On Tue 30 Jun 98 (10:01:02), chaiken@pa.dec.com wrote:

>The reason for going from the 65550 to the IGS was that the
>performanceof it *sucked* extremely badly---it wasn't that it had any

That's only part of the story:

   http://www.research.digital.com/SRC/iag/info/ctigs.html

We changed the chip because IT DIDN'T WORK at the specs that C&T
originally quoted to us.  In particular, when the power supply is at
5V and you crank up MCLK to get reasonable performance, the chip has a
bad habit of melting.

>>The rev5 boards also have some video problems with X11. They shimmer window
>>boarders like crazy if the window has a white background (which affects say
>>Neil more than me since I like to do the NCOS builds in say ETerm for some
>>really strange reason)
>
>I thought there's some info on Digital's DNARD pages that shows you how
>to fix this problem (requires some rework).

http://www.research.digital.com/SRC/iag/info/DNARDfix/index.html

While I'm posting, I'd like to apologize publicly for the ISA DMA mess.
At least I documented the horrible hacks: see arm32/isa_machdep.c.

Bunging an ISA bus on the side of a StrongARM is the root cause of
some of these hacks.  In defense of this architecture, note that shark
went from a blank sheet of paper to its first trade show in 6 months.
Shark achieved its target: $300 bill of materials with the same
performance as a 166 MHz Pentium (quite good at this time last year).

Believe me, we had plans to clean up both the hardware and the software.
The early demise of the shark project killed the clean-up efforts.
On the other hand, there wouldn't be so many free sharks available
if we hadn't been exploded.

dc

-- 
 __  __  __  __      __ ___   _____________________________________________
|__||__)/ __/  \|\ ||_   |   /
|  ||  \\__/\__/| \||__  |  /...Internet access for all Acorn RISC machines
___________________________/ ajdurham@argonet.co.uk