The 9 microsecond number pertains to an NVMe SSD containing 3D Xpoint chips. That's incomparable to DDR3 DRAM used as system memory. 3D Xpoint is said to have a latency in the "10s of nanoseconds", which is slower than DDR3 or DDR4 but not terribly so.
3D Guy: Before dealing with the architecture and processing problems at the block level it would be a little easier to assess the possibility of success if we had from INTEL some answers to a few questions at the stacked cell level.
Is there still a heater electrode or is a close coupled threshold switching filament used for that purpose?
Is crystallizesd active material still used as one electrode in the memory device?
What is the current density during RESET?
What is the melting temperature of the new recipe PCM material?
Is the threshold switch matrix isolation device deposited in the amorphous state if so does it require forming?
Is there a claim for bulk switching and "S" shaped negative resistance threshold switching without hotspots or filament formation?
Does the matrix isolation device threshold switch on every read cycle.
Does the cell design include any form of thermal barrier?
The lithographic node?
Cheer leader in waiting!!!
Rather than "marketing teams going wild" or an attempt at some form of juvenile deception, one interpretation might be they got caught up with the need to demonstrate Intel's diversification in light of the PC slowdown and now have to back off a bit.
Rebranded phase change memory with an OTS, as many of us knew. The challenges are enormous. Etch, clean, number of litho steps, low block sizes due to high current are just a few. If Intel and Micron struggled to get simpler 2D PCM to yield well, good luck trying to get 3D PCM to yield! This 3D XPoint PR, where they tried to not brand it as PCM to get more interest, is a classic example of marketing teams going wild.
@Ron Neale Yeah, it's strange they said in July they were in production because clearly they could not have been except for internal only samples at best. Others have noted it was strange they said it was not phase change and now admit to using the materials and techniques folks like you surmised earlier.
Blalock was indeed frank about the challenges -- from cross contamination to costs and single-sourced materials.
Could they fall on their faces? I suppose. But a separate Intel packaging talk showed 3-4 capex systems Intel developed in collaboration with partners. So Intel has the will to build production gear if they need to to get this stuff out the door -- and from the sound of it, they may need to at least in dry etch.
Any new projects in Engineering usually take longer to develop, XPoint is not any different, I thought Mr. Blalock gave a quite detail lowdown on all aspects of difficulty relative other older process before. I am glad it did not take longer than C1Q16 to get to the sample stage. If MU/Intel are to keep going at it, they will derive a tidy sum of revenue from these XPoint and 3D NAND in 2016, Cheers!
Rick- Something strange or is it it proving much more difficult than at first thought ""give the R&D guys a little more time to work out the kinks," he said. "" Yet the the 2015 press release that announced 3DXpoint had the following bullet points:
Intel and Micron begin production on new class of non-volatile memory, creating the first new memory category in more than 25 years.
Now in production, 3D XPoint technology is a major breakthrough in memory process technology and the first new memory category since the introduction of NAND flash in 1989.
3D XPoint technology will sample later this year with select customers, and Intel and Micron are developing individual products based on the technology.
Are the "kinks" in performance or fabrication and production? Are they new "kinks" or those that have been present for the past 50 years.
@resistion Blalock said Intel and Micron had to keep details quite in July when it was still in a late research phase. Now that it is moving into the fab third parties such as gear makers need to know more about the technology to work with them, so details are starting to come out.
Of course when chips ship and teardowns happen even more will be revealed
BTW, Blalock said XPoint manufacturing will ultimately be extended to a Micron fab in Singapore and Intel's Dalian fab, too.