Conference noted::ibmpc-95

    Um, along the lines of SCSI drivers.  I was checking out Quantum disks
    and noticed that they are listed as Ultra SCSI.  Their description of
    this off of their web page is that is is different that SCSI-2 and
    SCSI-3 - 
    
    Should I worry about that? (ie will SCSI controllers support it?)
    
    bjm

    re: .59 by hndymn.zko.dec.com::MCCARTHY
    
    Ultra SCSI disks should be hardware (and backward) compatible with
    existing FAST SCSI, you just won't be able to utilize the benefits.
    I don't think software cares, other than ultra SCSI host adapter
    drivers are probably different.
    
    /bill

While you were on the Quantum web page, you should read in the technical section
under white papers, they have a real good write up of Ultra SCSI. I also have a
mirror of that paper on my groups web page at;

http://whatsa.shr.dec.com/disks.htm

In a nutshell, to use Ultra you need Ultra drives and Ultra controllers with
half the cable length of Fast-10 busses. Folk will soon be coming out with
widgits to releive the length restriction, but, for Joe average you will
probably find that even if you have compliant controllers and drives, you'll
still need to restrict it to fast-10 to be able to cable in that old tape drive
and cdrom's and external rz55 box ;^)

roger.

Roger,

Thanks for the pointers.  I found the descriptions of Ultra on Quantum's and
Seagate's pages - Seagate's seemed to be a bit more descriptive and helpful
to me.

bjm

    Also worth taking a look at the Adaptec pages... Though they are
    serious biased towards SCSI and away from IDE/ATAPI.
    
    Not that that is such a bad thing....
    
    Chris
    

    
    re.62
    
    Your welcome. And if you find any other "must have" pointers, mail them
    to me so I can review and update my page.
    
    Roger.

    What is Ultra SCSI ?
    
    Olivier

>>    What is Ultra SCSI ?

Go check out the URLs back a few.  Those pages go into details.

Brian J.

An in a nutshell;

Ultra SCSI is SCSI with the clock rate during data transfer phase doubled.

> Ultra SCSI is SCSI with the clock rate during data transfer phase doubled.

Actually I think it's doubled twice.  Depending on your starting point.

Original SCSI moved bits at a 5 Mbit/sec/wire rate.
Fast SCSI uses 10 Mbit/sec/wire.
Ultra SCSI uses 20 Mbit/sec/wire.

One thing I didn't find described is the difference between (actually the 
limitations of) SCSI and wide SCSI.  I think the external connectors are
different but can I mix and match on the same cable?  I'm still limited to 
7 IDs right?

Brian J.

With wide SCSI in theory you should be able to have 15 IDs...at least the
standard allows it.  That's assuming all the software and hardware are up to
snuff.

Also from a logical point of view, a narrow disk works fine on a wide
SCSI...the mode is negotiated up front per device.  However, that assumes that
the cabling works out right.

Burns

Wide SCSI uses a mini-68 pin connector on both internal and external
devices.  Its ribbon cable for internal devices is actually physically
narrower than "narrow SCSI."

Most boards have at least one narrow connector (usually internal) on them,
in addition to the normal collection of wide connectors.  Mixing and
matching narrow and wide devices has been successful for me, as long as I
put the narrow devices on the board's narrow connector.  Trying to use
wide-to-narrow adapters on narrow devices has failed, probably because of
device ID conflicts (that is, narrow device ID 1 conflicting with wide
device ID 9, which isn't a problem if I leave the narrow devices on the
board's narrow connector).

Although the board will negotiate wide mode at initialization, this is
independent of a device's addressibility.  In other words, you can have a
wide device (which has the extra address line) answer to ID 9, but still
operate in "narrow" mode.  Why you'd want to is a different issue.

Re: Normal vs. Fast vs. Ultra

I believe every clock doubling reduces the maximum allowable bus length by
1/2.  Original was 9 meters (I think), fast 4.5, and ultra probably 2.25.

I've hit the maximum SCSI bus length on fast/wide.  Things don't work so
well when you hit it.

    I think the max lengths are 6m, 3m, 1.5m.  Not much room on a 1.5m
    cable for very many devices.  I've been seeing some ultra wide
    *differential* host adapters and devices lately.  That allows much
    longer signaling distances, but I don't know what the limits are.
    
    Something I haven't seen mentioned yet re: wide vs. narrow.  Perhaps I
    missed it.  Wide has a 16-bit datapath, narrow has 8-bit.
    
    /Bill

Re: Length

Yeah, that sounds more appropriate.  And, you're right; not much room for
devices, especially if you observe the restrictions on minimum distance
between devices.

