IDE RAID

[Ted Dawson]

Are there any 865 or newer boards with on-board IDE RAID, or is my only
choice SATA RAID?

 

[Paul]

Are there any 865 or newer boards with on-board IDE RAID, or is my only
choice SATA RAID?

For the 875/865 era, the Southbridge offers 4 x PATA drives and
2 x SATA drives, and for boards that have ICH5R (deluxe boards),
the SATA can be run in RAID mode. A number of these boards also
have a Promise 20378, which has 2 x SATA and one PATA cable. You
could RAID two IDE drives on a Promise 20378, but by sharing the
PATA cable, performance will be eroded.

On the 915/925 boards, I believe Intel has reduced the PATA interface
to a single cable on ICH6, while offering four SATA ports. This
is all part of Intel's plan to wean users off PATA. To compensate
for this deficiency, Asus offers the iteusa.com IT8212 PATA RAID,
which has two IDE connectors and allows four drives to be connected.
I haven't seen any benchmarks posted for this chip, so cannot say
whether it is a good chip for RAID or not. So, have a look in the
P5xxx part of the motherboard web page, as the specs for some of
them will include the IT8212. Some of the boards use PCI Express,
others AGP, some have DDR2 and others DDR, so you can search until
you get the right level of reuse of your current components.

You could always buy a separate controller card. The only one I
would avoid, is any product with a CMD0680 chip on it, as the
RAID on that one is "soft RAID". The down side of separate
controller cards, is getting them to work with other motherboard
disk subsystems. If you plan on booting from a separate controller
card, sometimes it means disabling other hardware on the motherboard
to get it to work.

As to whether any Southbridge has IDE RAID built in, I don't remember
any.

HTH,
Paul

 

[Ron Reaugh]

Are there any 865 or newer boards with on-board IDE RAID, or is my only
choice SATA RAID?

Why would you want anything but SATA RAID given that Raptors are SATA?

 

[Ron Reaugh]

-snip

You could always buy a separate controller card. The only one I
would avoid, is any product with a CMD0680 chip on it, as the
RAID on that one is "soft RAID".

All ATA RAID is software/firmware RAID except fancy/expensive cards like
3Ware and top Promise cards. I know of NO onmobo HW ATA RAID...it's all
software/firmware.


The down side of separate

 

[Leythos]

All ATA RAID is software/firmware RAID except fancy/expensive cards like
3Ware and top Promise cards. I know of NO onmobo HW ATA RAID...it's all
software/firmware.

ASUS PC-DL Deluxe has both Intel RAID 0/1 SATA controller and Promise
RAID 0/1 Controller (This is the better of the two).

 

[Ron Reaugh]

All ATA RAID is software/firmware RAID except fancy/expensive cards like
3Ware and top Promise cards. I know of NO onmobo HW ATA RAID...it's all
software/firmware.

ASUS PC-DL Deluxe has both Intel RAID 0/1 SATA controller and Promise
RAID 0/1 Controller (This is the better of the two).

BOTH are software/firmware RAID and not HW RAID however.

 

[Leythos]

All ATA RAID is software/firmware RAID except fancy/expensive cards like
3Ware and top Promise cards. I know of NO onmobo HW ATA RAID...it's all
software/firmware.

ASUS PC-DL Deluxe has both Intel RAID 0/1 SATA controller and Promise
RAID 0/1 Controller (This is the better of the two).

BOTH are software/firmware RAID and not HW RAID however.

The firmware handles the RAID completely on the Promise controller, the
software is just a driver, like the IDE, ATAPI, SCSI, etc... drivers you
need on other motherboards or that are included with the OS. Don't kid
yourself, the only thing that would make the RAID any better is if it
included a slot on the MB for cache.

 

[Ron Reaugh]

All ATA RAID is software/firmware RAID except fancy/expensive

cards
like

3Ware and top Promise cards. I know of NO onmobo HW ATA

RAID...it's
all

software/firmware.

