🚀 go-pugleaf

In article <533@myrias.UUCP>, bhm@myrias.UUCP (Brian Moore) writes:
> Has anyone out there done a table of comparison between the various
> micro chips available and their cost/chip, cost/mip, cost/mflop?

> How does the T800 compare with the M68030 or MIPS chips.

Well, Inmos mentioned Whetstones in the Electronics article, announcing it
as having the fastest floating-point unit. I don't have the figures in
front of me for the T800, but the MIPS R2000/R2010 gets 12 MegaWhetstones
single precision, and 9.3 MegaWhetstones double precision when run at 16.67
MHz, which was about double the figure for T800.

We'd rate a 16.67 MHz system with full size (64kbyte each) caches and
reasonably fast, local, memory at 12 MIPS (or 12 times faster than a
VAX-11/780). It appears the the T800 performance Whetstone figure has code
and data running out of the internal memory, and that the same figure with
off-chip memory and/or caches will be slower. - anyone have details?

Cost? MIPS quotes 5000 quantity prices of $195 for the R2000 CPU and
$295 for R2010 FPC. The price of the CPU is less than the 5000 quantity
price for a '386.

> This should lead to lots of debate on what is a mip wrt Occam ( risc??)
> and more traditional languages.

Not from me. I'd bet Occam disappears as fast as you can get C compilers
for the Transputer. Incidentally, does anyone know whether the the
T800 Whetstone figure was from an honest-to-golly Fortran compiler,
or did they "hand-code" the benchmark?

-- 
Craig Hansen
Manager, Architecture Development
MIPS Computer Systems, Inc.
...decwrl!mips!hansen