Do We Really Need To Answer The 64-Bit Question?

Is a 64-bit processor really a worthwhile concept? – Brian Bedingfield

It seems that there is a lot of buzz going around your website about the new
64-bit offerings from both Intel and AMD, and this is expected. Many people
like to discuss these new architectures, but they never think about the
possible benefits.

Is a 64-bit processor really a wothwhile concept?

First, we must ask, what are the benefits to a 64-bit architecture?

That is simple, it provides more addressing space (namely virtual).

What does this benefit?

Operations that manipulate a lot of data, namely multi-dimensional vector-spaces. Even a
multi-terabyte database seems insignificant from a 64-bit perspective. Not even the most avid
MP3 collector has a database that big.

Manipulation of complex data structures on a 32-bit chip requires many
cycles, while a 64-bit chip may be able to handle it in one cycle. However, while
the performance of complicated operations greatly improves, 64-bit actually degrades the
performance of most common operations. In a sentence, it’s sending a man to do a boy’s job.

It is ridiculous to expect a 64-bit architecture to scale as easily as a 32-bit
architecture. To cite just one example, while a pipelined adder can easily run in a clock cycle regardless
of word length, multiply operations grow on a logarithmic scale as the number of bits grow.
This requires additional stages in a pipeline to simply handle the basic
operation. Therefore, a pipeline must grow to a far greater length to
optimize common operations as the word length increases.

Therefore, we cannot expect a 64-bit processor to scale as well as our
current processor technologies and handle the most common system operations
as efficiently.

64-bit processors seem to be primarily oriented towards
research. I work with people like that everyday who look at me and say “if only we had a 64-bit processor.” However,
the calculations these people do is far beyond the scope of mere consumer applications.

For our typical purposes, a 32-bit processor is more than adequate (for
now and many years to come). Until consumer applications start to implement
complex user-oriented intelligent features, the concept of using a 64-bit
processor for solitaire (or games in general) is ridiculous. So, currently, 64-bit is
purely hype when it comes to a consumer market, and will remain so until consumer
software gets far more complicated than it is today.

I hope this does not sound like a rant, but I think that people should know
what we can expect from 64-bit.

This is just an introduction to the topic. If there is enough interest,
I have many details to back this up and can go more into detail in the future.

For now, just don’t assume that 64 is more than 32 and therefore must be better.

Brian Bedingfield