I/O transfer rate for multiple IDE devices

[RoamDog]

I have two IDE DVD writers.

Am using an Asus P5Q Pro MB with 1 IDE port (supporting a master and a slave device). And I have a Rosewill RC-200 PCI IDE (ATA) Silicon Image RAID (0/1/0+1/JBOD) Host Controller Card that will support an additional set of IDE devices via the PCI slot.

Am I better off connecting both IDE DVD writers to the IDE port on the mobo (which obviously does not use the PCI interface) as a master/primary and slave/secondary. Or am I better off connecting one to the IDE port on the mobo as a master and connecting the second writer as a master on the PCI slot IDE controller?

And why?

(I seem to remember that there is a performance penalty for putting both DVDs on the same channel as Master/Slave (esp. when trying to use both simultaneously for read/write) but was not sure whether that was a bigger penalty than going through the PCI interface).

 

[supremelaw]

My understanding of the old PCI bus is that all devices connected
to it must arbitrate bus access, because that bus can only service one
PCI device at any given time.

So, it shouldn't matter, as long as all of your IDE devices are
using the same PCI bus internally.

I'll try to find a block diagram for the P45 chipset, and
if I find anything to the contrary, I'll post an update here.



Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[supremelaw]

Here you go:

http://www.asus.com/products.a...l1=3&l2=11&l3=709&l4=0


Intel's P45 chipset block diagram does not show a "native" PCI bus, however:

http://supremelaw.org/systems/...pset.Block.Diagram.jpg


My guess is that ASUS added a Marvell controller for IDE devices:

Marvell 88SE6111
- 1 x UltraDMA 133/100/66 for up to 2 PATA devices


And, I'm not absolutely sure about this, but
"IDE Bus Mastering" functions spill over into firmware
and low-level device drivers, which manage a PCI
bus by only allowing one device to do I/O at a time --
up to a theoretical maximum of 32 bits @ 33 MHz
= 133 MB/second (MAX HEADROOM).

That's the same "133" as is found in the nomenclature "ATA-133".



Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[supremelaw]

So, if you want to achieve a bit of parallelism,
I would recommend that you search for a PCI-Express
card that supports IDE ports -- in order to maintain
backwards compatibility with aging IDE devices
that were designed to connect to the old PCI bus
i.e. long before PCI-Express became the industry standard.

I don't know if such cards even exist, but you should
be able to search for same using Google and
the right keywords e.g. try Addonics:

http://www.addonics.com

... and/or CoolDrives:

http://www.cooldrives.com/


A quick email message to their sales or tech support groups
should get you an answer promptly.


Look for such PCI-E to IDE adapters that plug into x1 PCI-Express slots,
because the latter have plenty of bandwidth for ATA-133 devices.


Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[supremelaw]

Voila!

http://www.addonics.com/produc...ller/adi2s3gpx1-2e.asp



Sincerely yours,
/s/ Paul Andrew Mitchell, Inventor and
Webmaster, Supreme Law Library

All Rights Reserved without Prejudice

 

[RoamDog]

Thank you for a very thorough and informative reply. Very helpful.