Moore's law only states that the number of transistors doubles for a specific size of silicon every 1.5 years.<p>When it does end, we will move onto other forms of computing, memristor based, graphene, photonic/optical, quantum (only really useful for some classes of problems), spintronics and such. Might have a bit of a dip off during the change over but then it will be business as usual.<p>We could start to see computers with more and more processors in them. The cost of producing a processor should continue to drop. Energy efficiency, (reduced heat) and cooling improved systems will keep being developed. So maybe we will end up with computers that are almost a solid mass of processors. That scalability isn't fantastic though, but it would work for maybe 10 years. Things like FPGA coprocessors could also help.<p>Finally we can look at breaking away from the current paradigms of computer science into esoteric things like biological/dna computing.<p>Most desktop systems are as fast as most people will ever need them. Remember now days all they are doing is opening a webbrowser and going to Facebook. The UI's keep getting more and more simplified.<p>Graphics in games is about the most demanding thing that happens on a desktop, which will likely keep going for some time until we have photo-realistic, realtime raytracing at retina DPIs on large screens, in 3D (maybe with voxels for some kind of future volumetric display).
As a student of analytic philosophy, this has always caused much confusion when talking to students of logic without a philosophical background. Gordon Earle Moore, of Moore's law, is incredibly frequently confused with George Edward Moore, of the "here is one hand" proof of an external world.<p>Part of me will be happy when G.E. Moore (George) get's a little more attention.
Of course Moore's Law will end eventually, but that doesn't mean that improvements in processor technology won't come from somewhere other than transistor density. Most likely, we'll see 2x+ improvements from one or more of graphene transistors, 3D chips, optical circuits, quantum computing...
The five performance bottlenecks: algorithms, programming language or compiler, operating system, processor, I/O system and devices.<p>One almost down I'm sure the I/O system will reach a wall soon as well. Only thing left is to optimize the software.
While Moore's Law clearly cannot go on forever, everyone in this trade should remember its end was prophesized many times before, every time backed up by good science.<p>Time will tell.