There are limits to what Moore's law can do.
You could buy a 4 GHz PC processor in 2004. In 2013 you could buy a 5 GHz PC processor. WooHoo! :-/ That period is 6 doublings in Moore's law. While it used to be that clock-speed had a Moore's law-like equivalent, you can't buy a 20+ GHz PC processor. More transistors is great, but there are limits (power dissipation among them - you can't dump more than about 50W into a square cm of Si before it gets too hot to work properly). E.g. the P4 was 217 mm^2 (2.17 cm^2) and maximum power was 115W or about 53 W/cm^2.
Higher density and lower power consumption generally means lower speeds. You're not going to see a 5 GHz MacBook Air anytime soon. ;-)
The ITRS is the bible on where the semiconductor industry thinks it is going and the challenges that need to be solved. The 2013 Roadmap has lots of details.
I'd be very skeptical that the advances we have seen in computing in the last 20 years are going to continue at the same pace for the next 20 years without some breakthroughs. But at some point, things will be "good enough". E.g. I can't see the need for 10,000 dpi video displays. And our brains due pretty well with a "clock speed" measured in Hz rather than GHz. :-) Parallelism can do a lot. More advanced sensors for interacting with computers would be great. But, as we know, the software to take full advantage of parallelism is going to lag...
Plus, there's the assumption that everything will keep getting cheaper the way it did in the past. In the past, Intel had competition from NEC, Zlog, AMD, and lots of other CPU vendors. Now, not so much - AMD is still struggling. Intel has licensed ARM (as has AMD and Apple), but their continuing lead in semiconductor processing technology gives them a big advantage apart from the instruction-set. If it gets to the point that Apple and Intel are the only companies that can make a profit on semiconductors, then don't expect the prices to drop any longer...
(As an aside - I'm continually astounded by all the effort put into voice recognition software and wonder if the developers have ever tried using them in a busy airport terminal. But I guess they haven't come up with anything better for portable phones and automated call centers yet...)
My $0.02.
Thanks.
Cheers,
Scott.
You could buy a 4 GHz PC processor in 2004. In 2013 you could buy a 5 GHz PC processor. WooHoo! :-/ That period is 6 doublings in Moore's law. While it used to be that clock-speed had a Moore's law-like equivalent, you can't buy a 20+ GHz PC processor. More transistors is great, but there are limits (power dissipation among them - you can't dump more than about 50W into a square cm of Si before it gets too hot to work properly). E.g. the P4 was 217 mm^2 (2.17 cm^2) and maximum power was 115W or about 53 W/cm^2.
Higher density and lower power consumption generally means lower speeds. You're not going to see a 5 GHz MacBook Air anytime soon. ;-)
The ITRS is the bible on where the semiconductor industry thinks it is going and the challenges that need to be solved. The 2013 Roadmap has lots of details.
I'd be very skeptical that the advances we have seen in computing in the last 20 years are going to continue at the same pace for the next 20 years without some breakthroughs. But at some point, things will be "good enough". E.g. I can't see the need for 10,000 dpi video displays. And our brains due pretty well with a "clock speed" measured in Hz rather than GHz. :-) Parallelism can do a lot. More advanced sensors for interacting with computers would be great. But, as we know, the software to take full advantage of parallelism is going to lag...
Plus, there's the assumption that everything will keep getting cheaper the way it did in the past. In the past, Intel had competition from NEC, Zlog, AMD, and lots of other CPU vendors. Now, not so much - AMD is still struggling. Intel has licensed ARM (as has AMD and Apple), but their continuing lead in semiconductor processing technology gives them a big advantage apart from the instruction-set. If it gets to the point that Apple and Intel are the only companies that can make a profit on semiconductors, then don't expect the prices to drop any longer...
(As an aside - I'm continually astounded by all the effort put into voice recognition software and wonder if the developers have ever tried using them in a busy airport terminal. But I guess they haven't come up with anything better for portable phones and automated call centers yet...)
My $0.02.
Thanks.
Cheers,
Scott.