>>>>> "m" == Michael <macallan%netbsd.org@localhost> writes:
>>>>> "dm" == Dave McGuire <mcguire%neurotica.com@localhost> writes:
>>>>> "gd" == Gert Doering <gert%greenie.muc.de@localhost> writes:
m> IIRC the U5/U10 doesn't like 3.3v only PCI cards. Neither do
m> older Macs for example.
doesn't make sense. There must be a longer story about what's really
going on.
dm> Look for the notches in the card-edge connector. If the
dm> notch is on the bracket end, it's a 3.3V-only card. If the
dm> notch is away from the bracket end, it's a 5V-only card. If
dm> both notches are present, it's a "universal" card
yes, but since the cards are chinese it's prudent to assume the
notches are a slop trough of lies. What's more important is the slot
in which the card normally works. We went around this track before
last time the rumor came up: there is *no such thing* as a
dual-voltage slot (think about it. it's a bus.), and all ordinary
32-bit chinese-compatible PCI slots are 5V, as are all the slots,
period, in the U5/U10, so...no, it's something else.
Then again, I just got something else wrong today so...
gd> (I've always been a fan of SCSI, but these days, my heap of
gd> incompatible SCSI cables and adaptors plus heaps of slloooww
gd> 1G/2G/4G disks really frustrates me...)
SAS is evolving to be pretty good: it has infiniband-like backplane
switches, in that they're capable of link bonding. And it seems to be
sort of single-vendor in practice: like Infiniband is Mellanox, SAS is
LSI. The icky part about it is the amount of proprietary code
involved: the software involved in the ``firmware image'' is really
out of control with too many knobs in there for stuff like multipath
and link-quality telemetry. Also I do not understand all this command
tunneling and translation that's going on all over the place these
days whenever you try to run smartmontools or wodim. it's like a
punting bucket brigade of unapologetic clowns.
Attachment:
pgpk0U7AjlMEZ.pgp
Description: PGP signature