SATA vs. IDE

On Mon, 4 Aug 2003 12:51:26 -0700
Bill wrote:

In article ,
says...
snip

You've snipped who said what, but I think it was me that said...


Scsi doesn't have DMA. AS far as I know, neither does SATA.

Hmmm....SCSI has had DMA transfers for at least the last 15 years.
My all SCSI systems use DMA and this is set from the SCSI BIOS
(no doubt Windows also has the DMA settings switch and, not being a
Win user I would have to experiment to see if Win ignored the BIOS
settings or not. No doubt another person could advise better on
this.

I don't know whether Scsi handles memory transfers through DMA or
some
other mechanism, so I'll take your word for it. I'm running SATA on a

Asus A7N8X Deluxe Ver. 1.06 MB and a Promise Fasttrak 100 Raid
controller in Raid 0+1 and neither show a DMA box under Device
Manager; both show as Scsi devices Win2000 .

It would have been more correct of me to say that, afaik, Windows
up to 2000, never ran XP, has no mechanism for enabling/disabling
DMA tranfers that I've ever found in Scsi devices, under Device
Manager.

That's because with SCSI host adapters you don't enable or disable DMA
any more than you would do it with a network card or video board. With
IDE drives, you enable or disable because the driver is
one-size-fits-all and some drives have DMA while others don't. With
SCSI DMA is on the host adapter, not the drive, and is handled by a
device driver unique to that host adapter rather than a generic driver,
hence no need to enable or disable.

Whether SCSI has or does not have DMA then depends on the particular
host adapter--a cheap one won't, a good one will.

I do have a box with a Tekram DC-395U Scsi controller but it's in
pieces getting a new Abit NF7-S MB installed, and used to have a
Fireport 40. I don't remember seeing DMA under either of the the bios

areas for these controllers. Could it be that DMA is only on faster
Ultra80/Ultra160/Ultra320 controllers? Or goes by a different name?

No, it's something that you don't futz with--there's no setting for it,
it's built into the hardware and drivers. According to Tekram the
DC-395U is a "Full 32-bit PCI DMA bus master", which means that it uses
DMA.

Bill



Regards,

Kevin


--
--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net)

J.Clarke wrote:
That's because with SCSI host adapters you don't enable or disable DMA
any more than you would do it with a network card or video board.
With IDE drives, you enable or disable because the driver is
one-size-fits-all and some drives have DMA while others don't. With
SCSI DMA is on the host adapter, not the drive, and is handled by a
device driver unique to that host adapter rather than a generic
driver, hence no need to enable or disable.

Good explanation John,

Ben
--
I'm not just a number. To many, I'm known as a string...

On Thu, 17 Jul 2003 07:59:06 +0100, "Ben Pope"
wrote:

Ransack The Elder wrote:
"Ben Pope" wrote in message
...

You're missing the point. Before you can say This SATA drive is
worse than that PATA drive, you need to be taking into account the
actual drive, not just its interface, not just the manufacturer.

1) I have the fastest 120 gig SATA drive on the market..today, anyway.

OK, but it's still not native SATA.

2) Everything I have read and all the charts I have seen say that
SATA is faster than IDE.

Read more.

3) My IDE drives are faster than my SATA drive.

OK.

So, does that mean:
A) My IDE drive are that damn good.

Yes. BUt because it's a fast drive mechanism and good algorithms, not
because it's PATA.

B) SATA is marketing hype that offers nothing other than a smaller
cable.

No! It offers many useful things such as hot-plugging. I don't know how
many times that would have been useful for me, at least,

Now, what you have is NOT a native SATA drive, you have a PATA drive
with a PATA to SATA convertor chip on it - this may explain why you
have experienced some degradation in peformance over the IBM drive.
Additionally, whereas your ATA interface is built directly into your
chipset, your the data flowing through your SATA interface has to
also traverse the PCI bus (limited in TOTAL to 133Megs/s).

Ummm..okay...so let's say my SATA controller is NOT built onto my
chipset. I would need a PCI SATA card which...bada bing...traverses
the PCI bus. So what's the difference?????

Re-read what I said.

Do you still think you are performing a fair comparison of SATA and
PATA?

Yep. I have the same setup as those two websites you sent.

Ok, then do you think that THEY were performing a fair comparison?

Before suggesting that SATA is worse than PATA we will have to wait
for chipsets and drives to natively support SATA, neither of which
are true for your system, but both of which are true for PATA on
your system.

So after all that you're saying that I'm right. Right now, SATA is no
better than PATA.

Yes. I never disagreed with you on the end result, all I was saying was
that SATA is not inferior merely because in your system PATA happens to be
faster - I'm saying that it's a result of the drive mechanics NOT the
interface.

I've said all along that the performance of a drive is not entirely a result
of the interface it uses to connect to the computer. If a drive can do
70Megs/s max, then it can do 70Megs/s oer SCSI, SATA, PATA (ATA/100,
ATA/133)...

Ben

On Sat, 23 Aug 2003 07:02:51 -0700, vegan
wrote:

On Thu, 17 Jul 2003 07:59:06 +0100, "Ben Pope"
wrote:

Ransack The Elder wrote:
"Ben Pope" wrote in message
...

You're missing the point. Before you can say This SATA drive is
worse than that PATA drive, you need to be taking into account the
actual drive, not just its interface, not just the manufacturer.

1) I have the fastest 120 gig SATA drive on the market..today, anyway.

OK, but it's still not native SATA.

2) Everything I have read and all the charts I have seen say that
SATA is faster than IDE.

Read more.

3) My IDE drives are faster than my SATA drive.

OK.

So, does that mean:
A) My IDE drive are that damn good.

Yes. BUt because it's a fast drive mechanism and good algorithms, not
because it's PATA.

B) SATA is marketing hype that offers nothing other than a smaller
cable.

No! It offers many useful things such as hot-plugging. I don't know how
many times that would have been useful for me, at least,

Now, what you have is NOT a native SATA drive, you have a PATA drive
with a PATA to SATA convertor chip on it - this may explain why you
have experienced some degradation in peformance over the IBM drive.
Additionally, whereas your ATA interface is built directly into your
chipset, your the data flowing through your SATA interface has to
also traverse the PCI bus (limited in TOTAL to 133Megs/s).

Ummm..okay...so let's say my SATA controller is NOT built onto my
chipset. I would need a PCI SATA card which...bada bing...traverses
the PCI bus. So what's the difference?????

Re-read what I said.

Do you still think you are performing a fair comparison of SATA and
PATA?

Yep. I have the same setup as those two websites you sent.

Ok, then do you think that THEY were performing a fair comparison?

Before suggesting that SATA is worse than PATA we will have to wait
for chipsets and drives to natively support SATA, neither of which
are true for your system, but both of which are true for PATA on
your system.

So after all that you're saying that I'm right. Right now, SATA is no
better than PATA.

Yes. I never disagreed with you on the end result, all I was saying was
that SATA is not inferior merely because in your system PATA happens to be
faster - I'm saying that it's a result of the drive mechanics NOT the
interface.

I've said all along that the performance of a drive is not entirely a result
of the interface it uses to connect to the computer. If a drive can do
70Megs/s max, then it can do 70Megs/s oer SCSI, SATA, PATA (ATA/100,
ATA/133)...

Ben