Re: MISC-d Digest V98 #44

[jfox@xxxxxxxx (Jeff Fox)]

Dear MISC readers:

>From: kragen@pobox.com (Kragen)
>To: misc
>Subject: hi - status?
>
>I'm back on the list after a year's absence.
>
>As far as I can tell, the i21 is working well enough to demo but not
>well enough to ship, 

The difference between demo and shipping is the person seeing the
demo wanting to spend money on product so product has ship to the
mass market.  iTV has been shipping small quantities of product
to corporte partners.  But have not made any announcements about
shipping in quantity for retail distribution.  this is really a
technical issue.

>the F21d is in the fab, and will be 500MIPS.  The

No.  I have never said F21 is a 500mip chip.  It has a 500mip cpu
going through a memory interface unit that shared by all I/O coprocessors.
The original spec published years ago said 200mips max throughput.
ITV literater calls i21 a 500mip chip.  My literatre has said
max is 333mips interal ROM, 222mips SRAM, 111mips DRAM, 6mips ROM
for F21.

>F21 software runs snappily at 5 MIPS on Jeff's desktop computer in an
>F21 simulator.  

the F21 simulator is now free.  It runs about 10K F21 instructions
per second.  The F21 emulator run NOPs at 4 mips on my laptop and
twice that on Stas' machine.  Running the video animation or BMP
file decode demos you can see that there is lots of slow data
access.  Both the simulated and real execution rates drop.  On
those programs the simulator only gets about 4k and emulator about
900K F21 instructions per second.  However after using simulators
with a max of 10k for a number of years something 100 times faster
is really nice.

When I was simlating F21 on a 1Mip laptop things were slow.  I
didn't simulate video and it took several minutes for a Forth
running no the simulated chip to interpret a string of text.
It is a please now to have a computer that is 100 times faster
and an emulator that is 60 times faster than the simulators.

>When the F21d comes back from the fab, it will probably
>work.  iTVCorp is offering microprocessor cores for licensing;
>presumably this means the i21.  They don't list any licensees.  The
>microprocessor core is currently 4 square millimeters in an 0.8 micron
>process.
>
>Is any or all of this correct?

The last part is correct.  The die is about 4 sq mm because of the pad
configuration.  With a different pad configuration it could be about
1 sq mm.  It just requires a different kind of bonding pattern for pads.

>What, exactly, is the relationship between the F21 and the i21?

F21 was started in 1990.  F21 is owned by Ultra Technology.  Ultra
Technology plans on selling F21.  Ultra Technology has provided full
technical information.

i21 was started in 1995.  i21 is own by the iTV Corporation.  The iTV
Corpration plans to use i21 in proprietary products.  They release
information about the products, not about the trade secret details
of the chips inside.  They have never intended to sell chips.

I cannot list all the technical details side by side to detail
"exactly" what is the relation in that sense.

I own Ultra Technology and I am an employee at iTV.  Chuck has
the patents and owns the designs.  The other companies own the
chips that he sold to them.  I own a 4 year marketing contract
on F21.

Jeff Fox
www.ultratechnology.com