Hi,

> Is that how it does 10Mb/S internal and 5Mb/S external (if
> that's right)?  Or is it simply clocked at 5MHz and the
> internal segment is 'slow, wide' (which seems unlikely)?

I think there's a bit of confusion. No Macs ever came with wide SCSI on
the motherboard. Nor is wide (non-ultra) SCSI even popular or common.

Most of the common PCI PowerMacs (7200 to 9600, generally) have two SCSI
busses, using two SCSI controllers. The one that is connected to the
external connector is configured to run at up to 5 MB/sec, which means
that the SCSI chip apparently likes to run five times faster than the bus:

esp0 at obio0 offset 0x10000 irq 12: NCR53C94, 25MHz, SCSI ID 7
scsibus1 at esp0: 8 targets, 8 luns per target

The internal bus is 10 MB/sec, but only when using syncronous (Fast SCSI)
mode.

mesh0 at obio0 offset 0x18000 irq 13: 50MHz, SCSI ID 7
scsibus2 at mesh0: 8 targets, 8 luns per target

NetBSD does not put try to use syncronous mode with this chip because
there are a number of bugs in the chip which make it not as reliable as
we'd like. Perhaps someday someone will write code to circumvent all of
those bugs, as I'm guessing that this is what the LinuxPPC people have
done:

> The Linux driver has either a run time or compile time option, I
> forget which.  Scary.

I'm pretty certain that this is talking about using syncronous mode, as
the actual clock doesn't appear to come from a software controllable clock
generator.

> I lay awake this morning wondering whether certain multi
> -media applications might benefit from a second (probably
> external) disk on the SCSI bus.  Suppose there was an MPEG
> decoder or CD writer on the same bus: am I right in thinking
> that long streams of data could take place between the SCSI
> devices, by-passing the mainboard and microprocessor?

Yes, but not much software is smart enough to do transfers between the
drives themselves. But you're assuming that the external and internal
busses are connected; they are not.

John Klos
Sixgirls Computing Labs