[geeks] The new IPC/LX, from Dell?

[Joshua Boyd]

On Thu, Nov 19, 2009 at 06:18:23PM -0500, Lionel Peterson wrote:

> When an eSATA drive is hooked up to mainboard controller port (in this  
> case an intel notebook chipset) it is as fast as the on-board controller 
> can go, up to the speed of the drive.
>
> On a USB 2.0 or FW port, there is an additional limiting factor, the  
> bridge chipset. A well-designed chipset could get out of the way, but I 
> don't think ultimately be able to be as fast as the native drive  
> interface. The processing overhead will approach, but never equal zero.

Certainly there is overhead.  

Back in the IDE days, it wasn't uncommon to find that an IDE drive
attached by firewire could be faster than that same IDE drive attached
to the internal IDE port.  IDE presumably has less overhead than
Firewire, yet clearly IDE was enough efficient that offloading the job
of talking to the IDE drive could accelerate the write process under
several types of tests.

eSATA isn't faster than SAS, even though SAS can talk to a lot of
drives, so without someone breaking open the firewire protocol and SATA
protocol documents, we don't know for certain that eSATA has less
protocol overhead than Firewire, we only know that eSATA has a much
higher clock rate.  For all I know, firewire could be more efficient
than SATA or SAS.