REGISTER | LOGIN
Breaking News
Blog

Imagining What’s Inside 3D XPoint

NO RATINGS
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 3   >   >>
resistion
User Rank
Author
Re: chalcogenide vs perovskite PCM
resistion   9/6/2015 7:15:13 PM
NO RATINGS
Maybe chalcogenides have less severe phase separation?

Ron Neale
User Rank
Author
Re: Partial barrier coverage leads to filamentary
Ron Neale   9/6/2015 2:56:28 PM
NO RATINGS
Resistion: The confirming clue maybe in the statement from Intel that the 3DXPoint (tm) devices were temperature sensitive. Most of the strong resistance transitions for the perovskites and magnetites are somewhere in the range 230K and below. If they have found a way of  extending the temperature range then that might make the devices marginal on temperature. If Intel have gone that materials route then I think, if memory serves me, it is also possib;le to make pn junctions from the perovskites so it might be possible to stay with the same family of materials for memory and matrix isolation at each cross-point stack.

Looks as though we might be heading back into CeRAM territory and correlated electron effects.  Thanks for the reference,

resistion
User Rank
Author
Re: Partial barrier coverage leads to filamentary
resistion   9/6/2015 10:41:19 AM
NO RATINGS
I attach this reference: http://arxiv.org/pdf/cond-mat/9909349v1.pdf,

although anyone can find more by searching for manganite phase separation. It is an interesting field dating since the beginning of the millenium or even a little earlier.

 

Ron Neale
User Rank
Author
Re: More like a tanatalum capacitor
Ron Neale   9/6/2015 7:38:30 AM
NO RATINGS
Resistion: If you are right, my first guess would be (Ag)m-La Ca MnO3, where silver is the mobile (m) species. Manganese Trioxide switches from its conducting to its insulating state when in contact with silver. You can replace the rare earth with others and the Ca with Sr, Ba, or Pb. and still have a perovskite.
The evidence of switching of MnO3 between its conducting and insulating state takes us a long way back to the early days of tantalum capacitors. For those devices conducting MnO3 was used as the conductor to the thin TaO film on the tantalum wire. When, to keep the series resistance low, silver was tried as a bulk electrode, with applied voltage it would cause a phase change of the MnO3 to its insulating state. That is why the early Ta wire capacitors used carbon as the buffer electrode to the MnO3. We know in other ReRAM/RRAMs silver can be a mobile species, perhaps for the 3DXPoint ™ and if it is a perovskite, Intel have found some some combination that can reverses the MnO3 phase change. All that assuming you do not mean some form of vertical colossal magnetic resistance (CMR) device.

If I move away from the Alice and Rabbit world of imagining and devices, although I must stress I have not seen the memo, I wonder if to demonstrate Intel diversification away from a PC company the word went out all divisions must put on a diversification show- in my view the rather pathetic 3DXpoint ™ presentation was the result. Including the statement it is not phase change and does not use electrons.

 

resistion
User Rank
Author
Re: Partial barrier coverage leads to filamentary
resistion   9/6/2015 1:31:14 AM
NO RATINGS
These perovskite or manganite materials suggested as non-PCM alternative explanation for 3D XPoint have been popular for CMR and show phase separation, which is an interesting and ironic reference back to these being inhomogeneous "phase change" materials after all.

Ron Neale
User Rank
Author
Re: Get rid of the crystal electrode- fast write
Ron Neale   8/27/2015 12:18:26 PM
NO RATINGS
Resistion: I think what you are suggesting is something like polymorphic states. Would that require a bidirectional write/erase pulse?  Don't know.

When I was writing this piece and doing my imagining I did think of the possibility that they might have gone back to an amorphous material with a lot of nucleating sites. Without the upper electrode being crystallized material. As I illustrated in my other "Why it's not PCM" paper with my graph that links SET  time to data retention time, when you grow the SET state from a single massive nucleating site, the upper electrode, you need high temepratures to achieve the sort of crystal growth rates required for short write times.

If you could embed  some material that would act as nano nucleating sites and would not be reset, less that 33% volume fraction, then even the slowest crystal growth rates would give a fast write time fiilling the gaps between. I suppose it is possible that heat treament could precipitate out a crystal species that would not be RESET while the materila surrounding it was RESET. All the atoms involved all the time rather than growth of a crystal front across a gap.

resistion
User Rank
Author
Two states of different bulk density = two different phases
resistion   8/27/2015 11:12:54 AM
NO RATINGS
They were saying all the atoms were affected in the resistance change. So I take it the bulk atomic density is different for different states. With different lattice constants, it really suggests different phases. Maybe not exactly the same amorphous vs. crystalline phases of GST. But necessarily a Tg (glass transition temperature) involved.

resistion
User Rank
Author
Re: Unity CMOx --> Rambus --> Micron-> PhasOx
resistion   8/25/2015 6:39:45 AM
NO RATINGS
That's hilarious! I love it, Ron.

Ron Neale
User Rank
Author
Re: Unity CMOx --> Rambus --> Micron-> PhasOx
Ron Neale   8/25/2015 4:52:10 AM
NO RATINGS
Resistion: For the moment I think we should call it "PhasOx"  as per my cartoon in the response  below.

Ron Neale
User Rank
Author
Re: Asking for information
Ron Neale   8/24/2015 4:56:07 PM
ROHIT115 At the moment the question you ask is one of the most difficult to answer. If you read all the recent articles published in EETimes about 3DXPoint (tm) you will get some idea of how little about the detail is known. If we construct a list of the knowns and unknowns from quotes etc. it looks something like this:

Materials                         * New (secret) Recipe
and Composition              *Based on the invention of
                                        unique material compounds

Switching                          *Not phase change
mechanism                       *Bulk memory effect uses all
                                         the atoms.
                                       *It's not a pure phase change
                                          because it uses bulk material properties.

Cross point                       *At least two materials
stack structure                      plus X-Y conductor electrodes

Performance                      * 1000 x faster than Flash.
                                        *1000x Write/erase endurance.
                                        *10x denser than Flash.
                                        *128G-bit memory in production.
                                        *Temperature sensitive.
                                        *Access time latency?
                                        * Data retention time equal Flash
                                        * Needs something similar to
                                           wear levelling algorithm (thermal?)
                                        *128G-bit memory in production.
                                        * Power dissipation of new chip?
                                         * Operating temperature range?

It is sad for a great company like Intel that when one looks at the information available to date what comes to mind is:



 

Page 1 / 3   >   >>
Most Recent Comments
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed