where will the action be ?



one old concept in scientific research (and other areas of endevour) is to go and work where the "action" is. in the "hot new fields".

a year ago this was C# and .NET.

it isn't any more - there have been a few developments outside of Borland's control that have changed this:

1. AMD64 has succeeded - technically and commercially - to the point where Intel's IA64 architecture is a dead duck. it will eventually replace IA32.

2. AMD64's multicore technology has succeeded technically, and shows every sign of doing so commercially by the end of the year.

3. Intel's NetBurst/P4 architecture has failed: it will not bring higher and ever higher clock speeds, so programmers are facing a speed issue.

4. so future CPU development will be towards greater parallelism (multicore), and clock speeds will stall around the 3 GHz mark. they may even decrease with decreasing feature size. Intel's future cores will be based on descendants of the P3 -> PM, and they will be lower in clock speed, but multi-core.

5. .NET, as elegant and beautiful as it is, is intrinsically slow (which may or may not affect a particular app), and there will not be any faster CPUs to fix its speed issues. even worse, it does not do parallelism very well, so it cannot take advantage of all those extra cores.

6. Microsoft is not a monolith (though it's nice to think of them as such). it is composed of a number of feuding kingdoms vying for the emperor's favour. a year ago the .NET kingdom had the emperor's ear; today the 64-bit kingdom is in ascendency.


There has been a paradigm shift. What was a wise business decision a couple of years ago (.NET) is now looking more and more like the OS/2 of our times. I predict that .NET will end up being a technology for web services for Microsoft's web server, and little else.



Microsoft has now indicated where it's major effort will be in the near future. I hope Borland takes note of this change of direction. I really don't want to have to migrate to Microsoft C++/64 products.




cheers,



Mat
.



Relevant Pages

  • Re: where will the action be ?
    ... Intel's NetBurst/P4 architecture has failed: it will not bring higher and ever higher clock speeds, so programmers are facing a speed issue. ... It takes 20MBytes of runtime memory to run, compared to most similar WIN32 packages that take 5MBytes or less. ... a year ago the .NET kingdom had the emperor's ear; today the 64-bit kingdom is in ascendency. ... I predict that .NET will end up being a technology for web services for Microsoft's web server, ...
    (borland.public.delphi.non-technical)
  • Re: where will the action be ?
    ... Microsoft is not a monolith (though it's nice to think of them as ... > is composed of a number of feuding kingdoms vying for the emperor's ... > year ago the .NET kingdom had the emperor's ear; ... > predict that .NET will end up being a technology for web services for ...
    (borland.public.delphi.non-technical)