|  |     Well, this is somewhat involved...
    The PCBU was FIS'ing an IDE DMA driver on NT 4.0 until we switched
    from Toshiba CD-ROMs to Goldstars, whereupon CD audio broke, so we
    went back to using ATAPI.  You can check by firing up the Services
    icon in Control Panel and looking for PIIXIDE.  If it's there (and
    ATAPI is disabled), you're doing DMA.  If ATAPI is enabled, you're
    not.
    
    The Seagate Barracuda (7200rpm) benchmarks faster than the Quantum
    EIDEs (even with DMA).  However, to save some (serious) bucks, the
    Barracuda was replaced with the Quantum Fireball Tempest SCSI, which
    is no faster than EIDE (in fact, now slightly slower than the equivalent
    2.1Gb EIDE model; same disk, now there's SCSI overhead).  If things
    are normal (i.e. FUBAR...), you can't tell which SCSI disk you'll get
    from the SKU order number (only the BOM) until the Barracuda's are
    completely used up.
    
    Keep in mind typical benchmarks don't tend to take advantage of
    DMA... the processor is just sitting there waiting anyhow.
    .02 K
 | 
|  | Do we not need to be very careful here? If we look at individual device
performance the issue of SCSI verse EIDE may lean to EIDE based on cost
performance. However, it is my understanding that SCSI will become more
attractive when you consider:
1)	Limitation (verses SCSI) of total number of EIDE devices
2)	Impact of concurrent I/O i.e. SCSI much better in this area.
3)	Variety of SCSI supported devices (eg. scanners) - no need for
	new interface cards (assuming <= 7 SCSI devices total).
4)	SCSI support for external devices (eg. Tapes, CD-ROM writers)
So besides individual disk performance you do need to consider overall
system performance and possible upgrades. 
Wayne
 |