Re: OpenBSD on 486 - Problem?

[Nick Holland <nhollan@home.com>]

Comments imbedded within...

Thomas Ribbrock wrote:
> 
> Hi all!
> 
> I have OpenBSD 2.6 up and running on an old 486. The other day, I upgraded
> the CPU from a 486DX2/66 (Intel Overdrive) to a SGS-Thompson ST486DX4-100.
> The BIOS recognises this CPU as a Cyrix CX486DX-S @ 66MHz . Supposedly, the
> ST468DX4 is the same as a Cyrix Cx486DX4-QP (I think - there aren't too many
> resources on these on the web).
> OpenBSD, however, recognises it as a Cyrix 486DLC and writes out a warning:
> 
> OpenBSD 2.6 (GENERIC) #696: Tue Nov  2 01:46:05 MST 1999
>     deraadt@i386.openbsd.org:/usr/src/sys/arch/i386/compile/GENERIC
> cpu0: Cyrix 486DLC (486-class)
> WARNING: CYRIX 486DLC CACHE UNCHANGED.
> [...]
> 
> I didn't find anything in the list archive or FAQ to that regard, but I did
> find something about the warning in the NetBSD FAQ:
> "This message indicates the kernel has detected a Cyrix 486DLC and since it
>  does not know if it is safe on this machine, the Cyrix special cache
>  settings have not been touched. This should only affect the Cyrix '386/486
>  hybrid' upgrade for 386 motherboards, and should not apply in any way to
>  true 486 (or later) processors from any vendor, including Cyrix."
> It then goes on to describe kernel options (CYRIX_CACHE_WORKS and
> CYRIX_CACHE_REALLY_WORKS) which one can try to set to improve performance on
> these CPUs.
> 
> Hence my questions:
> 
> 1) Am I right in assuming that the same explanation holds true for OpenBSD?

Got me, but it would be a good guess.  
Keep in mind, most people doing development get frustrated on a 486
system, so it is likely that no one has touched the Cyrix code in a
long time -- esp. the 386/486 code...

> 2) The system seems to be running fine with the ST486 - it even seems
>    slightly faster than before (subjectively - I haven't done any actual
>    measruements). Should I worry about that warning at all? Do I need to
>    worry about the mismatch in recognised CPU type vs. real CPU type? I'm
>    just a bit surprised, as the 486DLC was some kind of 386/486 hybrid,
>    whereas the ST486 is a "true" 486...

No.  Certainly do not worry.
This should probably be considered much more an "advisory" (i.e.,
information to you) than a true warning (i.e., things may cause
problems).  It is advising you that there is a feature this chip may
support that is not being utilized.  That's all.  My understanding is
that some of these non-Intel chips offer some improvements over the
Intel 486 internal cache, but this can conflict with the external
cache provided on many main boards.  

486 identification from software was an interesting art, as I
understand it.  The program made a few tests, and took a guess. 
Pentium-class systems have some kind of identifying string which
simplifies things substantially.  Considering there were many
different companies using the characters '486' for everything from
things that were mostly 286s (o.k., 386SXs, but I always considered
the 386SX to be more like the 286 than 386) to almost Pentium systems
(and if you count some of the Cyrix 6x86 chips which reported back to
some software as a 486, things get even more messy), it is easier for
me to forgive Intel's attempts to copyright the number 486.

As for performance increase, I tell my clients they should be looking
for a 2x performance upgrade before they consider spending money and
effort on an upgrade.  Anything less than that is unlikely to be
noticed by a person sitting at a keyboard and using the system, and if
the user can't tell, it doesn't matter.  This isn't autoracing, where
a couple inches can determine victory.  I wouldn't think you will get
any noticable improvement in performance if you were to tweek the
Cyrix cache, and if that mattered, I'd really recommend moving to a
newer Pentium class system. 8)

Nick.

-- 
http://www.holland-consulting.net/