Optical Drive On an IDE Expansion Card

  • Thread starter Thread starter Curt
  • Start date Start date
C

Curt

Is it possible to do this? If so, how much of a perfomance hit can I
expect...if any?

Thanks,

Curt.
 
Curt said:
Is it possible to do this? If so, how much of a perfomance hit can I
expect...if any?

Thanks,

Curt.
Click to expand...

First, you have to find an IDE expansion card (PCI or PCI Express)
that handles ATAPI. Not all do. Some will warn you that they don't
work well with optical drives. Check customer reviews on Newegg before
you buy something.

An optical drive, doesn't have a lot of performance to begin with.
To give an example, I was testing a DVD the other day on one of my
newer drives, and I was shocked it made it up to 8MB/sec on reads. I
expected less than that. A DVD drive can be handled well, even with USB2
bandwidth (30MB/sec). The 33/66/100/133MB/sec of IDE is hardly
taxed by an optical drive. While in the distant past, there were cases,
where mixing radically dissimilar devices on an IDE cable, resulted
in slowdowns, you shouldn't see that happening today. The claim
today is, the drive(s) function independently on the same cable.

So rather than performance issues, instead it's a compatibility issue.
If the card says "supports ATAPI and works with all optical drives",
then buy it.

A major problem now, is finding good cards for sale. IDE is obsolete.
Some chips for IDE have gone out of production. And this isn't a good
thing, because what remains is too small a stock of parts, to pick and
choose. (The CMD0680 and ITE8212 are gone. Promise IDE cards are gone.
VIA VT6421 remains.)

It might actually be better, if you had a PCI Express x1 slot on the
motherboard. Then, you might have access to a Jmicron controller.
This tiny card, uses a JMB368 and has one IDE connector. I think the
chip is similar to the one on my motherboard (which has worked with
my optical drive). You need a PCI Express slot for this. Or,
a company brave enough to use a PCI to PCI Express bridge chip with
one of these.

http://us.startech.com/product/PEX2IDE-1-Port-PCI-Express-IDE-Adapter-Card

This is another PCI Express x1 card, with a Jmicron JMB363. People seem
to be having problems with this, but it's probably the card design
(firmware chip), rather than the Jmicron part. Jmicron is used on
a lot of retail motherboards, to add an IDE port.

http://www.newegg.com/Product/Product.aspx?Item=N82E16815280010

Another option, is go with an IDE drive to SATA motherboard dongle.
But even those are disappearing. It's really not a happy situation
at all.

In the past, some of the chips for dongle design, were bidirectional
in design. That meant they could convert SATA to IDE or IDE to SATA.
This is an example. I've never heard of this particular chip or
the company that makes it.

http://us.startech.com/media/products/PATA2SATA3/Manuals/PATA2SATA3.pdf

What you'd need, is a SATA port on the motherboard, plug it to the
dongle, and the other side of the dongle goes to the IDE drive.
Set the jumper to Master or Slave, whatever the manual says.

Damn. No reviews yet.

http://www.newegg.com/product/product.aspx?Item=N82E16812200787

Paul
 
Thanks for the fast response and all the detailed information. My intention
is to install a third optical drive onto a new "Asus P4P800 SE Pentium 4
2.4ghz HT" 1gig dual channel memory XP-SP3 system using this method. I do
have a Promise Ultra66 ATA/66 PCI card and a SIIG UltraATA 133 PCI card. In
googling for both, the specs state that ATAPI devices are supported. Whether
the opticals agree or not, I'll have to wait and see. On this new system, 2
HD's will SATA. The 1st and 2nd opticals will be IDE as master/slave on the
secondary IDE channel and the 3rd IDE optical as master on one of the two,
(I'm leaning toward the Promise card), PCI cards. The SIIG card has it's on
BIOS and searches for attached drives to configure during boot.

I realize much of the hardware I plan to use is outdated, but I see no good
reason not to make use of functioning components that still meet my needs.
Plus, this way is easier on my wallet.

My current setup has two internal opticals plus an external USB optical. The
USB drive is beginning to make noises of impending failure. Plus installing
a third internal optical will free up some desk space...not much but some.
I'll post back when this all comes together in the way I hope it will.

Another question: Would it be your opinion that the early SATA motherboards
would have difficulty with the data transfer rates of todays SATA HD's?

Again, thanks for your time and assistance.

Curt.
 
Curt said:
Thanks for the fast response and all the detailed information. My intention
is to install a third optical drive onto a new "Asus P4P800 SE Pentium 4
2.4ghz HT" 1gig dual channel memory XP-SP3 system using this method. I do
have a Promise Ultra66 ATA/66 PCI card and a SIIG UltraATA 133 PCI card. In
googling for both, the specs state that ATAPI devices are supported. Whether
the opticals agree or not, I'll have to wait and see. On this new system, 2
HD's will SATA. The 1st and 2nd opticals will be IDE as master/slave on the
secondary IDE channel and the 3rd IDE optical as master on one of the two,
(I'm leaning toward the Promise card), PCI cards. The SIIG card has it's on
BIOS and searches for attached drives to configure during boot.

