1 of 2 pages
Competing with Multiple Legs in the Semiconductor Race
My opinion on why semiconductor companies will face many problems while trying to achieve performance gains using the multicore processor approach.
0 Comments on this document
Document Transcript:
COMPETING WITH MULTIPLE LEGS IN THE SEMICONDUCTOR RACE
On March 18, I found an interesting article titled "Racing to Gain Edge on Multicore
Chips" on the Wall Street Journal. The article talks about one of the challenges that Microsoft
and Intel are facing when trying increase computers' processing power.
During the last 50 years, semiconductor companies have been racing to produce the
fastest chip by increasing the number of processed instructions per second. Such approach,
however, is not so useful anymore due to heating problems; the faster the processor, the more it
heats, and manufacturers are already pushing the thermal limits of the processor's materials.
To cope with the heating problem, manufacturers are starting to create computers with
multiple core processors that are used in parallel. This approach allows computers to shift the
processing load from a single processor to multiple processors, effectively increasing their speed.
However, there is a catch: special software is required to use effectively multiple processors.
Enabling software to use multiple processors is a complex task. Consequently, Intel and
Microsoft have decided to enlist the help of the University of California at Berkeley to create
programming techniques that will simplify the software development. For this solution to work,
programmers must be able to write easily code that takes advantage of the multiple processors.
Intel has always marketed "faster" processors to remain competitive, but the next multi-
processor battle is not going to be an easy win. Software developers are used to writing software
for single processors. Except for specialized applications, the bulk of the software runs in single
processors even if multiple processors are available. The technical complexity necessary to
switch to multiprocessors is likely going to keep programmers away. The truth is that most ofthe people use computers for editing documents, downloading content off the internet, or
communicating, none of which would benefit considerably by using multiple core processors,
therefore, the extra effort will be difficult to justify.
Convincing programmers and development companies that they should switch to the
multiprocessor approach is going to be a daunting task. Although programmers have used
similar multiprocessor approaches for high-end super computers, this new approach means
changing the programming methodology that has been used for decades for most of the personal
computer industry. The University of California at Berkeley, Microsoft, and Intel will have to
come up with an easy-to-use solution if they want such solution to work. Only time will tell if
this is the correct approach.
On March 18, I found an interesting article titled "Racing to Gain Edge on Multicore
Chips" on the Wall Street Journal. The article talks about one of the challenges that Microsoft
and Intel are facing when trying increase computers' processing power.
During the last 50 years, semiconductor companies have been racing to produce the
fastest chip by increasing the number of processed instructions per second. Such approach,
however, is not so useful anymore due to heating problems; the faster the processor, the more it
heats, and manufacturers are already pushing the thermal limits of the processor's materials.
To cope with the heating problem, manufacturers are starting to create computers with
multiple core processors that are used in parallel. This approach allows computers to shift the
processing load from a single processor to multiple processors, effectively increasing their speed.
However, there is a catch: special software is required to use effectively multiple processors.
Enabling software to use multiple processors is a complex task. Consequently, Intel and
Microsoft have decided to enlist the help of the University of California at Berkeley to create
programming techniques that will simplify the software development. For this solution to work,
programmers must be able to write easily code that takes advantage of the multiple processors.
Intel has always marketed "faster" processors to remain competitive, but the next multi-
processor battle is not going to be an easy win. Software developers are used to writing software
for single processors. Except for specialized applications, the bulk of the software runs in single
processors even if multiple processors are available. The technical complexity necessary to
switch to multiprocessors is likely going to keep programmers away. The truth is that most ofthe people use computers for editing documents, downloading content off the internet, or
communicating, none of which would benefit considerably by using multiple core processors,
therefore, the extra effort will be difficult to justify.
Convincing programmers and development companies that they should switch to the
multiprocessor approach is going to be a daunting task. Although programmers have used
similar multiprocessor approaches for high-end super computers, this new approach means
changing the programming methodology that has been used for decades for most of the personal
computer industry. The University of California at Berkeley, Microsoft, and Intel will have to
come up with an easy-to-use solution if they want such solution to work. Only time will tell if
this is the correct approach.
