On Tuesday 04 March 2003 08:26 am, Anne Wilson wrote:
> On Tuesday 04 Mar 2003 3:05 pm, et wrote:
> > On Tuesday 04 March 2003 09:29 am, Anne Wilson wrote:
> > > I have tried without success to install onto a very old box, IBM/Cyrix
> > > cpu. It really isn't worth fighting too much, but I'd like to have one
> > > more go at it.
> > >
> > > All I have got, so far, is as far as the initial screen, F1, and first
> > > try, typed 'linux nopentium noapic'.  This brought just a blank screen.
> > >
> > > I then tried to run one of the alternative kernels(?), but putting alt1
> > > into the same statement, but this was obviously wrong, because just
> > > before the blank screen I saw that it was loading the alt0 kernel.
> > >
> > > a)  What is the correct syntax for trying the alternatives?
> > > b)  Is there anything else worth trying - before I kick the box out?
> > > <g>
> > >
> > > Anne
> >
> > also if it has an onboard video card, giving the "mem=xxxM" statements
> > where xxx= the amount of installed memeory minus the amount "shared" with
> > the video. as well as remembering to turn off Plug and play aware OS?
>
> The on-board video is turned off - I have always preferred separate sound
> cards, though I believe on-board audio is improving.
>
> I'm beginning to suspect that Mark is right, and it simply doesn't support
> the cpu.
>
> Anne

Actually the 6x86 is fully i586 compatible....  But remember....

The Pentium had the f00f bug 

The AMD K6 had a bug which had no linux kernel workaround

The Cyrix had the coma bug  (the two execution units could be bus-locked in a 
loop each waiting for the other to complete, and both masking ALL interrupts)
Note that his is a hardware/microprogramming/architecture flaw, not a kernel 
limitation.  When certain code hits the processor at the right time and gets 
split between the two execution units in the wrong way, a logic interlock 
occurs.

Most likely the newest kernel is not set up to avoid the coma bug--it is 
subtle and cannot be avoided in compiled languages, only by inspecting the 
assembly or machine output could we see if it is invoked, and that is a LOT 
of machine code to examine.

I have used Cyrix processors extensively.  They were all 6x86 or M series and 
all performed well on most tasks though they had a tendency to run hot.

The non-i686 compliant Chips one might be thinking of is the continutaion of 
the IDT Winchip series -- the VIA C3 which is actually emulated by the 
Transmeta.  There We DO have a problem with the kernel not properly 
identifying the processor (i686 when it is only i586) but only the 9.0 kernel 
and there is a workaround.

Yes an earlier kernel (alt1,alt2) is certainly worth trying.

You do not need "nopentium" which is a dodge for the K6/K7 to avoid a memory 
paging error.

Civileme


Want to buy your Pack or Services from MandrakeSoft? 
Go to http://www.mandrakestore.com

Reply via email to