ASUS PC-DL Deluxe has both Intel RAID 0/1 SATA controller and Promise
RAID 0/1 Controller (This is the better of the two).

BOTH are software/firmware RAID and not HW RAID however.

The firmware handles the RAID completely on the Promise controller,

Right, that firmware is x86 code contained in the mobo's BIOS just like the
Intel RAID firmware inside that same mobo BIOS.

the
software is just a driver,

Right, just like firmware is a driver...both are x86 code executed on the
host's x86 CPU.

That Promise firmware and Intel firmware are used ONLY during POST. After
that in an OS like XP ALL activity with both the Promise and Intel RAID are
handled by OS x86 drivers.

like the IDE, ATAPI, SCSI, etc... drivers you
need on other motherboards or that are included with the OS.
Right.

Don't kid
yourself, the only thing that would make the RAID any better is if it
included a slot on the MB for cache.

HUH?

Hardware RAID is completely different. By definition HW RAID offloads all
RAID processing to the RAID card like a 3Ware. All RAID 5 parity calcs are
done on the RAID card by its own resident RISC CPU.

In RAID 1(mirroring) the data gets written to each drive(two identical
redundant writes). In software/firmware RAID 1 like Promise/Intel that data
moves over the hosts I/O bus structure TWICE for the two writes and that's
done by TWO cycles through the host's x86 driver code. In HW RAID that
write data goes over the host's I/O bus structure just ONCE and the host's
x86 driver code deals with that data just once. The HW RAID card then does
the two writes to the two RAID 1 drives using its onboard RISC CPU.

In RAID 0(striping) the host's CPU driver does all the stripe cutting
calculations and then divies the data to be written/read to the individual
drives. In HW RAID all that is done onboard the RAID card by its RISC CPU.

HW RAID always makes the whole array look like a single HD to the PC from
POST time through the OS load and operation.

The following Promise cards are software/firmware RAID:
FastTrak TX2200
FastTrak TX4200
Inexpensive RAID cards like Highpoint and SIIG are also all
firmware/software RAID and all are >$100..

Really neat SATA HW RAID cards are:
http://www.3ware.com/products/serial_ata9000.asp
They cost $350 and up. They however offer little advantage for RAID 0 and
RAID 1 as that extra x86 processing is minimal for RAID 0 and RAID 1. RAID
5 is a whole different ball game and is where these 3Ware gadgets shine.
Also that is why there is little advantage in using the mobo's RAID for W2K3
or W2K as these OS's offer great RAID 1 support in the OS.

The Promise chip on the mobo is nothing more than a fancy ATA controller
chip with NO significant RAID functionality on it. All on mobo Promise RAID
is firmware/software in x86 code hosted by the host's x86 CPU.

 

[Leythos]

The Promise chip on the mobo is nothing more than a fancy ATA controller
chip with NO significant RAID functionality on it. All on mobo Promise RAID
is firmware/software in x86 code hosted by the host's x86 CPU.

I would be interested in seeing where you get this information from. In
reviewing the Promise RAID 0/1 controller on the motherboard of the ASUS
PC-DL Deluxe board, I've only seen that the "driver" is a stub that
allows the OS to recognise the controller (much like the SCSI RAID
Controllers that we use in HP or Compaq servers).

 

[Leythos]

I would be interested in seeing where you get this information from. In
reviewing the Promise RAID 0/1 controller on the motherboard of the ASUS
PC-DL Deluxe board, I've only seen that the "driver" is a stub that
allows the OS to recognise the controller (much like the SCSI RAID
Controllers that we use in HP or Compaq servers).

Still waiting on a reply for where you're getting the idea that ALL
motherboard based RAID is as you describe. As I've seen it, there are
several good RAID implementations of onboard SATA RAID and SCSI RAID
controllers and many bad implementations, but I've not found one that
requires the driver to make two writes, that's built into the chipset
firware (as it is in a off-board scsi card).

