[rescue] 36G SCA drives still looking

[Nathan Raymond]

On Sat, 5 Mar 2005, Charles Shannon Hendrix wrote:

> What I don't understand is why the industry wasted so much time IDE.
>
> SCSI was better, and at the same quantities, the price should be about
> the same.

SCSI never sold in the same quantities as IDE, which is why IDE was always 
less expensive.  Also, for systems with 1-2 drives, a single IDE 
controller was less expensive than SCSI.  Especially when you factor in 
that IDE was only PIO and didn't even have DMA until relatively late in 
it's life...

> I don't see any reason for IDE to have ever existed.

It was simpler and therefore cheaper and people didn't care.

> Also, since IDE drives have to have a controller on each drive,
> shouldn't they technically cost more than an identical SCSI drive?

Slave/Master was a simple (and stupid) way of making IDE even less 
expensive.  If the majority of computers sold actually had 5-7 storage 
devices per system, SCSI absolutely would have been cheaper, but computers 
shipped with 1-2 storage devices per system for a long time.

> Or do the extra capabilities of SCSI mean that their PCB is just as
> complex?

DMA, scatter-gather, device disconnect, and bus mastering were all common 
technologies with Ultra-Wide SCSI when IDE was still crawling on the 
ground in PIO mode.  SCSI was more expensive to develop and produce 
(especially at the controller level) for a long time, cabling was more 
expensive because it had to handle higher speeds and greater quality and 
you had to deal with things like termination, and most people didn't know 
any better or care at the consumer level.

Now we have SATA, which takes a subset of SCSI (ATAPI) and marries it to 
fiber channel packet technology, and is now inheriting other SCSI 
technologies like NCQ.  Unforunately there is a problem with a few 
different standards bodies for SATA, resulting in lack of industry 
cohesion.  And then folks like NVIDIA go and disable SATA-II capabilities 
in the nForce 4 chipset, but have it enabled in the nForce 4 Ultra, for 
reasons of product differentiation (and good luck trying to find a PCI or 
PCI-Express SATA-II card right now).  Also, NVIDIA's implimentation does 
not scale as it should in terms of NCQ performance, while Intel's chipset 
implimentation does.  Intel sees SATA-II as a way to push chipset sales, 
as a value-add for motherboard integrators, as much as any storage 
technology on it's own.  Intel doesn't hesitate to have a PCI-Express SCSI 
RAID storage card, but doesn't carry a single PCI-Express SATA card.

Ah, the games companies play...

- Nate