SATA On XP and Win2K

NTDEV
1

Hello everybody,

this is a real lay man’s question, but still if some one could throw some
light on the issue.

I recently assembled a P-IV machine for myself.

The configuration is…

P-IV, 3 GHz, HT
Intel 915 board (original), FSB 700 MHz
2 *512 MB DDR RAMs at 400 Mhz
1 SATA 80 GB Seagate HDD

Rest all standard components…

Well, I had initially put up Windows 2000 Pro, SP4 on the machine, even
after tweaking the box, and loading the original Intel supplied drivers, I
found the machine “not very astounding”. My older celeron or P-III gave me
equiv performance when I copied data etc.

Opon asking the hardware manufacturer, I was told to install Windows XP as
“win2K won’t deliver good performance on SATA”!!!

This statement I do not understand and the tech support guy just whined the
same tune again and again.

My question is, what has XP, or 2K got to do with copying data, when I am
installing the original drivers of the hardware, both should be equivalent.

Accepted, that XP might have some improved high level data cacheing algo
implemented but apart from that all should depend on the driver, or am I
wrong?

Can some one tell me as to why windows 2K should not perform well on SATA
drives???

Thanks in advance,

–

  • Developer
2

Developer wrote:

Hello everybody,

this is a real lay man’s question, but still if some one could throw
some light on the issue.

I recently assembled a P-IV machine for myself.

The configuration is…

P-IV, 3 GHz, HT
Intel 915 board (original), FSB 700 MHz
2 *512 MB DDR RAMs at 400 Mhz
1 SATA 80 GB Seagate HDD

Rest all standard components…

Well, I had initially put up Windows 2000 Pro, SP4 on the machine,
even after tweaking the box, and loading the original Intel supplied
drivers, I found the machine “not very astounding”. My older celeron
or P-III gave me equiv performance when I copied data etc.

There are too many variables here. Copied what? From what to what?
How do you know the performance is “equiv”? Just your gut feel? Or do
you actually have some megabytes-per-second numbers?

Serial ATA is not an automatic performance windfall. Is it possible
your old system had a “hot” hard disk? A great IDE disk with a good
cache will perform just about the same as an average Serial ATA disk

Copying is not a good measure of system performance. Your bottleneck is
the disk, so you get no benefit at all from your faster processor.

Opon asking the hardware manufacturer, I was told to install Windows
XP as “win2K won’t deliver good performance on SATA”!!!

This statement I do not understand and the tech support guy just
whined the same tune again and again.

My question is, what has XP, or 2K got to do with copying data, when I
am installing the original drivers of the hardware, both should be
equivalent.

Accepted, that XP might have some improved high level data cacheing
algo implemented but apart from that all should depend on the driver,
or am I wrong?

It isn’t that easy. When you access a disk, you aren’t using just one
“driver”. Instead, your requests are passing through a driver “stack”.
The disk driver is in the middle of that stack, with (possibly) a bus
driver below, and file system drivers and filters above. The rest of
the stack is part of the operating system.

Can some one tell me as to why windows 2K should not perform well on
SATA drives???

Note that you are still working from hearsay here. The only way you
could know whether the operating system is part of the problem is to do
the upgrade and redo your benchmarks.

–
Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.

3

Who’s ATA HBA drivers are you using? The dog-slow performance sounds a
lot like PIO-configured MS drivers are in action here. Unless my memory
is broken (again!) when the Win2k in-the-box ATA driver is running then
the disk is possibly running in PIO mode, and for SATA even though data
is always transferring by DMA the overhead for the whole process will
bring your machine to it’s knees. Verify this by looking at the disk
controller settings in the device manager and make sure the channels are
operating DMA and not PIO. Good luck, MKE.

-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Tim Roberts
Sent: Monday, November 21, 2005 10:05 AM
To: Windows System Software Devs Interest List
Subject: Re: [ntdev] SATA On XP and Win2K

Developer wrote:

Hello everybody,

this is a real lay man’s question, but still if some one could throw
some light on the issue.

