You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Winbond Electronics has kicked off mass production of specialty DRAM memory using 20nm process technology, with plans to transition to a newer 16nm process, according to the Taiwan-based chipmaker.
www.digitimes.com
Winbond is planning a transition to a 16nm DRAM process. Development has been ongoing for a while.
This is their first 1x nm DRAM, yes? How is it that a trailing edge memory manufacturer can avoid getting buried by firms that are using those cheaper new nodes? Is it by selling things like DDR3/4 while others use their lower power and faster speeds to make higher margin DDR5/HBM dies? Certifications for embedded applications that someone like Micron can't get with current process maturity and or won't yet bother getting on something like 1-beta?
Also correct me if I am wrong Fred, but don't you work at Winbond? If so, congratulations! Always exciting to see a node you've spent years working on finally come to fruition.
This is the 3D product, right? The one designed to support vias and custom hybrid bonding to a logic chip on top? The memory industry really needs a "foundry" mentality of being willing to make products to support the new ideas of their customers, without trying to steal the IP of the customers. This looks like an intelligent step in that direction.
This is their first 1x nm DRAM, yes? How is it that a trailing edge memory manufacturer can avoid getting buried by firms that are using those cheaper new nodes? Is it by selling things like DDR3/4 while others use their lower power and faster speeds to make higher margin DDR5/HBM dies? Certifications for embedded applications that someone like Micron can't get with current process maturity and or won't yet bother getting on something like 1-beta?
Also correct me if I am wrong Fred, but don't you work at Winbond? If so, congratulations! Always exciting to see a node you've spent years working on finally come to fruition.
Yes, basically, avoiding the direct competition with commodity DRAM at the bleeding edge nodes, doing specialized applications. Although there is still competition such as from CXMT, etc. Though happy to see Winbond make this transition, I am actually only very indirectly involved with the DRAM goings on.
This is the 3D product, right? The one designed to support vias and custom hybrid bonding to a logic chip on top? The memory industry really needs a "foundry" mentality of being willing to make products to support the new ideas of their customers, without trying to steal the IP of the customers. This looks like an intelligent step in that direction.