Author: S J J
Date: 06:16:56 04/19/05
Go up one level in this thread
I have been very fortunate to be part of the semiconductor industry for the last 25 years and have seen ebbs and flows of the business, but, a continuing ability to low cost and increase speed of devices. However, the trend is clear that the ability of the industry to continue on this path has been reduced and will continue to be reduced. Many people think that the ability to economically scale silicon will end around 2015, others think around 2020. 25 years ago you could build a good fab for $20M. A mask set (the master tooling for a chip design) cost about $1,000 and you could get a wafer for $120. Today, leading fabs can easily cost two to three billion dollars. A mask set will cost $500,000 and you can get a leading edge wafer for $2,500. The financial burden has meant that most companies 25 years ago had their own fab to where only an elite group can have their own fab today. Even the top tier suppliers are now pooling their money to make fabs together to spread out the cost. If the cost of a fab went up by (only) a factor of 10 in the next 10 years, it would cost 20 to 30 Billion dollars. Certainly, the number of leading edge fabs being built will be reduced greatly by 2015. The scale has reduced from 3 microns (10,000 atoms) down to .09 microns, more than a factor of 30. At present levels, transistors are only 300 atoms across. While this sounds big, it is very very difficult to make a reliable and controllable device below 100 atoms. Each step forward will be more difficult and more expensive than the past. Keep in mind that Moore originally predicted a doubling of the number of transistors on a chip every year. That was later revised to every 18 months then every two years. The rate of increase has been slowing for some time. Predictions on the future? IMHO, by 2015 there will be silicon in production at less than .02 microns. Progress beyond that will be very slow. However, since the rate of going to the next generation will be slowed, that will mean that fabs will have a longer useful life. Since they will be depreciated over a higher number of wafers, this will help offset somewhat the higher cost of fabs, masks, etc. In effect, the rate of technology improvement will slow to the rate at which the investment can be economically recovered. Rather than doubling the speed of a processor every 2 years, we will be seeing more multi-core processors as well as putting more processors per system. This will help serve the growing need for speed. Regards, Steve Very few companies in the world can spend billions on a fab.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.