Re: 16 bit:

Yes.  Then again, most hard drives have a native transfer rate (how fast
the heads actually fly over the bits) that could easily be handled by fast
SCSI.  Wide SCSI doesn't buy you any real disk transfer rate improvements,
although with proper buffering in the drive, it does cut down on bus
contention, and it is nice to have 8 more devices before having to buy
another SCSI adapter.

Why does wide device 9 conflict with narrow device 1?  You don't put the
address out in binary...you raise the data line whose number corresponds to
the device number.  If you raise data line 9, a narrow device shouldn't even
see it.

Burns

    re: .74 by skylab.zko.dec.com::FISHER
    
    >> Why does wide device 9 conflict with narrow device 1?
    
    It would have to be because the SCSI ID is encoded.
    
    1 = 0001
    9 = 1001
    
    A narrow device looks at just the lower 3 bits hence the conflict.
    
    /Bill

    Nope, that's not how it works....
    
    The ID is put on the bus in bitmap form, not encoded.
    
    High Bus			Low Bus
    15 14 13 12 11 10 9 8    7 6 5 4 3 2 1 0
    
    0   0  0  0  0  0 0 0    0 0 0 0 0 0 1 0	= ID 1
    
    0   0  0  0  0  0 1 0    0 0 0 0 0 0 0 0    = ID 9
    
    
    
    Priority of Id's is as follows:  7 (highest) - 0, 15 - 8 (lowest)
    
    Any device wanting to use the bus must arbitrate for the bus, and
    not take it if any device of a higher priority is arbitrating.
    
    For a narrow bus, you are limited to 8 ID's. The controller is
    typically at ID 7, the highest ID.
    
    For a wide bus, you have 16 ID's, and the controller is also typically
    at 7, the highest ID priority.
    
    What this gives you is the ability to mix narrow and wide devices on
    the same bus. A narrow device can only be and see id's 0-7, but as all of
    these are higher than 8-15, it will always win. The wide devices can be 
    at any ID, but if they are on 8-15, they must monitor the low bus ID's,
    and not win arbitration if any ID on the low bus is asserted.
    
    Refer to the Adaptec web site for a more in-depth review of cabling,
    etc... www.adaptec.com
    
    Chris
    

    re: .76 by LEFTY::CWILLIAMS
    
    Thanks, Chris.  Makes sense not to dedicate lines to SCSI ID
    when the datapath will do.
    
    I bow to the SCSI gods...  ;-)
    
    /Bill

The real reason have trouble with wide drive on narrow busses is;

Most old narrow busses/cables/terminators/controllers comply to the SCSI-1 rules
that say "all unused signals should be GROUNDED". On a narrow bus that means
that the pins that are used to carry data bits 8-15 and p2 are gounded. When you
introduce a wide drive onto that bus during ARBRITRATION phase the wide device
sees this as devices 8-15 (bits 8 through 15 are = ground = asserted) are on the
bus before him at a higher priority and the wide device will back down. This in
term means you (the initiator) will never get a response from the wide drive, he
is for all purposes invisible.

That is why you can put a RZ28M-VA (narrow) a BA356 box (wide) but you cannot
put a RZ28M-VW (wide) in a BA350 box (narrow).

Now if your wide device has on-board split termination as some of our newer
solid state disks that are coming out have, you can unterminate the lower 8 bits
for use on a narrow bus and sever/terminate the upper 8 bits with a jumper on
the device.

roger. 

    I'm running a combo bus of a wide disk with everything else narrow,
    going dual port off a AHA2940UW... Theres a wide bus going to the disk,
    terminated at the disk. There's a narrow bus going the rest of the way,
    with the high bus terminated at the controller, and the rest terminated
    at the end of the narrow bus.
    
    The possible problem here is signal skew due to the 2 halves of the bus
    being different lengths. I'm not seeing a problem, but it can exist,
    and is real. A better config is 1 wide bus, with 68-50 pin adapters
    connecting the narrow devices to the wide bus. Those adapters are not
    cheap, though ($15-40 each), and you need one for each narrow device.
    In my case, I'd need 5. Ouch.
    
    Chris
    

> Most old narrow busses/cables/terminators/controllers comply to the SCSI-1 rules
> that say "all unused signals should be GROUNDED". On a narrow bus that means
> that the pins that are used to carry data bits 8-15 and p2 are gounded. When you

I'm not sure this makes sense (to me) either.

The extra byte that makes the bus wide are new signal pins, aren't they? 

    Yes, that's true. The problem came in StorageWorks bricks - the
    internal flex cable designs grounded pins on the custom connector that
    were re-defined for wide busses. 
    