I recently assembled a P-IV machine for myself.

The configuration is…

P-IV, 3 GHz, HT
Intel 915 board (original), FSB 700 MHz
2 *512 MB DDR RAMs at 400 Mhz
1 SATA 80 GB Seagate HDD

Rest all standard components…

Well, I had initially put up Windows 2000 Pro, SP4 on the machine,
even after tweaking the box, and loading the original Intel supplied
drivers, I found the machine “not very astounding”. My older celeron
or P-III gave me equiv performance when I copied data etc.

There are too many variables here. Copied what? From what to what?
How do you know the performance is “equiv”? Just your gut feel? Or do
you actually have some megabytes-per-second numbers?

Serial ATA is not an automatic performance windfall. Is it possible
your old system had a “hot” hard disk? A great IDE disk with a good
cache will perform just about the same as an average Serial ATA disk

Copying is not a good measure of system performance. Your bottleneck is

the disk, so you get no benefit at all from your faster processor.

Opon asking the hardware manufacturer, I was told to install Windows
XP as “win2K won’t deliver good performance on SATA”!!!

This statement I do not understand and the tech support guy just
whined the same tune again and again.

My question is, what has XP, or 2K got to do with copying data, when I

am installing the original drivers of the hardware, both should be
equivalent.

Accepted, that XP might have some improved high level data cacheing
algo implemented but apart from that all should depend on the driver,
or am I wrong?

It isn’t that easy. When you access a disk, you aren’t using just one
“driver”. Instead, your requests are passing through a driver “stack”.
The disk driver is in the middle of that stack, with (possibly) a bus
driver below, and file system drivers and filters above. The rest of
the stack is part of the operating system.

Can some one tell me as to why windows 2K should not perform well on
SATA drives???

Note that you are still working from hearsay here. The only way you
could know whether the operating system is part of the problem is to do
the upgrade and redo your benchmarks.

–
Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.

Questions? First check the Kernel Driver FAQ at
http://www.osronline.com/article.cfm?id=256

You are currently subscribed to ntdev as: xxxxx@intel.com
To unsubscribe send a blank email to xxxxx@lists.osr.com

4

Tim, Michael,

Thanks for the replies, I will try and explain the grey parts again.

There are too many variables here. Copied what? From what to what?

I did a very simple test. I copied the same amount of data (2.72 GB) from a
partition of the SATA disk to another partition. Keeping everythiung
constant during the series of tests. I defragmented the drives before
starting.

How do you know the performance is “equiv”? Just your gut feel? Or do
you actually have some megabytes-per-second numbers?

Well, I did clock the time taken. Ofcourse I didn’t have h2bench of Iometer
to help me, it was just for a basic home based test, so a stop watch was all
I had. I could ofcourse find out the numbers by doing the calculations.

Serial ATA is not an automatic performance windfall. Is it possible
your old system had a “hot” hard disk? A great IDE disk with a good
cache will perform just about the same as an average Serial ATA disk

No it didn’t. It was one of samsungs low end drives that ship in India which
you can acquire for a few dollars. 7200 RPM if I am not mistaken, and more
over the older machine was a celeron ( I also tried a P-III), celeron having
a max FSB speed of 100 Mhz. Thus this point is ruled out.

Copying is not a good measure of system performance. Your bottleneck is
the disk, so you get no benefit at all from your faster processor.

Yes I know, but I have seen P-IVs working a ziupping fast speeds before
WITHOUT any sort of tweaking at all. I am surprised that the same copy of
the Windows XP CD ( I have licences, please don’t start away in that
direction )is not getting me the same results.

It isn’t that easy. When you access a disk, you aren’t using just one
“driver”. Instead, your requests are passing through a driver “stack”.
>The disk driver is in the middle of that stack, with (possibly) a bus
driver below, and file system drivers and filters above. The rest of
the stack is part of the operating system.