 

[Ron Reaugh]

I would be interested in seeing where you get this information from. In
reviewing the Promise RAID 0/1 controller on the motherboard of the ASUS
PC-DL Deluxe board, I've only seen that the "driver" is a stub

Nope. No clue where you see this as there is full driver support there in
two different flavors. One RAID and one NOT.

that
allows the OS to recognise the controller (much like the SCSI RAID
Controllers that we use in HP or Compaq servers).

Any OS install requires a F6 driver load just like any other RAID card that
the OS doesn't already know about.

All my assertions are obvious once one thinks about it. Look at the specs
for a real HW RAID like a 3Ware and notice the onboard uP.

Look at the SATA card:
http://www.promise.com/product/product_detail_eng.asp?productId=126&familyId=3#

Look at its nearly identical sibling RAID card:
http://www.promise.com/product/product_detail_eng.asp?productId=107&familyId=2#

Both use the same Promise SATA controller. The only significant difference
is the onboard firmware chip, which contains x86 code. One has RAID
functionality and the other doesn't.

Such is well known.

 

[Ron Reaugh]

Still waiting on a reply for where you're getting the idea that ALL
motherboard based RAID is as you describe.

It's well known for mobo ATA RAID. Ask anyone that knows about such.

As I've seen it, there are
several good RAID implementations of onboard SATA RAID and SCSI RAID
controllers and many bad implementations,

Some onmobo SCSI RAID maybe HW but that requires an add-on card.

but I've not found one that
requires the driver to make two writes, that's built into the chipset
firware

Nope, not for ATA.

 

[Leythos]

Nope. No clue where you see this as there is full driver support there in
two different flavors. One RAID and one NOT.


Any OS install requires a F6 driver load just like any other RAID card that
the OS doesn't already know about.

All my assertions are obvious once one thinks about it. Look at the specs
for a real HW RAID like a 3Ware and notice the onboard uP.

Look at the SATA card:
http://www.promise.com/product/product_detail_eng.asp?productId=126&familyId=3#

Look at its nearly identical sibling RAID card:
http://www.promise.com/product/product_detail_eng.asp?productId=107&familyId=2#

Both use the same Promise SATA controller. The only significant difference
is the onboard firmware chip, which contains x86 code. One has RAID
functionality and the other doesn't.

I'll contact promise - I'm interested to really know if the OS/Driver
has to make two writes in RAID-1 or if the firmware handles it on it's
own. Since I have a large number of PC-DL servers with Dual 250GB SATA
in RAID-1 mode it will be interesting to see their response.

 

[Tim]

Ron,

I suggest you read a lot more. Go to the intel site and ferret out the specs
on the ICH5R (82801ER chip). Read there that it states clearly that it is
Hardware Raid. Also note that the manifestation of difference between ICH5
and ICH5R is either a) if it is [perhaps] packed as a ICH5 only and so has
RAID turned off / omitted within the chip, or b) if RAID is turned on or off
by the bios at boot time - the chip will report itself as 82801EB if RAID is
off and 82801ER if RAID is on. IE ICH5R = ICH5 with RAID turned on.

The presence of RAID Firmware in the bios is no indication of
a) x86 code running in place of RAID functionality (IE soft raid) or
b) that the bios performs the RAID functionality, or
c) that all the code is x86 code - it could be a mix of x86 and whatever
code set the RAID controller uses internally for its own private firmware
that it receives on every boot from the bios.

You will see in the Intel documentation that the bios code provides specific
supporting funcitonality ie:
a) configuration and management of RAID volumes,
b) boot time access to the RAID drive, and
c) detection of RAID status in the event of failure.
The Intel documentation does not say it does anything else. IE it does not
state that it implements soft raid.

If you also read up about windows drivers you will also learn that bios
functionality is not used within Windows XP when the system is running. The
purpose of device drivers is to provide windows with software interfaces to
hardware devices that conform to a specific predefined model so that Windows
knows how to use the device correctly and automatically. The responsibility
of the device driver writer is to marry the specific device(s) to the
interface in conformance with the chosen and stated standard (IE you can
take a device that controls SATA drives and implement it as SCSI if you
wish). You are bound to have noticed that when a SATA RAID controller is
configured as RAID the device is present as a SCSI device. This is because
the native SCSI functionality is a more appropriate device model for RAID
and also that the underlying IDE and SATA interfaces are no longer visible
(see the Intel programmers reference for ICH5R for more details, or Windows
Device Manager). Any functionality provided in the bios (EG boot time
support) is minimal functionality - single threaded reading / writing to the
device, boot time disc access is not a multithreaded high performance
environment. Bios support is designed for pre-boot execution (EG checking
RAID integrity) or boot: DOS or DOS equivalent access modes (IE boot, and EG
Nortons Ghost).

Having a hardware vendor implement soft raid is unheard of here. All reviews
of such hardware would be condeming as it would be a poor perfoming,
deceitful product to claim RAID for a device when the device does not
implement it. In this country, any vendor of such a product would be legally
liable for such deceit. If you want soft raid then use the in-built Windows
soft raid functionality on *any* stock IDE or SCSI drive - no special
controller is needed.

The OP's original reference was to Intel 865 based motherboards. Making
generalisations about Intel 865 or 875 based hardware when someone somewhere
*may* have done what you claim on totally different hardware is misleading
at the least. Your references to the Promise hardware indicate your
confusion. On one hand you link to a SATA controller and the other a RAID
controller. What was your point? Which chips contains the on-board
microprocessor? Do you know? "The only difference is the onbaord firmware
chip". What is the make and model number of this chip? I suspect you are
attributing microprocessor functionality to a Flash RAM chip. Onboard or
Onchip controller that needs firmware can be configured to get their
firmware out of the BIOS chip (or the bios supplies it to them somehow).

Please, get your facts straight.

- Tim

You can find Intel at www.intel.com
for information on flash memory chips, see www.atmel.com
for information on windows device driver model etc. see
http://www.microsoft.com/whdc/devtools/ddk/default.mspx

 

[Ron Reaugh]

I'll contact promise - I'm interested to really know if the OS/Driver
has to make two writes in RAID-1

It does.

or if the firmware handles it on it's
own.

The firmware is hosted on the x86 host CPU so if x86 hosted firmware does it
then it PROVES my point.

Have you ever investigated the speed of code running from a ROM which is
where firmware is held vs code in RAM were device drivers are held? Ever
hear of the concept of shadowing the video BIOS or other BIOS segments in
RAM and what that's all about?

Since I have a large number of PC-DL servers with Dual 250GB SATA
in RAID-1 mode it will be interesting to see their response.

Why? You are chasing a ghost.

The percentage of writes is usually small so that extra host overhead in
RAID 1 is not very important. To avoid it one must use full hardware RAID
like 3Ware. To avoid it the RAID controller must have onboard buffering and
there is none for the Promise nor other onmobo ATA RAID solutions. All
are firmware/software RAID and all do two host I/O bus writes for RAID 1
writes.

I use P4C800-E Dlx mobos on some SBS2003 sites of mine. I like W2K3's
intrinsic SW RAID 1. I don't even bother with the onmobo RAID but just use
the OS's intrinsic RAID 1. There is NO performance disadvantage compared to
onmobo SW/firmware RAID 1. Little performance would be gained by using a
$350 3Ware card for true HW RAID 1 here but RAID 5 with it's parity calcs is
a horse of a different color. Then I do 3Ware or SCSI HW RAID 5 with
Fujitsu MAS3735s.

 

[Ron Reaugh]

Ron,

I suggest you read a lot more. Go to the intel site and ferret out the specs
on the ICH5R (82801ER chip). Read there that it states clearly that it is
Hardware Raid.

Nope, it states NO such thing.

See the ICH5R datasheet (8 MB):
ftp://download.intel.com/design/chipsets/datashts/25251601.pdf
5.17.3
The only special RAID functionality regards RAID 0 and not RAID 1.

Goto the Intel site and select advanced search and search on the exact
phrase 'hardware RAID'. Study the results. That will give you a good idea
of what HW RAID is. You will NOT find the ICH5R in that search.

Also note that the manifestation of difference between ICH5
and ICH5R is either a) if it is [perhaps] packed as a ICH5 only and so has
RAID turned off / omitted within the chip, or b) if RAID is turned on or off
by the bios at boot time - the chip will report itself as 82801EB if RAID is
off and 82801ER if RAID is on. IE ICH5R = ICH5 with RAID turned on.

Just a subtle difference not related to the difference between SW/FM and
hardware RAID and only related to RAID 0 processing.

The presence of RAID Firmware in the bios is no indication of
a) x86 code running in place of RAID functionality (IE soft raid) or

Quite true in the most general sense however in this case that firmware is
x86 code and it is what handles ALL the RAID functionality until the OS
device driver takes over.

b) that the bios performs the RAID functionality, or

True, until the OS's DD takes over.

c) that all the code is x86 code - it could be a mix of x86 and whatever
code set the RAID controller uses internally for its own private firmware
that it receives on every boot from the bios.

Incredible nonsense.

You will see in the Intel documentation that the bios code provides specific
supporting funcitonality ie:
a) configuration and management of RAID volumes,
b) boot time access to the RAID drive, and
c) detection of RAID status in the event of failure.

It does all that plus all the other RAID functionality.

The Intel documentation does not say it does anything else. IE it does not
state that it implements soft raid.

Nevertheless it does.

If you also read up about windows drivers you will also learn that bios
functionality is not used within Windows XP when the system is running.

DUH! Read my other post where I talk about that in some detail.

The
purpose of device drivers is to provide windows with software interfaces to
hardware devices that conform to a specific predefined model so that Windows
knows how to use the device correctly and automatically.

The 2nd purpose is to locate the code in RAM instead of very slow ROM where
firmware lives.

The responsibility
of the device driver writer is to marry the specific device(s) to the
interface in conformance with the chosen and stated standard (IE you can
take a device that controls SATA drives and implement it as SCSI if you
wish).

Ever write a DD? I'm a DD developer and have written disk device drivers.

You are bound to have noticed that when a SATA RAID controller is
configured as RAID the device is present as a SCSI device.

That's purely a function of the device driver.

This is because
the native SCSI functionality is a more appropriate device model for RAID
and also that the underlying IDE and SATA interfaces are no longer visible
(see the Intel programmers reference for ICH5R for more details, or Windows
Device Manager). Any functionality provided in the bios (EG boot time
support) is minimal functionality - single threaded reading / writing to the
device, boot time disc access is not a multithreaded high performance
environment. Bios support is designed for pre-boot execution (EG checking
RAID integrity) or boot: DOS or DOS equivalent access modes (IE boot, and EG
Nortons Ghost).

Go back and read up in more detail such that you can understand what's
really said in such places.

Having a hardware vendor implement soft raid is unheard of here.

HUH, just clueless!

Virtually all ATA RAID was SW/FM and not HW until 3Ware started doin HW
RAID.

All reviews
of such hardware would be condeming as it would be a poor perfoming,

Pure nonsense. SW/FW RAID performs about as well as HW RAID for RAID 0 and
RAID 1.

deceitful product to claim RAID for a device when the device does not
implement it.

Just NO!

In this country, any vendor of such a product would be legally
liable for such deceit.
Clueless.

If you want soft raid then use the in-built Windows
soft raid functionality on *any* stock IDE or SCSI drive - no special
controller is needed.

Exactly, the difference is that one can't boot from such a RAID 0 array
without that RAID firmware but RAID 1 makes most these cards of little value
IF the OS supports RAID 1.

The OP's original reference was to Intel 865 based motherboards. Making
generalisations about Intel 865 or 875 based hardware when someone

somewhere > *may* have done what you claim on totally different hardware is
misleading

at the least. Your references to the Promise hardware indicate your
confusion. On one hand you link to a SATA controller and the other a RAID
controller. What was your point? Which chips contains the on-board
microprocessor? Do you know? "The only difference is the onbaord firmware
chip". What is the make and model number of this chip? I suspect you are
attributing microprocessor functionality to a Flash RAM chip. Onboard or
Onchip controller that needs firmware can be configured to get their
firmware out of the BIOS chip (or the bios supplies it to them somehow).

Your background on these issues is very shallow. Go back to school.

 

[Leythos]

Then I do 3Ware or SCSI HW RAID 5 with
Fujitsu MAS3735s.

Have you looked at the Promise SX-6000 ATA RAID-0/1/5 controller card.

 

[Ron Reaugh]

Have you looked at the Promise SX-6000 ATA RAID-0/1/5 controller card.

Don't know. Some of those Promise cards are semi-HW RAID. Where the 'semi'
is mostly relevant to RAID 5 because of a HW XOR engine.

 

[Tim]

From the intel document "Why_Raid.pdf":

"Industries first desktop RAID controller integrated into the chipset"
"RAID BIOS ROM: Integrated system BIOS, enables pre-OS RAID creation,
mapping, and deletion"

From 25272901.pdf:
"... includes an integrated RAID controller that utilizes the dual Serial
ATA ports for high-performance RAID Level 0.... By integrating the Intel
RAID controller into the I/O controller hub there are no PCI bandwidth
limitations..."

From 25251716.pdf, page 17 (addendum to 25251601.pdf):
"RAID Level 1 is supported in addition to RAID Level 0."

From 25267102.pdf page 28 (somewhat offtopic):
"... does not support surprise removals. ... a device can be powered down by
system software... allowing removal and insertion of a new device". Which is
as I said a while back: No SATA Hot Swap - the implementation is incomplete.

82801 *is* hardware. If the RAID controller is "integrated into" the chip,
then are you saying that Intel is deliberately misleading everyone into
thinking this is not actually a RAID controller at all, that it is some
lesser device that only can do individual IO's to individual volumes?

Now, Ron, how about some evidence - real evidence? You have none so far and
your logic is horribly flawed. You have this horrible habit of saying things
like "No", or "Clueless" and so on. You never substantiate any of your
claims, never even offer any reasoning or documentation for your claims. All
you have offered in all the posts I have read from you are the two
hyperlinks below that point to two different controllers that are
"different". How about it? Show us some documentation that proves beyond all
doubt that your own assertions are correct when they run contrary to Intel's
own documentation? How about proving that the ICH5 firmware is *not* held in
the bios? How about proving that the Windows OS actually isses 2 writes from
the OS to the controller to achieve a write onto 2 physical volumes with
Intel's RAID 1? That is what you are claiming. I don't appreciate being
called clueless, and don't appreciate being run down by someone that claims
to have ddk knowledge when there is nothing evident in their thinking at
all. I would like to know the answers, but at the moment your answers are
useless to everyone as they convey no knowledge or reference to any
knowledge whatsoever. Consequently your posts are more akin to flames and
that is how I have treated them to date.

"> Incredible nonsense."? What are microcode updates then? Answer:
proprietary code that is loaded into the Intel CPU at boot time. So why such
nonsense? If the bios can update the microcode at boot time, what is
stopping a bios writer from supplying firmware to an on board manufacturer
designed-in controller such as the ICH5R? Why does this have to be x86 code?
Why does this have to run out of the flash chip (that would be stupid,
memory is cheap remember)? Why does the bios have to contain exclusively x86
code when clearly microcode can be anything that the proprietor of the CPU
decides (it could be x86 code...).

You would do well to read up on the documentation on some of the flash chips
over at atmel. You will find that they have updated quite a bit over the
years and now include support within the chip for 'partitions' so that
different areas of the memory can be erased / updated separately and safely,
can serve different purposes, so that boot blocks can be safe guarded, and
lastly to integrate with the motherboard chips more. IE to serve as the
repository for not just bios but also on board firmware for chips such as
RAID controllers. When you have noted some of these product numbers then go
back and have a look at the difference getweent the two promise cards
previously - the chips may not be atmel or the same as those for your
motherboard bios, but you will find they have similar capabilities and serve
the same purpose. Try a google search on the chips to find out what they
are...

- Tim

Ron,

I suggest you read a lot more. Go to the intel site and ferret out the specs
on the ICH5R (82801ER chip). Read there that it states clearly that it is
Hardware Raid.

Nope, it states NO such thing.

See the ICH5R datasheet (8 MB):
ftp://download.intel.com/design/chipsets/datashts/25251601.pdf
5.17.3
The only special RAID functionality regards RAID 0 and not RAID 1.

Goto the Intel site and select advanced search and search on the exact
phrase 'hardware RAID'. Study the results. That will give you a good
idea
of what HW RAID is. You will NOT find the ICH5R in that search.

Also note that the manifestation of difference between ICH5
and ICH5R is either a) if it is [perhaps] packed as a ICH5 only and so
has
RAID turned off / omitted within the chip, or b) if RAID is turned on or off
by the bios at boot time - the chip will report itself as 82801EB if RAID is
off and 82801ER if RAID is on. IE ICH5R = ICH5 with RAID turned on.

Just a subtle difference not related to the difference between SW/FM and
hardware RAID and only related to RAID 0 processing.

The presence of RAID Firmware in the bios is no indication of
a) x86 code running in place of RAID functionality (IE soft raid) or

Quite true in the most general sense however in this case that firmware is
x86 code and it is what handles ALL the RAID functionality until the OS
device driver takes over.

b) that the bios performs the RAID functionality, or

True, until the OS's DD takes over.

c) that all the code is x86 code - it could be a mix of x86 and whatever
code set the RAID controller uses internally for its own private firmware
that it receives on every boot from the bios.

Incredible nonsense.

You will see in the Intel documentation that the bios code provides specific
supporting funcitonality ie:
a) configuration and management of RAID volumes,
b) boot time access to the RAID drive, and
c) detection of RAID status in the event of failure.

It does all that plus all the other RAID functionality.

The Intel documentation does not say it does anything else. IE it does
not
state that it implements soft raid.

Nevertheless it does.

If you also read up about windows drivers you will also learn that bios
functionality is not used within Windows XP when the system is running.

DUH! Read my other post where I talk about that in some detail.

The
purpose of device drivers is to provide windows with software interfaces to
hardware devices that conform to a specific predefined model so that Windows
knows how to use the device correctly and automatically.

The 2nd purpose is to locate the code in RAM instead of very slow ROM
where
firmware lives.

The responsibility
of the device driver writer is to marry the specific device(s) to the
interface in conformance with the chosen and stated standard (IE you can
take a device that controls SATA drives and implement it as SCSI if you
wish).

Ever write a DD? I'm a DD developer and have written disk device drivers.

You are bound to have noticed that when a SATA RAID controller is
configured as RAID the device is present as a SCSI device.

That's purely a function of the device driver.

This is because
the native SCSI functionality is a more appropriate device model for RAID
and also that the underlying IDE and SATA interfaces are no longer
visible
(see the Intel programmers reference for ICH5R for more details, or Windows
Device Manager). Any functionality provided in the bios (EG boot time
support) is minimal functionality - single threaded reading / writing to the
device, boot time disc access is not a multithreaded high performance
environment. Bios support is designed for pre-boot execution (EG checking
RAID integrity) or boot: DOS or DOS equivalent access modes (IE boot, and EG
Nortons Ghost).

Go back and read up in more detail such that you can understand what's
really said in such places.

Having a hardware vendor implement soft raid is unheard of here.

HUH, just clueless!

Virtually all ATA RAID was SW/FM and not HW until 3Ware started doin HW
RAID.

All reviews
of such hardware would be condeming as it would be a poor perfoming,

Pure nonsense. SW/FW RAID performs about as well as HW RAID for RAID 0
and
RAID 1.

deceitful product to claim RAID for a device when the device does not
implement it.

