Re: Apple PowerMac G5
On Wed, Jun 25, 2003 at 01:12:35PM +0200, Sven Luther wrote:
> On Wed, Jun 25, 2003 at 11:03:51AM +0000, paubert wrote:
> > On Wed, Jun 25, 2003 at 10:32:10AM +0200, Sven Luther wrote:
> > > On Wed, Jun 25, 2003 at 09:33:33AM +0100, Bastien Nocera wrote:
> > > > On Wed, 2003-06-25 at 06:26, Sven Luther wrote:
> > > > > On Wed, Jun 25, 2003 at 12:55:04AM +0100, Bastien Nocera wrote:
> > > > > > On Wed, 2003-06-25 at 00:24, Leandro Guimarães Faria Corsetti Dutra
> > > > > > wrote:
> > > > > > > Anyone has it working already with Debian?
> > > > > >
> > > > > > Read the announcement properly, they won't ship until September in the
> > > > > > US...
> > > > >
> > > > > Err, august that is, no ?
> > > >
> > > > Same difference, it's not shipping for quite some time :)
> > >
> > > Notice that IBM also plan to ship some ppc 970 using boxes, so these
> > > would probably make decent linux boxes and support would be provided.
> > >
> > > This would not help for the applce specific chipset though.
> >
> > The question is how much of the chipset is Apple or IBM specific
> > and how much will be shared. I don't believe that doing two
> > completely independent designs makes much sense.
>
> So you think that the north-bridge could be an Apple-IBM common design ?
Not exactly since there are several claims that this is an Apple design,
but the 970 bus "was designed exclusively for Apple" according to IBM.
I'm speculating, but I would not be surprised if in exchange IBM had
the right to use the northbridge in its own designs (they
also manufacture it in the same process as the 970).
In any case, this would probably be the cheapest way for IBM to build
its own PPC970 based systems. The other possibility is that they
build a hostbridge which includes only the processor and memory controller
of Apple's ASIC and modify the rest to better suit their needs, for example
by removing AGP and HT to make room for PCI-X interfaces on the host brigde.
Anyway, we'll see when IBM ships the 970 based products.
Note that I'm not very worried by the host bridge chip, once it is configured
(and the firmware should do most of it) it should work (I'm however a bit
disappointed by the fact that there is no ECC memory: 8GB without ECC
looks a bit foolish.).
> The rest is not as important, as it is connected trough a HyperTransport
> bus, so any HT chip will do. The funny thing is that they have a special
> PCI-X bridge sitting between the north bridge and the super IO chip.
This sems to be a chip with 2 HT links (to the big ASIC) and 3 PCI busses,
one to the Super-I/O, one to the 133MHz slot and one to the 2 100MHz slots.
Even sharing 2 slots on the same PCI bus at 100MHz is far from being
conservative IIRC.
> I guess this one is just a off-the-shelve piece, while the IO chip is of
> more common apple lineage. It has serial ATA, standard ATA for the
> SuperDrive (this one is a guess), USB 2.0, Firewire 800, networking and
> audio.
The schematic in the technical overview shows a link between the IO
chip and the superdrive, so your guess is probably correct.
>
> Naturally, one could even build a PPC 970 motherboard with This
> IBM/Apple northbridge,
Provided the chips are offered to third parties, that's a big if.
> some PCI bridge, and Nvidia's HT connected
> southbridges, not that we have much drivers for them though.
Drivers for which part? HT configuration is very similar
to PCI AFAIR. Many devices nowadays (USB, FW) have a few
standardized interfaces. Only network interfaces still seem
to be very varied (at least of the kinds of devices which
are important to me).
Regards,
Gabriel
Reply to: