Standing the Transistor Up: What 4F² DRAM Changes

For years, mainstream DRAM has used the same basic cell layout. In October, Chinese memory maker CXMT said it will bring out a server memory module built a different way by the end of this year. What actually changes when you stand a transistor on its end? This is an announced plan: no speed, capacity or outside test results have been shown, and no one outside CXMT has tested it. What a DRAM cell is, and what "F" means 6F² vs 4F²: about a third smaller on paper The news, and what we still don't know Samsung and SK hynix are working on 4F² too — we don't say who's first Caveats, and what to watch at year-end Figures are labeled on screen as company claim or report. Not investment advice. Next week: where does the light come from? The laser lives outside. Chapters 0:00 The same cell layout for years 0:34 Stack vs shrink 1:13 What a DRAM cell is 2:02 6F² vs 4F² (on paper) 3:39 The news 4:48 Not from nowhere 5:21 Everyone wants 4F² 6:19 What we still don't know 6:47 Caveats 7:21 Prediction: what changes if it works 8:06 What to watch 8:21 Next week + final thought Sources Tom's Hardware / TrendForce / Bloomberg / DIGITIMES / Seoul Economic Daily / The Herald Business / Yicai (as reported) / Embedded.com / IEDM 2024 short course (Oh) / The Elec (as reported) Update (Oct 2026): At 5:21 we say “everyone wants 4F².” That overstates it. Samsung and SK hynix are still developing 4F². Micron said in a Oct 21 tech webinar it will not adopt 4F², and plans to stretch 6F² then move to 3D DRAM. CXMT’s year-end plan is unchanged; no capacity, speed, or outside tests yet.