[plug] Serial ATA

[Craig Ringer]

Cameron Patrick wrote:
> On Sat, May 17, 2003 at 05:48:26PM +0800, Craig Ringer wrote:
> | SATA is mainly nice because of its hot-pluggability, nice-to-work-with 
> | cabling, airflow improvements, and lack of reliance on the awful 
> | "master/slave" scheme. The speed just isn't much of a benefit, and quite 
> | frankly I don't think there will be a drive that can properly flood 
> | ATA100 before SATA rev 2 comes out at 300+mbps.
> 
> I was under the impression that (a) most current SATA hardware uses
> PATA-to-SATA converter chips anyway

Most controller hardware does, but there's good hardware that doesn't 
and more chipsets can be expected to implement it natively soon. 
Especially if people make a point of buying the proper SATA 
implementations which are 150mbps capable and hot-swappable instead of 
ATA133 with a different connector.

> and (b) the PCI bus has less bandwidth
> than SATA anyway - 133 MB/s vs 150 MB/s IIRC.

Well, 33MHz/32bit does, but /good/ PCI will be at least 33MHz/64bit if 
not 66MHz/64bit. Then there's more serious PCI (becoming available on 
"consumer" hardware at the very, very top end now) like PCI-X 100MHz and 
133MHz 64bit-wide PCI.

So I s'pose the very original PCI can't take it, but you can expect 
higer-bandwidth PCI types to connect the ATA controllers in the future.

Anyway, the actual drive bandwidth can't flood a current PCI bus, not by 
a mile. It'll matter a lot if you have a high-speed NIC as well, though.

Craig