At last week’s Analyst Conference, Micron’s VP of R&D, Scott DeBoer, provided an update (link downloads presentation pdf) on Micron’s Next Generation Memory Technology (entitled ‘Accelerating Next-Generation Technology’). It is 6 months since the 3D Xpoint announcement last July. I was curious to see how Micron’s cryptic roadmap had changed and whether 3D Xpoint would be revealed as memory A or Memory B from previous roadmaps in these Analysts’s presentation. It turns out I was to be disappointed!
Scott described successful 2015 Technology deployments of 3D NAND (gen 1) to Fab 10 in Singapore and 3D XPoint introduction to Fab 2 in Lehi, Idaho. 3D Gen 2 (sorry, Gen 1 Rev 2) is underway with an unspecified TLC endurance improvement putting it well above 16nm planar. Later in the presentation 3D NAND Gen 2 is described as having 30% cost reduction. So it looks like Rev 2 is related to a cell level improvement and Gen 2 is related to more layers (and scheduled for Summer 2016).A most impressive TEM is shown labelled 32 layers; there are actually 38 which presumably means there are a selector and two dummies at either end. CMOS is clearly shown under the 3D NAND cell array (which is why selectors are needed in the 3D array itself). Micron have been confident enough to show a wafer yield map claiming >90% yield. There seem to be a few non-yielding die around the edge and a cluster just off center. Nonetheless impressive if it is indeed a typical result.
There is no such wafer yield map for 3D Xpoint but ‘Rapid yield and manufacturing capability improvements’ are claimed along with ‘Currently sampling with select partners’. So there must be some chips being exercised in various applications out there! Deployment (interesting choice of words) is a key technology focus area for 2016.
So, any clues as to what Micron has referred to as New Memory A and New Memory B in the past? Is 3D Xpoint A or B? Well it appears it is neither as the following graphic from the presentation suggests under the title of ‘Future Memory Technology Options’.
So New Memory A is still the DRAM alternative and New Memory B the NAND alternative.
Christie Marrian ReRAM Forum Director.
p.s Andy Walker has some predictions for 3D NAND in 2016 (and is bold enough to revisit his 2015 predictions). He has unearthed a Micron patent application from 2013 which describes how you might manufacture 3D floating gate NAND cells. It’s not simple and I’m sure it is actually even more complex than the application describes.