Anyway, a few recent excerts, follow-up on what was posted at the above link
(> quoting me in earlier mail)...
(me)> Silicon itself may
be dirt-cheap, but there are many production line cost factors to think of
that are more significant: printed circuit card and extra component/wiring
mounting cycles, component tracking, alternate component options for when
just-in-time ends up being just-out-of-reach... I noted that the manufacturers
were awfully quick to drop the few components that handled selection of SCSI
bus parity or not -- so quick that their tech support was caught off guard
and the data sheets remained incorrect for almost two
years.
(Bob) Good points. But, still, that's not enough to
explain why an ATA version of a drive is $150 and the SCSI version $350.
(my reply) Could also be they all know they're selling
to a high-endian sort of market, rather than going head-to-head on the
consumer-pc market, and so feel no great incentive to drop prices.
I think a lot of people don't realize that the pricing model for computer
hardware is a lot closer to that of software than in almost any other
manufacturing endeavor. The actual incremental unit cost for a hard disk,
CPU, or memory module is trivially small. Just as with software, the vast
majority of the cost is overhead. The actual incremental cost to produce
a hard drive is probably $25; and for a CPU or memory module, it's perhaps
$2. The rest of the selling price goes to pay R&D, amortize the cost
of the production lines, etc.
> More or less, after all the disks as
such are pretty much the same, it's the controller that differs. That said,
I would think we are after the physical disk/motor mounting and clean-room
sealing, dealing with different production lines, and perhaps even different
line organizations.
I doubt that there's quite that much difference. Basically, it's a
matter of which circuit board/connector assembly you slap onto the HDA. If
you look at SCSI and ATA versions of a drive, you'll find that the only
difference is that one component, and even it has the same mounting points,
etc. I'd think that one could run a single line that randomly put out SCSI
or ATA versions.
> Well, true, the years do zip by... :)
And that's hardly a sustained rate, or even typical.
The reference to rate here was from Bob's earlier statement:
-
"... Perhaps so, but ATA achieves 33 M/sec on an inexpensive
unbalanced 40-wire cable with header-pin connectors."
Well, the ATA interface can sustain it, but that kind of sucks down
the entire PCI bus while transfers are going on....
> Still, with SCSI, we're dealing with
a daisy-chain bus that can extend many meters, driving numerous units at
sustained high throughput. That does put somewhat different demands on cables
than the short to very short internal ATA connections.
Good point.
Probably the biggest cause today (cost difference) is simply
total volumes sold. The price difference between ATA and SCSI wasn't all
that great some years back, I recall something like 20-30% more for the same
capacity in SCSI, when we were still talking 100 to 400 Mb, but the ATA drive
prices fell faster and SCSIs became somewhat rare to find with the
suppliers.
Yes, that's the point I was making earlier. Back when drives were quite
expensive a "large" (for then) ATA drive might be $800, and the same drive
in SCSI was $1,000. If the percentage had stayed the same, that $150 ATA
drive would be available in SCSI for $187.50. I'd happily pay that additional
25%/$37.50 for SCSI, as I suspect many people would. But the problem is,
the absolute $200 difference has remained, and I'm sure not going to pay
133%/$200 more for SCSI.
> Could also be they all know they're selling
to a high-endian market...
You may be right, but that's just stupid. On the other hand, drive
margins on ATA are next to nothing, while they remain high for SCSI. A
manufacturer who makes one nearly always makes both, so perhaps they figure
they'll win either way--they won't lose a sale to someone who wants ATA,
but they'll make some bucks from someone who insists on SCSI.
Pretty much sums up the situation. SCSI is today a product for people willing
to pay the premium price for what they perceive is the premium performance.
In other words, not that many, given the "good enough" ATA performance seen
by your average consumer. I seriously doubt SCSI will die completely, at
least any time soon.