    It's not a generic issue. 
    
    Chris
    

Chris is correct. The pins in StorageWorks were not "new" to wide. In generic
you run the opposite problem, if you use wide drives on a narrow bus with 50-68
pin adapters, the upper bits may be "float" instead of ground. That causes it's
own problems.

    .76:
    
    I did not know that.  Thank you Sir!
    
    (It clears up at least one misconception of mine.)

    
    completly unrelated to the previous discussion, but can anyone tell me
    WHY external SCSI cases are so @#%$* expensive?
    
    You can by a full tower w/250 watt power supply and 7 internal bays for
    $45 but a much smaller 7 bay encloseur with a 200W power supply for
    SCSI drives is $180 - $260!! 
    WHAT's the DEAL?
    
    Garrison
    (and NO, volume is not a believable reason.)

    Garrison,
    
       Grow your own out of a tower case.  You will probably have to make
    the cable and bulkhead as well.  Make sure to put a bunch of LED's on
    the front for the activity lights.  Thats the problem with computers
    these days - not enough blinking lights!
    
    Kurt

Building one's own is certainly a good idea.

I think they're just used to selling to the high end and server markets. 
You know, where transactions start with a purchase order, rather than check
or credit card, and price gouging is the name of the game?

    Re .86
    
    Your right about the lights, you would think that computer
    manufacturers had never seen the original Star Trek.  Almost nothing
    BUT lights! Now THOSE were computers! :)
    
    As far as the tower case, yes that would certainly work, but you end up
    with a monster size case to do the same thing.  Those 7 bay SCSI
    enclosures are compact!  Stuff one under a desk on top of your
    mini-tower system and still have knee room.
    
    Yeah, I guess it must just boil down to pure GREED.
    
    Hmmm..I'm greedy, maybe I should start selling SCSI enclosures
    as a new sideline. :)
    
    Garrison

    >>                       Stuff one under a desk on top of your
    >>mini-tower system and still have knee room.
    
    If it did have a bunch of lights, you wouldn't be able to see them
    with it parked under the desk. :-)
    
    LEDs/lights are typically only for maintenence purposes, and as
    such don't get a lot of demand by customers, anymore. Ninty-nine
    percent of the time customers/consumers want a box they can park
    somewhere out of the way, and only concern themselves with the
    monitor, keyboard, mouse egonomics. There are limited reasons/purposes
    where a few LEDs/lights are needed for the consumer/customer to see on
    a regular basis.
    
    
    Bob

    Roaming a bit further off-topic: in fact "lites" are making a comeback
    in a big way - at least in the list of customer requirements for
    servers. 
    
    Combining the "downsizing" of corporations (which affects the number of
    hardware gods on site), the always-present desire to minimize
    down-time, the advent of hot-swap SCSI, and software management
    "agents" running across LANs and even WANs, we're being asked to
    visually identify a faulted storage device. A system may call for help
    to replace a drive that's going down the tubes - and even provide the
    FRU to bring, but when the non-skilled humanoid arrives, having the
    system point to the drive to be replaced makes a lot of sense (rather
    than have said non-skilled humanoid try to figure it out)...
    
    Buy stock in companies that make orange LEDs as this industry is
    gonna be using a lot of 'em ;^)
    
    /dave

    >Your right about the lights, you would think that computer
    >manufacturers had never seen the original Star Trek.  Almost nothing
    >BUT lights! Now THOSE were computers! :)
    
    Ah, but as technology marched on, the flashing lights and all those
    buttons and levers disappeared to be replaced by displays (2D and
    holographic) and touch panels - both in the Federation and in the
    Romulan, Cardacean (sp?), and Klingon Empires (ref: STTNG, DS9, ST
    Voyager, and the ST movies).  Obviously, history repeats itself!

I wish my system had more lights.  My PeeCee Power & Cooling room-warmer
box only has three leds: Power, HDD, Turbo.  I've wired up the HDD and
Turbo lights to my two SCSI controllers.  I must say, when I'm scanning
directly to disk, or doing a backup, it is quite satisfying to see both
busses light up.

Then again, with the cabinet's 4 fans, the CPU fan, and the 2 7200 RPM hard
drives (all of which conspire to hit the resonant frequency of the
enclosure), the Power LED tells me something that my neighbor could
probably tell me sooner.

Maybe I'll yank the Power LED and hook it up to the IDE controller...

    Never mind the lights, how often could you have used a "write protect"
    button ? I want my "write protect" buttons back...

    And toggle switches for toggling programs directly into
    memory.  Bring back the PDP-12; enough toggle switches
    and lights to satisfy anybody!

the -12 was truly a hacker's machine!