Sunday, July 16, 2006
A slight shift in the PC Paradigm is about due
Over the last twenty years or so we've had incredible advances in technology ... particularly in digital consumer electronics ... Personal Computers ... digital cameras, printers, cell phones and GPS ... Etcetera. Oh! And HDTV and digital TV broadcasting!
Now we have dual core processors. With high clock speeds combined with impressive bus and memory data transfer rates ... We are about due for a shift in the PC paradigm.
PCs, and their Operating Systems, have been able to utilized multiple processors for sometime.
Essentially the CPU still handles nearly all user interfaces and program execution.
The result, between the CPU and the OS, was that simple user functions ... such as moving the mouse around vigorously... resulted in a marked negative performance impact.
On early Windows based systems you could watch CPU utilization increase dramatically just by moving the mouse around the desktop. With faster processors and corresponding increases in bus performance this is not as noticeable as it once was just six years ago or so. But the inherent architectural problem remains ... the CPU processes almost all of the user interface (GUI) and program execution.
At some point, hopefully very soon, we should anticipate that the OS will begin to utilize these multiple processors to divide the workload between the GUI and program execution.
The result should be that the GUI should not be impacted by program execution and overall system management. Inversely, neither should the processing of "work" be negatively impacted by the user interface.
Changing the user interface such that the GUI processes are dedicated to a single processor and the actual performance of program execution be allocated to other individual processors (CPUs) should increase performance and stability.
The time for the end of a "hung" system and/or GUI should be considerably closer to being a reality.
It seems to me that additional hardware changes would be needed to better achieve this goal ... more buses, shared and dedicated memory would probably make the design and system management easier.
Hopefully a much smarter IO sub-system would finally become a reality ... quick example ... defragmention of the HDD shouldn't be a normal end-user activity ... if the CPU needs data for a program, fine, a data request is made and the main CPU doesn't concern itself with where the data lives or what format it is in ... it should work similar to how graphics cards work today ... it is predominantly an externally managed process.
I would think, hope, that Linux/Unix operating systems would have an easier time achieving this than Windows ...
Regardless, I would anticipate changes of this nature within the near future ... say 3 years ... maybe a little longer ... this assumes that it is even on the radar of the system architects.
Now we have dual core processors. With high clock speeds combined with impressive bus and memory data transfer rates ... We are about due for a shift in the PC paradigm.
PCs, and their Operating Systems, have been able to utilized multiple processors for sometime.
Essentially the CPU still handles nearly all user interfaces and program execution.
The result, between the CPU and the OS, was that simple user functions ... such as moving the mouse around vigorously... resulted in a marked negative performance impact.
On early Windows based systems you could watch CPU utilization increase dramatically just by moving the mouse around the desktop. With faster processors and corresponding increases in bus performance this is not as noticeable as it once was just six years ago or so. But the inherent architectural problem remains ... the CPU processes almost all of the user interface (GUI) and program execution.
At some point, hopefully very soon, we should anticipate that the OS will begin to utilize these multiple processors to divide the workload between the GUI and program execution.
The result should be that the GUI should not be impacted by program execution and overall system management. Inversely, neither should the processing of "work" be negatively impacted by the user interface.
Changing the user interface such that the GUI processes are dedicated to a single processor and the actual performance of program execution be allocated to other individual processors (CPUs) should increase performance and stability.
The time for the end of a "hung" system and/or GUI should be considerably closer to being a reality.
It seems to me that additional hardware changes would be needed to better achieve this goal ... more buses, shared and dedicated memory would probably make the design and system management easier.
Hopefully a much smarter IO sub-system would finally become a reality ... quick example ... defragmention of the HDD shouldn't be a normal end-user activity ... if the CPU needs data for a program, fine, a data request is made and the main CPU doesn't concern itself with where the data lives or what format it is in ... it should work similar to how graphics cards work today ... it is predominantly an externally managed process.
I would think, hope, that Linux/Unix operating systems would have an easier time achieving this than Windows ...
Regardless, I would anticipate changes of this nature within the near future ... say 3 years ... maybe a little longer ... this assumes that it is even on the radar of the system architects.
David of One, 7/16/2006 03:22:00 PM
|
