I have been following this thread for a long time now, and at present
I still have an ARM610.

I may be getting a StrongARM in the not too distant future, so one
natural question is what the revision of the 233 MHz StrongARMs
Acorn are selling now is. Does anyone know?

If these are in the clear, then there should only be around
10k bad StrongARMs out there now, and much more good ones.
IIRC Acorn made an initial batch of 10k units to get the
price down in order to make sure that most people would
take this opportunity to get a faster machine.

When this batch ran out, the 233MHz chip was available from
Digital Semiconductor, so they started using this chip. From
the errata sheet I see that the good chips were available
around Acorn World '96, so the 233MHz ones should be in the
clear.

Another thing I have been wondering about is that the problem
boils down to paging. Wouldn't the best workaround be to get
enough physical memory so as to avoid paging?

Or, is there another sort of paging in NetBSD that I don't know
about?

If I were in Kim's shoes and had the same problem, I would
definitely go out and get myself a new StrongARM chip, or
enough memory not to swap. The Risc PC can take up to 256MB,
and I think revision 3 motherboards can take up to 512 MB.

Of course the latter would only be a solution if there is
no other paging going on than when the physical memory
runs out.

I will hereby volunteer to collect emails from people
who are interested in replacing their chip with a new one.
I will not take any action on this, but I can put the
emails I get as a mailbox on my W3 pages or mail it to
people who are interested.

The email address to use is mailto:strongarm@thomassen.priv.no.
I will filter that to a separate mailbox.

At present I have 64 MB DRAM in my Risc PC 600 (revision 1),
so I should be able to do most without paging.

Another good thing about the revelation of this bug is that
I know that I can upgrade to 1.3.1 without having fears of
any problems.

The reason why the 1.3_ALPHA cc1 and cc1plus works and the 1.3
ones does not seems to be be the result of this bug. At
least that is the conclusion I have made from this thread.

I really feal sorry for those who have a buggy StrongARM,
but if I were you, I would go and pester Intel about it.
(FYI: Intel bought the StrongARM part of Digital.)

I seem to remember that they replaced the bad Pentiums that
had a math bug free of charge, but I may remember incorrectly.

That's all for now. Sorry to have wasted your time.

Kjetil B.
mailto:kjetil@thomassen.priv.no
http://home.eunet.no/~kjetilbt/