This kind of manufacturing cost data is pretty hard to reconcile with DRAM makers' recent high profitablity. For sure, Sam, Hyn, MU and Elp have reduced their manufacturing costs substantially for product they make themselves; only their Taiwanese partners (Ino, Nan, Pchp) have had significant process node transition glitches that would temporarily raise costs. Net net: I am skeptical of any industry-wide assertions of rising DRAM manufacturing costs
If you have been tracking the progress at nano-science field. there has been no major progress (in size) in last 10 years sth. most people (or everyone) stopped at 10nm (a single wall NT).
look back at history, so many things will hit a limit soon or later, cars, air plane, rockets... drams..
Does anybody know any specific physical/technical reason?
I heard that DRAM capacitor stopped scaling because it's not possible to make dielectric any thiner. Of course, somebody might figure out how to different material but it could require a lot of time and money.
I bet some of them still believe in the science will never end theory, and the trend is working cool till recently.
some of them could be getting a plan B, ie, in case tech stops at xx nm.
yeah, i think plan B should be interesting, not much tech difference between me and you, pure cost...
will dram bussiness become more like motherboard bussiness in the future?
insider info: 40nm is tough now,
if you can make something with 10% yield it's useless.
Moore's law won't run forever. the bottom line is ... 10nm? and the cost curve will hit at ??? nm.
good news for fabs in china etc. the tech advance is slowing down, u can catch up soon now.
It is going to fall in the next two Qs if the economy does not stabilize. But if it continues to grow, that will cause a major dislocation in the DRAM business model and what it is all said about Moore's law
Blog Doing Math in FPGAs Tom Burke 18 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...