Tim, I am into storage driver development myself, though a newbie still, I
have developed 2 disk filter drivers, so I guess I you can be a little more
technical with me. What I don’t know here, is what all OS drivers the Intel
installer package replaces. The CD, has three sections, video, audio (these
two are not of any concern) an the third are th chipset drivers. I don’t
know if they replace the entire stack, or add a special driver for the SATA
controller and I/O chipsets.

My laymans understanding goes like this…if I install the drivers that come
with a particular device, I stand better chances of getting better service,
as they are custom written.

Note that you are still working from hearsay here. The only way you
could know whether the operating system is part of the problem is to do
the upgrade and redo your benchmarks.

I did, I started off with windows 2K pro basic installation without intel
915 drivers. Did the copy operation noted the time. Now I installed the
drives, and did the test again. Not much improvements.

Then I installed Windows 2K3, without drivers did the test there was a
significant difference in time around 30 seconds gain on the same data, same
copy operation. then I installed the drivers on this OS, and ran the test,
this time there was a gain of 35 seconds on the 2K times.

So can I conclude that the OS caching algos play a role here. Oh yes, I more
thing, the CD just had one installer MSI, I don’t know whether it had
custom drivers for each versoin of the OS or not.

Who’s ATA HBA drivers are you using?
Michael, I guess I am using the Microsoft supplied ones, unless your
installer does away with it. I have to do a recheck on this one.

The dog-slow performance sounds a
lot like PIO-configured MS drivers are in action here.

NO IT IS NOT. Yes, I am using caps, as I AM ABSOLUTELY sure on this. I
myself configured it to use DMA.

Unless my memory
is broken (again!) when the Win2k in-the-box ATA driver is running then
the disk is possibly running in PIO mode, and for SATA even though data
is always transferring by DMA the overhead for the whole process will
bring your machine to it’s knees.

Correct.

Verify this by looking at the disk
controller settings in the device manager and make sure the channels are
operating DMA and not PIO. Good luck, MKE.

Yeah I did, and I did change it to use DMA and reboot, before all tests.

Any pointers???

amitrajit

5

Hello,

I’m running an SATA Mirror under W2KSp4 (and XP as well).
It’s a Via Chipset with Via Raid driver and it goes like hell.
There’s no dedicated MS SATA support in W2K, XP.
The internal protocols are still compatible with PATA,
but to achieve proper performance of course you need the right drivers.
Thus the disk stack should look like
“Chipset -> Chipset-whatever disk.sys -> Microsoft Disk.sys …”.
The chipset driver would replace atapi.sys (if we talk about PATA),
mine creates SCSI LUNs.
Just have a look at your (disk) stack with DevceTree.

Else

|---------±-------------------------------->
| | Developer |
| | |
| | Sent by: |
| | bounce-226476-16691@li|
| | sts.osr.com |
| | |
| | |
| | 11/22/2005 07:28 AM |
| | Please respond to |
| | “Windows System |
| | Software Devs Interest|
| | List” |
|---------±-------------------------------->
>-----------------------------------------------------------------------------------------------------------|
| |
| To: “Windows System Software Devs Interest List” |
| cc: |
| Subject: Re: [ntdev] SATA On XP and Win2K (Unsigned Mail) |
>-----------------------------------------------------------------------------------------------------------|

Tim, Michael,

Thanks for the replies, I will try and explain the grey parts again.

>There are too many variables here. Copied what? From what to what?

I did a very simple test. I copied the same amount of data (2.72 GB) from a
partition of the SATA disk to another partition. Keeping everythiung
constant during the series of tests. I defragmented the drives before
starting.

>How do you know the performance is “equiv”? Just your gut feel? Or do
>you actually have some megabytes-per-second numbers?

Well, I did clock the time taken. Ofcourse I didn’t have h2bench of Iometer
to help me, it was just for a basic home based test, so a stop watch was
all I had. I could ofcourse find out the numbers by doing the calculations.

>Serial ATA is not an automatic performance windfall. Is it possible
>your old system had a “hot” hard disk? A great IDE disk with a good
>cache will perform just about the same as an average Serial ATA disk

No it didn’t. It was one of samsungs low end drives that ship in India
which you can acquire for a few dollars. 7200 RPM if I am not mistaken, and
more over the older machine was a celeron ( I also tried a P-III), celeron
having a max FSB speed of 100 Mhz. Thus this point is ruled out.

>Copying is not a good measure of system performance. Your bottleneck is
>the disk, so you get no benefit at all from your faster processor.

Yes I know, but I have seen P-IVs working a ziupping fast speeds before
WITHOUT any sort of tweaking at all. I am surprised that the same copy of
the Windows XP CD ( I have licences, please don’t start away in that
direction )is not getting me the same results.

>It isn’t that easy. When you access a disk, you aren’t using just one
>“driver”. Instead, your requests are passing through a driver “stack”.
>The disk driver is in the middle of that stack, with (possibly) a bus
>driver below, and file system drivers and filters above. The rest of
>the stack is part of the operating system.

Tim, I am into storage driver development myself, though a newbie still, I
have developed 2 disk filter drivers, so I guess I you can be a little more
technical with me. What I don’t know here, is what all OS drivers the Intel
installer package replaces. The CD, has three sections, video, audio (these
two are not of any concern) an the third are th chipset drivers. I don’t
know if they replace the entire stack, or add a special driver for the SATA
controller and I/O chipsets.

My laymans understanding goes like this…if I install the drivers that
come with a particular device, I stand better chances of getting better
service, as they are custom written.

>Note that you are still working from hearsay here. The only way you
>could know whether the operating system is part of the problem is to do
>the upgrade and redo your benchmarks.

I did, I started off with windows 2K pro basic installation without intel
915 drivers. Did the copy operation noted the time. Now I installed the
drives, and did the test again. Not much improvements.

Then I installed Windows 2K3, without drivers did the test there was a
significant difference in time around 30 seconds gain on the same data,
same copy operation. then I installed the drivers on this OS, and ran the
test, this time there was a gain of 35 seconds on the 2K times.

So can I conclude that the OS caching algos play a role here. Oh yes, I
more thing, the CD just had one installer MSI, I don’t know whether it had
custom drivers for each versoin of the OS or not.

>Who’s ATA HBA drivers are you using?
Michael, I guess I am using the Microsoft supplied ones, unless your
installer does away with it. I have to do a recheck on this one.

> The dog-slow performance sounds a
>lot like PIO-configured MS drivers are in action here.

NO IT IS NOT. Yes, I am using caps, as I AM ABSOLUTELY sure on this. I
myself configured it to use DMA.

>Unless my memory
>is broken (again!) when the Win2k in-the-box ATA driver is running then
>the disk is possibly running in PIO mode, and for SATA even though data
>is always transferring by DMA the overhead for the whole process will
>bring your machine to it’s knees.

Correct.

>Verify this by looking at the disk
>controller settings in the device manager and make sure the channels are
>operating DMA and not PIO. Good luck, MKE.

Yeah I did, and I did change it to use DMA and reboot, before all tests.

Any pointers???

amitrajit

— Questions? First check the Kernel Driver FAQ at
http://www.osronline.com/article.cfm?id=256 You are currently subscribed to
ntdev as: unknown lmsubst tag argument: ‘’ To unsubscribe send a blank
email to xxxxx@lists.osr.com

6

yes the chipset driver does get replaced after I install the intel set of
drivers, but I still cannot find out why I cannot see any performance gains.

One reason could be that I am again using seagates low end SATA drives,
which don’t come with their own set of drivers, maybe if it had, and I got a
replaced disk.sys things would be better.

I will also check the seagate site for baracuda drivers.

for the the argument that there is not mcuh of a difference between 2K and
XP regarding inbuilt support, that is what I told the tech support guy also,
but he was very sure there is, may be seniors in this group may be able to
answer better.

amitr0

7

On my Asus P5P800 (the chipset is Intel 865PE) the XP’s OS-provided driver
works very fine with SATA.

Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation
xxxxx@storagecraft.com
http://www.storagecraft.com

----- Original Message -----
From: “Else Kluger”
To: “Windows System Software Devs Interest List”
Sent: Tuesday, November 22, 2005 1:39 PM
Subject: Re: [ntdev] SATA On XP and Win2K

> Hello,
>
> I’m running an SATA Mirror under W2KSp4 (and XP as well).
> It’s a Via Chipset with Via Raid driver and it goes like hell.
> There’s no dedicated MS SATA support in W2K, XP.
> The internal protocols are still compatible with PATA,
> but to achieve proper performance of course you need the right drivers.
> Thus the disk stack should look like
> “Chipset -> Chipset-whatever disk.sys -> Microsoft Disk.sys …”.
> The chipset driver would replace atapi.sys (if we talk about PATA),
> mine creates SCSI LUNs.
> Just have a look at your (disk) stack with DevceTree.
>
> Else
>
>
>
> |---------±-------------------------------->
> | | Developer |
> | | |
> | | Sent by: |
> | | bounce-226476-16691@li|
> | | sts.osr.com |
> | | |
> | | |
> | | 11/22/2005 07:28 AM |
> | | Please respond to |
> | | “Windows System |
> | | Software Devs Interest|
> | | List” |
> |---------±-------------------------------->
>
>------------------------------------------------------------------------------
-----------------------------|
> |
|
> | To: “Windows System Software Devs Interest List”
|
> | cc:
|
> | Subject: Re: [ntdev] SATA On XP and Win2K (Unsigned Mail)
|
>
>------------------------------------------------------------------------------
-----------------------------|
>
>
>
>
> Tim, Michael,
>
> Thanks for the replies, I will try and explain the grey parts again.
>
> >There are too many variables here. Copied what? From what to what?
>
> I did a very simple test. I copied the same amount of data (2.72 GB) from a
> partition of the SATA disk to another partition. Keeping everythiung
> constant during the series of tests. I defragmented the drives before
> starting.
>
> >How do you know the performance is “equiv”? Just your gut feel? Or do
> >you actually have some megabytes-per-second numbers?
>
> Well, I did clock the time taken. Ofcourse I didn’t have h2bench of Iometer
> to help me, it was just for a basic home based test, so a stop watch was
> all I had. I could ofcourse find out the numbers by doing the calculations.
>
> >Serial ATA is not an automatic performance windfall. Is it possible
> >your old system had a “hot” hard disk? A great IDE disk with a good
> >cache will perform just about the same as an average Serial ATA disk
>
> No it didn’t. It was one of samsungs low end drives that ship in India
> which you can acquire for a few dollars. 7200 RPM if I am not mistaken, and
> more over the older machine was a celeron ( I also tried a P-III), celeron
> having a max FSB speed of 100 Mhz. Thus this point is ruled out.
>
> >Copying is not a good measure of system performance. Your bottleneck is
> >the disk, so you get no benefit at all from your faster processor.
>
> Yes I know, but I have seen P-IVs working a ziupping fast speeds before
> WITHOUT any sort of tweaking at all. I am surprised that the same copy of
> the Windows XP CD ( I have licences, please don’t start away in that
> direction )is not getting me the same results.
>
>
> >It isn’t that easy. When you access a disk, you aren’t using just one
> >“driver”. Instead, your requests are passing through a driver “stack”.
> >The disk driver is in the middle of that stack, with (possibly) a bus
> >driver below, and file system drivers and filters above. The rest of
> >the stack is part of the operating system.
>
> Tim, I am into storage driver development myself, though a newbie still, I
> have developed 2 disk filter drivers, so I guess I you can be a little more
> technical with me. What I don’t know here, is what all OS drivers the Intel
> installer package replaces. The CD, has three sections, video, audio (these
> two are not of any concern) an the third are th chipset drivers. I don’t
> know if they replace the entire stack, or add a special driver for the SATA
> controller and I/O chipsets.
>
> My laymans understanding goes like this…if I install the drivers that
> come with a particular device, I stand better chances of getting better
> service, as they are custom written.
>
>
> >Note that you are still working from hearsay here. The only way you
> >could know whether the operating system is part of the problem is to do
> >the upgrade and redo your benchmarks.
>
> I did, I started off with windows 2K pro basic installation without intel
> 915 drivers. Did the copy operation noted the time. Now I installed the
> drives, and did the test again. Not much improvements.
>
> Then I installed Windows 2K3, without drivers did the test there was a
> significant difference in time around 30 seconds gain on the same data,
> same copy operation. then I installed the drivers on this OS, and ran the
> test, this time there was a gain of 35 seconds on the 2K times.
>
> So can I conclude that the OS caching algos play a role here. Oh yes, I
> more thing, the CD just had one installer MSI, I don’t know whether it had
> custom drivers for each versoin of the OS or not.
>
>
> >Who’s ATA HBA drivers are you using?
> Michael, I guess I am using the Microsoft supplied ones, unless your
> installer does away with it. I have to do a recheck on this one.
>
> > The dog-slow performance sounds a
> >lot like PIO-configured MS drivers are in action here.
>
> NO IT IS NOT. Yes, I am using caps, as I AM ABSOLUTELY sure on this. I
> myself configured it to use DMA.
>
> >Unless my memory
> >is broken (again!) when the Win2k in-the-box ATA driver is running then
> >the disk is possibly running in PIO mode, and for SATA even though data
> >is always transferring by DMA the overhead for the whole process will
> >bring your machine to it’s knees.
>
> Correct.
>
> >Verify this by looking at the disk
> >controller settings in the device manager and make sure the channels are
> >operating DMA and not PIO. Good luck, MKE.
>
> Yeah I did, and I did change it to use DMA and reboot, before all tests.
>
> Any pointers???
>
> amitrajit
>
>
>
>
> — Questions? First check the Kernel Driver FAQ at
> http://www.osronline.com/article.cfm?id=256 You are currently subscribed to
> ntdev as: unknown lmsubst tag argument: ‘’ To unsubscribe send a blank
> email to xxxxx@lists.osr.com
>
>
>
> —
> Questions? First check the Kernel Driver FAQ at
http://www.osronline.com/article.cfm?id=256
>
> You are currently subscribed to ntdev as: xxxxx@storagecraft.com
> To unsubscribe send a blank email to xxxxx@lists.osr.com

8

>One reason could be that I am again using seagates low end SATA drives,

which don’t come with their own set of drivers, maybe if it had, and I got a

Drivers are necessary for the controller only, not the disk.

Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation
xxxxx@storagecraft.com
http://www.storagecraft.com

9

Rotation speed and cache - low end disk drives are slow and
under-cached.

-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Maxim S. Shatskih
Sent: Tuesday, November 22, 2005 2:54 PM
To: Windows System Software Devs Interest List
Subject: Re: [ntdev] SATA On XP and Win2K

One reason could be that I am again using seagates low end SATA drives,
which don’t come with their own set of drivers, maybe if it had, and I
got a

Drivers are necessary for the controller only, not the disk.

Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation
xxxxx@storagecraft.com
http://www.storagecraft.com

Questions? First check the Kernel Driver FAQ at
http://www.osronline.com/article.cfm?id=256

You are currently subscribed to ntdev as: xxxxx@stratus.com
To unsubscribe send a blank email to xxxxx@lists.osr.com

10

I wouldn’t bother. There are no drivers for the Barracuda, or any other Seagate SATA/ATA disc. The only driver we produce for discs would be the USB storage devices and the drives we use for Smartcard/FDE.

–
Gary G. Little
“Developer” wrote in message news:xxxxx@ntdev…
yes the chipset driver does get replaced after I install the intel set of drivers, but I still cannot find out why I cannot see any performance gains.

One reason could be that I am again using seagates low end SATA drives, which don’t come with their own set of drivers, maybe if it had, and I got a replaced disk.sys things would be better.

I will also check the seagate site for baracuda drivers.

for the the argument that there is not mcuh of a difference between 2K and XP regarding inbuilt support, that is what I told the tech support guy also, but he was very sure there is, may be seniors in this group may be able to answer better.

amitr0