I realize much of the hardware I plan to use is outdated, but I see no good
reason not to make use of functioning components that still meet my needs.
Plus, this way is easier on my wallet.

My current setup has two internal opticals plus an external USB optical. The
USB drive is beginning to make noises of impending failure. Plus installing
a third internal optical will free up some desk space...not much but some.
I'll post back when this all comes together in the way I hope it will.

Another question: Would it be your opinion that the early SATA motherboards
would have difficulty with the data transfer rates of todays SATA HD's?

Again, thanks for your time and assistance.

Curt.
Click to expand...

The connection between the Northbridge and Southbridge is sometimes a limit.
The SATA port itself can be a limit too.

Processor
|
865PE
|
| Hub bus (266MB/sec)
|
IDE --- ICH5R --- IDE
| |
SATA SATA
~125
max
each

The SATA on ICH5R is SATA I, running at 1.5Gbit/sec. I've done a
benchmark with my drive jumpered to SATA I (the Force150 jumper),
and it just starts to clip at around 125MB/sec. So 125MB/sec
seems to be the usable limit for a SATA I interface. The drive I
tested, was a cheap 500GB. So it is possible to "bump your head"
on the limitations of SATA I. But in my case, it barely makes
a difference.

http://img829.imageshack.us/img829/842/500gb3500418ascomposite.gif

The hub bus is the next limitation. It might be an issue, if you
had two drives in RAID 0. It really depends on what other traffic
has to travel over the hub bus.

One thing you can't be sure of (short of testing it), is whether
a device in the Southbridge, is bridged to PCI or bridged to the
hub directly. Many crappy designs in the past, would be bridged
to the PCI bus, which would cause an internal 133MB/sec bottleneck.
Even though a much faster hub bus was available.

The Intel hub, is one of the slower ones in the industry. Other
companies have used faster hubs. Even today, Intel has bumped
their "DMI" up a bit, so it's no longer the pathetic 266MB/sec
limit. But it still represents a limit. Especially considering
the stuff they hang off a modern Southbridge.

In rough numbers, my benchmark above shows I go from the 135MB/sec
max my drive will offer, down to the 125MB/sec the SATA I port will
allow. The drive is a SATA II, but it can be jumpered down to SATA I
rates. So you might lose 10MB/sec.

If you were doing RAID0, perhaps you wouldn't get the whole thing
there either.

Now, if you bought a SIL3112 card for the PCI bus, then you're clipped
down to PCI bus limits, or perhaps 110MB/sec. The Southbridge
is still your best bet (not the Intel IDE though, just the SATA ports
are good). The PCI bus is eventually bridged to the hub
bus. If enough things go active at the same time, then the hub is
the limit. For example, if you transfer from a SATA port, to a PCI
card with SATA drive, then you'd have to consider whether they
conflict at the hub. I think the hub is simplex in this case, and
your 266MB/sec is in a single direction at a time. (A disk to disk
transfer, would alternate the burst direction.)

For an example

http://gathering.tweakers.net/forum/list_messages/384220/11

"2 x Maxtor diamondmax 9 SATA 120 gb Raid0 op de Promise 20378 onboard"

versus

"2 x Maxtor diamondmax 9 SATA 160 gb Raid0 op de ICH5R onboard"

and the latter setup using ICH5R wins by 10MB/sec. The Promise 20378 hits
the practical PCI limit (they could have squeezed a bit more out of it,
by raising the burst size, but then, that does nasty things to sound
on the motherboard).

There is a second example here. This is to show the ICH5R SATA ports
are bridged to the hub directly. Two 80MB/sec drives are put in RAID0
and give 160MB/sec on the ICH5R SATA ports. And the only way you could
get 160MB/sec, is if the SATA ports are connected to the hub. And that means
you should be able to get up to around 250MB/sec best case, perhaps
less depending on how much overhead the hub protocol has. I wasn't
able to find details on the hub protocol - it probably has an
address in the header of each burst, and the longer the burst is,
the more efficient the bus will be.

http://www.overclockers.com/forums/showthread.php?p=5282536

"single 250gb seagate perp on ICH5R" 80MB/sec
"2x 250gb seagate perps on ICH5R raid 0" 160MB/sec

I think you're in good shape, and will get to enjoy the
majority of your SATA performance.

Once you get it running, you can post an HDTune 2.55 result
on imageshack.us .

HTH,
Paul
 
Wow....thanks. That's a lot of good information to digest, but I'm feeling
better about my system make-over. I must say that I've never dabbled with
RAID...always felt it was a bit to complicated for me. I may delve into it
though just for the experience of learning something new.

You've been tremendously helpful,

Curt.
 
Back
Top