It is a more complex engineering issue than that. Video
accelerators seem to be a good idea for now because of the enormous
bandwidth used, which is increasing faster than bus bandwidth due to
increasing resolution, pixel depths, and richer user interfaces that
involve modifying a larger proporion of the screens pixels. However,
frame buffers themselves represent an elimination of another
coprocessor, terminals, which were eliminated shortly after the
elimination of another coprocessor, the "front end" computer, and I
think we're likely to see video "coprocessors" and even graphics
memory moved all the way back across and into the CPU die itself in
the future as microprocessor designers look for more things to sell as
much silicon as they can and cash strapped DRAM vendors start putting
logic cores on DRAM chips. It could be argued that the Cyrix MediaGX
chip is a step in this direction. It certainly has dropped a bomb on
the low end of the computer market. There have also been moves toward
some simplication of external processors, for example, the rise of the
IDE bus. In general, I do not see a pervasive trend of
reapportionment of complexity from CPU to coprocessors or vice-versa,
but rather continuing pragmatic adjustment of that balance as
improvements in technology make different tradeoffs optimal.
Winmodems are an example of this. It may be that by saving some money on
the modem and spending it on a faster CPU, you get a more powerful system
at some price points, even when the modem is running full blast.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html