I have just finished a tour with Intel in Europe. The message around processor performance is becoming very clear.
In the past software developers and to some degree operating systems vendors simply relied on the ever increasing clock cycles of the processors from Intel and AMD to boost performance. However this approach is clearly not going to work moving forward, with pervious statements from Intel saying that if we have to keep up with Moore’s Law on a single processor, the processor core will be hotter than a Nuclear Reactor in three years.
The future for Intel around processor scalability ( much the same as AMD ) is that they have to move to multi-core. New laptops are coming out with dual-core architectures already and servers will be multi-core enabled by the end of the year.
What does this all mean? It sounds fast, but ultimately now to get the performance from these multi-core machines, developers are going to have to write well architected applications which support multiple threads to take advantage of the performance of the new multi-processor architectures ( even if they are on the same silicon ).
I am not sure whether anyone is aware of this yet, but bottom line, the poorly architected, single threaded applications are going to suck in the new multi-core world and engineers are going to have to become a little more aware of the applications they are building if they need to get ultimate performance. Sound like a salary increase to me. Moore’s Law continues …..