Wide ULTRA 3 SCSI 18.2 gb How Much?????

[stopthegop]

They're probably server drives.  If thats the case, most likely these have an "SCA" type connector which is 80 pin (male on devices, I think).  Anyway, yes these will work with any Amiga scsi controller, but with a couple of catches.  First you need an SCA adapter to transform the drive into whatever type of scsi connector your internal passthrough cable and your Amiga controller have.  No Amiga controllers (that I am aware of) have 80 pin SCA connectors.   For example, you'd need an SCA(f)->68pin(f) if using the controller on a Cyberstorm MKIII.  In this instance, you would also need a passthrough cable equal to the TOTAL number of drives you plan to connect PLUS THREE.  Four drives -> 7 position cable.  Next, the SCA adapters usually have terminating resistors installed by default.  These must be removed before using them in an Amiga.  Finally, be aware that the speed of these drives means nothing if you don't hook them up to a SCSI II or III controller which is specified as "WIDE" (16 bits per word vs. 8 bits per word - or "narrow").   Any scsi controller can talk to these drives; the problem is the entire bus will automatically synchronize down to the speed of the slowest device on the bus.  If the controller itself is the slowest device, then guess what??  Everything goes that speed.  With scsi, its "follow the slow-poke".     My advice is this:  yes, by all means hook them up to an Amiga.  But only if you have a SCSI II or III controller such as the Commodore 4091 or the onboard controller on Cyberstorm cards.  And definately do not connect crap like scanners, cdr, or zip drives to this bus!  One slow device on the chain will just kill your throughput.  Best of luck.      

[stopthegop]

You're mistaken.  If you have a "Narrow" (8 data bits per cycle) device on a bus with "Wide" (16 bits per cycle) devices and "Wide" controller(s), then the overall width of the data path is 8 bits.  How are you going to cram 16 bits through only 8 wires in a single clock cycle?  The answer is you can't.  You can do it in two clock cycles, but not one.  That effectively cuts your overall performance in half, which is what I said earlier.  I know this is true.  I work for a manufacturer of SCSI periphreals.  I've done tests myself in the lab and in the field.  Granted, there is one way to "cheat" that I know of; that is to get decent performance on a Wide bus even with a Single-Ended device attached.  In order to work, the "slow" device must not reside physically between the "target" and "initiator" hosts, both of which are presumably Wide.  Second, it won't work with active termination.  If these two conditions are met, then yes. it is possible to get 'close' to rated throughput.  This is more true with a short bus length using LVD (Low Voltage Diff) devices. This is a dirty solution, imo.  And it won't work on the Amiga/Cstorm controller anyway because it requires active termination.