problem with UDMA mode on XP1000

Jameel Akari jakari at bithose.com
Tue Apr 15 08:55:22 PDT 2003


On Tue, 15 Apr 2003, Kevin A. Pieckiel wrote:

> On Mon, Apr 14, 2003 at 09:22:00AM -0400, Andrew Gallatin wrote:
>
> >   Rant aside, ATA cables are VERY sensative to EMI noise.  What's
> >   happening to you is that the cable is either badly twisted, running
>
> Are these types of issues IDE specific, or are there similar issues with
> SCSI cabling as well?

	Yes and no.. IDE's big failing (IMO) is that it's basically
unterminated.  It's tuned for a certain signal level and cable length in
the hope that you don't need termination.  There are a lot of variables,
but in my experience IDE is more susceptable to radiated noise than SCSI
cabling.  I was always under the impression it was tuned for an 18" cable
length but I'm seeing all sorts of sizes these days; I seriously doubt the
36"+ cables I see around are within spec.

	I've run max-length SCSI cables internally (with 5-6 attached
devices) in a rather noisy system without any ill effect.  OTOH, plain
40-pin IDE didn't fare well in that system.. CDROM was ok, hard disk was
flaky.  Switching to "UDMA" cables (80-conductor) resolved the issue, even
though it wasn't a UDMA aware controller or drive in question.

	My Tivo used to randomly crash until I replaced its cheap 40-pin
cable with a new UDMA one.  Now it randomly crashes because I forgot to
re-enable swap, but that's another matter... ;)

	I had the same sort of problem in an Alpha PC164, but in the end
the onboard IDE was so terrible and Tru64 refused to run a hard disk on
it, so I bought an ISP1040 and went SCSI, and it's been solid.

--
#!/jameel/akari
sleep 4800;
make clean && make breakfast



More information about the freebsd-alpha mailing list