Re: tlug: Some hardware questions...

[Karl-Max Wagner <karlmax@example.com>]

> SCSI contollers often take most of the load of controlling the
> drive away from the CPU.  I haven't seen IDE do this yet, but if
> they're starting to, great!

There seem to be some around. Not widely spread at all.

> The line numbers are off.  First 8, then 16, data lines.  The
> next steps are only improving quality, and speed, not # of lines.
> And some are (sort of) compatible.

Could be. I didn't follow things much after SCSI II lately.... 

> * By compatible, I mean you can use all the devices together.  
> With Fast-20, the fast devices will go fast, the slow ones slow.
> LVD reverts back to Fast-20.
> 
> Now you all know why SCSI can be so confusing, especially when 
> sales people themselves mix things up and create new names.

There are also cases when some makers interpret the specs in an
intergalactic way. There were some Quantum drives in the past
that violated the timing specs.

> DEFINITELY.  I had to study them at IBM.  Good sleeping material.
> They're available at:
> ftp://ftp.symbios.com/pub/standards/io/x3t10/drafts/

Good to know. I used to have a paper copy but I don't have
access to it right now.

> However, I've yet to see an Ultrawide-SCSI adapter card that didn't
> also include a narrow adapter plug internally.  My guess is that you 

Ah ! That is good news....

> One other confusing thing to note.  Internal SCSI cables:
> The 'wide' cables with more lines are physically narrower than the
> 'narrow' cables.

In standard cables each second line is actually ground. So they
probably use a ground ribbon in the new ones ( better anyway ).

> > What about things like Ultra-ATA that claim a 33 MB/sec tranfer 
> > rate, the same as UltraWide SCSI?  Do they really do this, or 
> > is this a theoretical capability of the drive that cannot be met 
> > by the adapter card in the computer?

> Ultra-ATA's 33 MB/sec is the theoretical capability of the adapter
> card in the computer and drive, that rarely (read almost never) can
> be met by the physical drive.  The fastest drives (SCSI of course) 
> out there can sustain a speed of 10-15 Meg a second.  All drives have
> buffers in them (< 1 meg), and that's what might actually use the
> 33MB/s  bandwidth, assuming you're OS doesn't do any caching (yeah,
> right.) Typical drives max out at sustained rates of 8MB/s or so, I
> think. So why all this fuss over fast bus speeds (33 or 80 MB/s), when
> the drive can't keep up?!?  For SCSI, it makes sense.  7 devices, all
> wanting to be used, at ~10MB/s each = 70MB/s.  Excellent for a RAID
> system.

There is another catch: IDE traditionally doesn't use
termination. For 33 MByte/s to work OK, the cables have to be
very short or the data get mangled.

> BTW, I'll take this moment to HIGHLY recommend Linux's software RAID
> setup for anyone with SCSI.  If you have 2 drives, you can get about
> 1.5 times the speed.  (If it was ideal, it would be 2 times the speed.)
> And don't think you have to lose space.  That's only if you want 
> increased reliability (which the driver doesn't easily provide anyway,
> last time I checked).    I use it for some ancient small (~100 meg ea.)
> drives to bring them to a reasonable partition size & speed.

That's gonna be needed for my two Apricot 600's, too.....

================================================================
"It was hell. They knew it.          Karl-Max Wagner
  But they called it                 karlmax@example.com
    W-I-N-D-O-Z-E"
================================================================
---------------------------------------------------------------
Next Meeting: 10 October, 12:30 Tokyo Station Yaesu central gate
Next Nomikai: 20 November, 19:30  Tengu TokyoEkiMae 03-3275-3691
---------------------------------------------------------------
Sponsor: PHT, makers of TurboLinux http://www.pht.co.jp