REGISTER | LOGIN
Breaking News
Blog

Intel 3D XPoint Found in Optane Module

NO RATINGS
View Comments: Newest First | Oldest First | Threaded View
Ron Neale
User Rank
Author
Re: PCM! Not ReRAM.
Ron Neale   6/14/2017 8:25:39 AM
NO RATINGS
The "brand new" could relate to the way the Optane(R) PCM is operated in a dual crystalization mode to speed up the write (SET) time. I have discussed this possibility in my latest contribution to EETimes.

http://www.eetimes.com/author.asp?section_id=36&doc_id=1331894&

tankmemory
User Rank
Author
Re: Impressions
tankmemory   6/13/2017 5:01:27 PM
NO RATINGS
Differences would be because of processes and materials used in manufacturing I would think.

resistion
User Rank
Author
Re: Impressions
resistion   6/9/2017 11:52:37 PM
NO RATINGS
The price of a 3D XPoint chip is not easy to pin down, since it is not sold standalone (yet). But 20 nm DRAM does not have as many SADP layers and the materials are less exotic, so I would expect it to be cheaper to make than 3D XPoint. But 3D XPoint is likely cheaper than a 7nm foundry processor chip.

double-o-nothing
User Rank
Author
Re: Impressions
double-o-nothing   6/9/2017 10:43:05 PM
NO RATINGS
3D XPoint expected cheaper than DRAM: http://www.computerworld.com/article/3194147/data-storage/faq-3d-xpoint-memory-nand-flash-killer-or-dram-replacement.html

resistion
User Rank
Author
Re: Impressions
resistion   6/9/2017 3:41:13 AM
NO RATINGS
The FEOL looks much looser indeed. But the lower tier of the 3D crosspoint array is also quite different from the upper tier. For example, you can see gaps, there are width differences, etc.

Shinobee
User Rank
Manager
Re: Impressions
Shinobee   6/9/2017 3:00:47 AM
NO RATINGS
The size difference may be due to the fact that the FEOL appears to be made on a much older larger process node.  This is possible probably because there appears to be only 2 layers of memory cells which reduces the number of column decoding circuitry required.

It also seems to confirm my suspicion that n layers of 3dxpoint memory cells requires at least 2n+1 lithography + SADP steps, which I think puts a pretty heavy limit on its vertical scalability.

resistion
User Rank
Author
Impressions
resistion   6/9/2017 12:50:43 AM
NO RATINGS
Well, I would be impressed if they can pull this off and call it a relatively "cheap" process. Obviously, there are the equivalent of four 20 nm half-pitch SADP layers here (two tiers, two directions each). That said, there are quite obviously differences between the upper and lower tiers.

resistion
User Rank
Author
Re: PCM! Not ReRAM.
resistion   6/8/2017 11:40:48 PM
NO RATINGS
There was outright denial at the announcement Q&A, abominable. Otherwise, how could there be considerations of ReRAM or other memories?

jeongdongchoe
User Rank
Blogger
PCM! Not ReRAM.
jeongdongchoe   6/8/2017 10:58:56 PM
NO RATINGS
What is the memory element for XPoint? It's a question that many of us have asked to Intel and Micron. Now we found that it's a PCM (Phase Change Memory), not the other memory types such as ReRAM. Two years ago, Micron announced at FMS that it was a brand new PCM. We found it's a GST-based PCM. It's not brand new, from PCM materials view point. However, since it has a double stacked and double layered 'PCM + OTS' memory/selector structure, we would say it's a brand new one.        

Most Recent Comments
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed