Re: Some tests with my Macintosh Centris 650
On Fri, 23 Jan 2015, John Paul Adrian Glaubitz wrote:
>
> > Failing that, can you send me a photo of the crash? The serial console
> > log would be a lot better than a photo but you'd need a suitable
> > cable. (I use an Apple StyleWriter cable with a Keyspan
> > USB-to-mini-DIN8 adapter.)
>
> Here you go: http://imgur.com/a/luLEV The last screen is actually where
> it stops. It looks the same when booting 3.14 but this kernel continues
> to boot fine after the pink screen.
That's a very early crash. It could be caused by a NuBus slot interrupt
being asserted, since these are non-maskable IRQs on some models. This can
happen with SONIC ethernet, prior to the Linux macsonic driver being
loaded.
Try booting MacOS with no extensions loaded (hold down shift key when you
hear the boot chime, and release it when you see the words "Extensions
disabled").
If this really is a regression, since 3.14, an obvious suspect would be my
earlyprintk patch, commit 7913ad1ad834, which appeared in 3.16. But this
crash looks to be too early.
>
> > You might want to order a 16.666 MHz crystal as well, in case of
> > instability at 40Mhz.
>
> Good idea!
I've been there myself. I once clock chipped a PowerBook 520 CPU board. I
desoldered the SMT 68LC040 CPU and replaced it with a 68040, and replaced
the crystal at the same time. It worked, but the extra transistors and the
extra frequency generated far too much heat for the passively cooled
laptop to deal with. It needed a high-efficiency cooling system from a
modern laptop, but there's no room for that.
>
> > The Quadra 650 (33 MHz) has a big heat sink on the CPU; I don't recall
> > seeing one on the slower machines.
>
> The Centris 650 has a big heat sink as well which is why the guys from
> the clockchipping website claimed overclocking them wouldn't be an issue
> at all...
Fair enough. I wonder if you need heat sinks on other chips -- who knows
what clock lines could be affected by that crystal.
> > For reliability, I think it would be prudent to do the whole
> > conversion (resistors and crystal, soldered) just so that MacOS knows
> > what clock is fitted, because it probably needs to know which timing
> > loops to use. It seems likely that drivers besides serial will
> > benefit. Did you find any reports of working Ethernet with a partial
> > conversion?
>
> I could do that later on. First I want to see what results I get.
Sounds like a good plan.
--
Reply to: