Because I've received the `arcdiag' result from Ross and Adam in the past.
(Adam, sorry for not responding your question....)
I know some information about their machines.

> 3.If a step 2 above you now suspect possible wrong clock code
> then to 'fix-it' you need to change it in sys/arch/arc/arc/clock_mc.c..
> However the gotch is YOU NEVER SHOULD HAVE GOTTEN
> THIS FAR!!.  The reason I claim this is the way the 'arc' port
> configures for different models is a function in 'arcbios.c'
> reads the 'machine type' from the arc firmware and sets
> a global variable 'cputype' to one of the values defined
> in 'arctype.h'. The currently defined types in
> Netbsd-current are :
	:
> If you closely you see Risc Express server in NOT LISTED!
> What's happening is the cpu-type is being reported as
> some other type already listed above BUT the Express
> varies somewhat from the type.

According to arcdiag output from Ross, RISCserver Express looks 
very closer to RISCserver 2200.
Since it's system type ID is "NEC-R96" and vendor strings is "MIPS DUO"
(In other words, RISCserver Express is exactly same with RISCserver 2200
 in this point), RISCserver Express should match NEC_R96.
Only difference between RISCserver 2200 and RISCserver Express I saw
is L2 cache size. RISCserver 2200 has 2MB cache, but RISCserver Express
only has 1MB cache. So, perhaps chaging line 769 if arc/arc/machdep.c
from "#if 0" to "#if 1" may makes difference.

Adam's RISCstation 2200 is different.
It's system ID is "NEC-R94", and vendor string is "NEC W&S",
Adam's machine should match NEC_R94.
About NEC_R94, probably clock_mc.c should treat it as just like
NEC_R96....
--
soda