Just NO!

In this country, any vendor of such a product would be legally
liable for such deceit.
Clueless.

If you want soft raid then use the in-built Windows
soft raid functionality on *any* stock IDE or SCSI drive - no special
controller is needed.

Exactly, the difference is that one can't boot from such a RAID 0 array
without that RAID firmware but RAID 1 makes most these cards of little
value
IF the OS supports RAID 1.

The OP's original reference was to Intel 865 based motherboards. Making
generalisations about Intel 865 or 875 based hardware when someone

somewhere > *may* have done what you claim on totally different hardware
is
misleading

at the least. Your references to the Promise hardware indicate your
confusion. On one hand you link to a SATA controller and the other a RAID
controller. What was your point? Which chips contains the on-board
microprocessor? Do you know? "The only difference is the onbaord firmware
chip". What is the make and model number of this chip? I suspect you are
attributing microprocessor functionality to a Flash RAM chip. Onboard or
Onchip controller that needs firmware can be configured to get their
firmware out of the BIOS chip (or the bios supplies it to them somehow).

Your background on these issues is very shallow. Go back to school.

Please, get your facts straight.

- Tim

You can find Intel at www.intel.com
for information on flash memory chips, see www.atmel.com
for information on windows device driver model etc. see
http://www.microsoft.com/whdc/devtools/ddk/default.mspx

 

[Ron Reaugh]

From the intel document "Why_Raid.pdf":

"Industries first desktop RAID controller integrated into the chipset"
"RAID BIOS ROM: Integrated system BIOS, enables pre-OS RAID creation,
mapping, and deletion"

From 25272901.pdf:
"... includes an integrated RAID controller that utilizes the dual Serial
ATA ports for high-performance RAID Level 0.... By integrating the Intel
RAID controller into the I/O controller hub there are no PCI bandwidth
limitations..."

From 25251716.pdf, page 17 (addendum to 25251601.pdf):
"RAID Level 1 is supported in addition to RAID Level 0."

From 25267102.pdf page 28 (somewhat offtopic):
"... does not support surprise removals. ... a device can be powered down by
system software... allowing removal and insertion of a new device". Which is
as I said a while back: No SATA Hot Swap - the implementation is incomplete.

82801 *is* hardware. If the RAID controller is "integrated into" the chip,
then are you saying that Intel is deliberately misleading everyone into
thinking this is not actually a RAID controller at all, that it is some
lesser device that only can do individual IO's to individual volumes?

Now, Ron, how about some evidence - real evidence?

Clueless. You simply posted from the spec sheet similar to what I already
cited. You provided NOTHING new beyond what I already showed.

Nothing in that spec sheet says anything about HW RAID. The ICH5R does NOT
do HW RAID. There is no uP and NO buffer. Such are requirements of HW
RAID.

How about proving that the ICH5 firmware is *not* held in
the bios?

Wacko...I was the one that said that the ICH5R firmware WAS in the BIOS.

How about proving that the Windows OS actually isses 2 writes from
the OS to the controller to achieve a write onto 2 physical volumes with
Intel's RAID 1?

It's intuitively obvious. What does the processing if there's a device
error or retry on one drive and not the other? Where does that
functionality live? So when the data must be resent to one drive and not
the other in RAID 1 then what handles that processing. It's all x86 code.
It's all done over the x86 I/O bus. Get a clue. Read up on what real HW
RAID looks like from www.3ware.com

That is what you are claiming. I don't appreciate being
called clueless,

Get use to it when you go out of your depth up against someone who actually
knows something.

and don't appreciate being run down by someone that claims
to have ddk knowledge when there is nothing evident in their thinking at
all. I would like to know the answers, but at the moment your answers are
useless to everyone as they convey no knowledge or reference to any
knowledge whatsoever. Consequently your posts are more akin to flames and
that is how I have treated them to date.

Wacko speak....