developing all kinds of alternatives too. In case of Samsung they're developing both charge trap and PCM in addition to floating gate.
Still for as long as charge trapping and more or less related techniques (so Mirrorbit, SONOS, and all...) would scale with a similar level of difficulty as PCM they should do just fine because as it looks now their cell sizes will be significantly smaller. Still very long term this statement contains a big presumption because we don't know well enough yet how easy it is to scale PCM to newer nodes and if >=4b/c ever could become possible. For the next four years or so it's unlikely to become an issue though.
Not related to cost but also important are performance advances for PCM. PCM will become significantly faster with every process node because the smaller the bit area that needs to be heated the less time it takes to flip bits. The question is what how fast speed will ramp up compared to speed increases for charge trapping. As far as I know this is not publically known yet. If it ever would be needed Spansion can get equal access to Ovonyx and Molecular fafai can see.
For now Spansion has a reasonable window of opportunity with:
- NOR @ 65nm / 300mm before Numonyx
- a mature range of charge trapping products
- a good product roadmap
- a mediocre process node roadmap short term (Numonyx beat them to 45nm by a full year), possibly slightly better process roadmap a node or two after that; not holding my breath here
- a more or less reasonable pricing environment now Numonyx is cut loose from its parents and Samsung is cooperating which has a somewhat reasonable chance to last for a while
- a rather undesirable balance sheet
Still as everyone here knows all too well they managed to keep a couple of previous windows firmly shut... They better execute this time. (I should think this is part of their common cautionary statement for a reason)
Regards,
Rink |
