KOSH [Kommunity Orientated Software Hardware]

Weekly Summary

Week:  17th May, 1999

Number:  014

Mailing List:  kosh-hardware-o

In the mailing list this week, the following items were discussed.
Please do not email the scribe regarding any of these topics, it is
not his job to answer these questions, but merely to report the topics
of conversation. If you have any queries about this summary, please
email summaries@kosh.convergence.org, stating the Summary Number, and
Mailing List Name and he will try to answer your queries.

a)

Subject: Amd K6-2 versus Intel Celeron

Summary of Debate:  The K6-2's FPU can actually process a single
                    instruction faster than the celeron, but the
                    celeron's FPU is pipelined, so it's FPU is
                    effectively much faster.  The K6-2's 3DNow!
                    instruction set in some ways makes up for its
                    fpu deficiencies.   The K6-2 has more
                    cache than the celeron, but the celeron's
                    cache runs faster.  There is a special
                    configuration under which celerons can be
                    run dual.

b)

Subject:  WinChips

Summary of Debate:  These are x86 chips, primarily intended for
                    laptops where their lower power usuage is an
                    advantage.  They run at lower clock speeds
                    than the current Intel chips, are somewhat
                    similar to the Cyrix chips.  Rather than
                    trying to keep up with Intel on the clock
                    speed front, their maker, IDT, focused
                    on improving their integer support, at the
                    expense of their fpu.  The latest versions
                    have 3DNow! support.  In Germany they are
                    called c6, w2, and w2a because winchip is
                    copywritten in Germany.

c)

Subject:  Single Unit Measurements have limited use

Summary of Debate:  Measuring the performance of a complicated
                    system with a single unit of measurment
                    isn't always satisfactory.  One example
                    would be the demonstration of 1Ghz CPU
                    x86 systems, which in practice are
                    tremendously bottlenecked by RAM speeds.
                    Another example would be measuring hard
                    drive's strickly by RPMs.  In some hard
                    drive uses, streaming data or large files
                    for example, seek time can be equally
                    important.