News Micron's Fab Goes Offline for One Hour, DRAM Prices Go Up

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

jkflipflop98

Distinguished
Each layer of a chip requires several steps to make, some of these steps take a few hours and each chip has 10+ layers. Hours add up. If a single wafer went from start to finish, it could probably get through in a week or so but handling wafers one by one is grossly inefficient. For volume, you want to run batches of wafers through each process for higher throughput and better repeatability.


To expound upon this, it takes 10,000 - 20,000 process layers to make a wafer of DRAM. When we say "it has 10 layers", that's how many layers of interconnects there are in the backend of the process line. Each one of those interconnect "layers" is 100's of operations to create. It takes over 10,000 process operations to make the transistors in the front end alone. And it really does require 2.5 months of process time within the fab facility to go from a blank wafer to a fully functional part. It wouldn't matter if you only had a single wafer in the entire facility, it would still take months.
 

bkuhl

Distinguished
Dec 31, 2007
24
19
18,515
This is silly. 1 hour outage and prices trend upwards. Shows that people are just very antsy.

Not when you consider that a 1 hour outage could effectively scrap 2.5 months of production at that factory and that factory produces approx 10% of the worlds production.
That could be pretty significant. RAM is usually not stock-piled, its usually delivered JIT.
 

InvalidError

Titan
Moderator
Not when you consider that a 1 hour outage could effectively scrap 2.5 months of production at that factory and that factory produces approx 10% of the worlds production.
The outage mainly affects wafers that were undergoing a critical process at the time of the outage and a couple more wafers during restart for calibration check purposes. Most wafers in a plant are in transit between stations where they are safe from most disruptions, which is why the disruption is only expected to affect production for a few days instead of two whole months.