Power reduction and not performance now driving semiconductor design
http://www.electroiq.com/articles/sst/2012/10/why-node-shrinks-are-no-longer-offsetting-equipment-costs.html
Unfortunately, that 30% reduction from a 450mm wafer-size transition "represents about only three or four years of increasing equipment costs, and consequently, delays the inevitable," the analysts say. And even if new advanced technologies pan out as promised and reduce costs -- and that's a big "if," getting semiconductors down through the teens and into the single-digit nodes -- that too will be only a temporary reprieve: "The reality is that rising costs are a permanent part of the industry, and the fundamental economics of the industry may start changing sooner rather than later.
Lower costs per transistor aren't even the real driving factor for semiconductor manufacturers, it's reducing power consumption while maintaining performance, the Gartner analysts note. That's the key functionality driving semiconductor content in mobile devices, and that's what is driving the market now. (It's a little different for memory which does rely on lower unit/production costs even if it's a slight improvement.) "Overall, the semiconductor and electronics industry will have to come to grips with the fact that traditional cost reductions with each new node are in jeopardy, and in the future, higher performance may very well come at a higher price," the analysts note.
And one more factor to consider: semiconductor average selling prices (ASPs) have benefitted from the relentless node-shrink cost reductions, but semiconductor manufacturers are unlikely to (nor should they) swallow all these cost increases by themselves for the good of the rest of the supply chain. "What is more likely is that ASP trends will reverse after years of decline, and that reversal will change the supply-and-demand economics of the industry in ways that we don't really understand today," the Gartner analysts admit.
Basically, chipmaking is getting too expensive to do anything about actually raising performance. Also, the chips themselves may become more expensive. The cost consequences alone can bring the performance increase effect of Moore's Law to a screeching halt. I don't see a forced increase in ASPs as being in any way conducive to demand. Chip manufacturers have no choice but to absorb the cost increase if they want to retain demand, which is already bad enough the way it is now.
Unfortunately, that 30% reduction from a 450mm wafer-size transition "represents about only three or four years of increasing equipment costs, and consequently, delays the inevitable," the analysts say. And even if new advanced technologies pan out as promised and reduce costs -- and that's a big "if," getting semiconductors down through the teens and into the single-digit nodes -- that too will be only a temporary reprieve: "The reality is that rising costs are a permanent part of the industry, and the fundamental economics of the industry may start changing sooner rather than later.
Lower costs per transistor aren't even the real driving factor for semiconductor manufacturers, it's reducing power consumption while maintaining performance, the Gartner analysts note. That's the key functionality driving semiconductor content in mobile devices, and that's what is driving the market now. (It's a little different for memory which does rely on lower unit/production costs even if it's a slight improvement.) "Overall, the semiconductor and electronics industry will have to come to grips with the fact that traditional cost reductions with each new node are in jeopardy, and in the future, higher performance may very well come at a higher price," the analysts note.
And one more factor to consider: semiconductor average selling prices (ASPs) have benefitted from the relentless node-shrink cost reductions, but semiconductor manufacturers are unlikely to (nor should they) swallow all these cost increases by themselves for the good of the rest of the supply chain. "What is more likely is that ASP trends will reverse after years of decline, and that reversal will change the supply-and-demand economics of the industry in ways that we don't really understand today," the Gartner analysts admit.
Basically, chipmaking is getting too expensive to do anything about actually raising performance. Also, the chips themselves may become more expensive. The cost consequences alone can bring the performance increase effect of Moore's Law to a screeching halt. I don't see a forced increase in ASPs as being in any way conducive to demand. Chip manufacturers have no choice but to absorb the cost increase if they want to retain demand, which is already bad enough the way it is now.