KOSH [Kommunity Orientated Software Hardware]

Weekly Summary

Week:  4th January 1999

Number: 004

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:  3D Laser Projection Display Systems

Summary of Debate:  As a possible method of terminating the laser 
		    beams at the edge of the image being projected,
		    quick vaccum pumps could be employed.

b)

Subject:  AMD Chips

Summary of Debate:  The K6-2 and K6-3 are nice chips, their weakness 
		    being that their FPU's are not fully pipelined.
		    The K7 will use RamBus and the EV6 bus which is
		    currently used by the Dec Alpha, and have a
		    backside memory bus of up to 8MB L2 cache. 

c)

Subject:  New Instruction Sets

Summary of Debate:  3D Now allows SIMD processing of instructions, is
		    supported in some programming packages and by
		    Active X.  AltiVec goes further in using a 
		    seperate set of 32 128bit registers.  It is
		    speculated that Intel's Katmai will use similar
		    instructions to AltiVec.  
		    Possible ways for KOSH to take advantage of all
		    of these include "plugin" objects for different
		    processors, standard function libraries that 
		    could be reimplemented for each set, or slim
		    binaries.

d)

Subject:  Expandable versus self contained systems.

Summary of Debate:  Some people would like a "console" system with 
		    fixed hardware, while others like the 
		    expandability of PC style computers.  If KOSH is
		    properly scalable it could run on both.

e)

Subject:  Custom hardware versus "off the shelf" hardware.

Summary of Debate:  Current PC hardware has some "legacy" features,
		    left over for backwards compatability.  None of
		    these features, however, effect the critical path
		    (CPU, bus, memory)  Some feel that custom 
		    hardware may be used advantageously in an untapped
		    niche or as a replacement for PC architechture 
		    when that arch hits its limits.  At this point the
		    majority seem to feel that producing custom 
		    hardware is impractical.  One because the hardware
		    market is very competetive right now (cheaper
		    and cheaper graphics and sound cards which do
		    more and more.)  Two because with recent additions
		    the current PC architecture is fairly powerfull.
		    And three, because the initial investment would be
		    immense.

f)

Subject:  Hardware performs better without Windows

Summary of Debate:  Two examples were the QNX demo disk and the fact
		    that some companies choose to use a 486 and linux
		    for a mailserver rather than NT on a PII.

g)

Subject:  Standards Wars

Summary of Debate:  Intel has made the CPU interface field more 
		    complicated by having the Celeron A use socket
		    370.  Creative and Diamond are competing for
		    a 3D sound standard.

h)

Subject:  Multibus architecture

Summary of Debate:  Future busses may use the mainframe concept of a 
		    bus in which all subsytems can simultaneously 
		    access main memory.  This may be in progress at
		    SGI, who would have access to the tech, having
		    bought Cray.

i)

Subject:  Drivers

Summary of Debate:  Windows drivers do a lot of abstraction.  Perhaps
		    we should focuss on finding a way to harness them
		    and then begin replacing them with kosh native 
		    drivers.