SATA vs. IDE

First off, I realize this is not a hard drive newsgroup. However, I never
visit any hard drive groups, and I know the folks here know what's going on
so I'll ask here...plus I have an AMD system

I just upgraded to a Chaintech motherboard with the Promise SATA controller.
I WAS excited about getting my Maxtor Diamondmax 9 120 gig SATA drive going,
until I benchmarked it with SiSoft Sandra. It runs almost identical speeds
as my IDE Hitachi/IBM 120 gig drive. Both of them were almost as fast as an
IDE RAID setup that Sandra had listed to compare to. However, I expected
SATA to be much faster than IDE. So I'm quite dissapointed so far.

My question is this: the controller has 2 SATA ports and one IDE port. My
main drive (The IBM IDE 120) is on that IDE port, and the Maxtor SATA is of
course on one of the SATA ports. Does having this IDE drive on that
controller limit the speeds to ATA/133 speed?? I was planning on replacing
the IDE drive with SATA, but not if it's no faster. But if doing that will
'open the flood gates' so to speak, I'll do it.

I'm eager for answers, and in the meantime (when I get some time) I'm going
to move the IDE drive off the Promise controller and put it on the other IDE
ports that the board has and leave the SATA drive dedicated to the
controller and see what happens.

 

Just an update..

I moved the IBM drive to a seperate controller, so each had their own
controller. The IBM is still a hare faster than the SATA drive, isolating it
on the controller did not help with the speed. So I can only conclude two
things now:

1) IBM drives are just that damn good. I tested some Western Digitals, and
they were much much much slower than the IBM.

2) SATA seems to be hype at this point. When my IDE drive is faster than
SATA, what else can I think?

 

If it's speed that you're after, get yourself a Western Digital Raptor
drive.

 

Nah. I hate WD drives. Plus 36 gigs is just not big enough for me,
especially for the outrageous prices they sell these drives for.

I'm not looking for extreme speed, I'm just looking for speed faster than
ATA/100, which is what SATA is supposed to do.

 

So I should not expect SATA to be any faster than IDE? Well that sucks.

 

Only if you put a drive on the SATA that's capable of higher that ATA-
100 transfer speeds. When you get the aureal density of a Hitatchi/IBM
drive with a rotational speed of 10,000 to 15,00 rpm platter, then you
will see an increase in data throughput. Or you could run 2 drives
striped Raid0 for an increase in speed.

Bill

 

Yep, I have a SATA controller and a SATA drive. Everything works fine, SATA
is just slower than IDE for some reason.

 

Yep. Not fun to get suckered by the marketing types, is it? The
numbers you want to look at are the "Buffer to disk" transfer rate and
the average access time, which is the sum of the average seek time and
the average latency. There's also a settling time in there but it's
usually relatively small and not always available from the drive
manufacturer.

For a 250 gig 7200 RPM WD drive, the maximum buffer to disk transfer
rate is 93.5 MB/sec, while for the WD Raptor it's 102 MB/sec. Average
access time on the 250 gig drive is 13.1 ms for reads (usually longer
for writes) while for the Raptor it's 8.19 ms.

Now, those are _maximum_ transfer rates, which means that they are for
the outermost track (modern drives use Zone Bit Recording, which is a
Seagate trademark, but others use the same technology with different
names), which put more data on the longer outer tracks than on the
shorter inner ones, hence the transfer rate for the innermost tracks is
lower. It also means that they do not include any seek time. Figure
that in and the transfer rate goes _way_ down. For example, for the
raptor, the thing transfers 102 MB/sec max. Divide that by the RPM and
multiply by 60 sec/min and you end up with 61 MB per revolution, or 610
kB per track on the outermost track. Assume that the drive can switch
heads in zero time, and that any transfer reads an entire track, and you
can then read 122 MB without a seek. That transfer takes 2 revolutions
or 12 ms. Add in a seek and that goes to 20.19 ms. So you transfer 610
kB in 20.19 ms or 60 MB/sec. If you play Fun With Numbers with the
Cheetah X15, which I believe is currently the fastest drive on the
market, you get 92 MB/sec.

But it gets worse. Drives have internal formatting that is used to
allow the drive to find whatever sector it's supposed to find, and to
deal with hot-sparing, error correction, and the like. These cut into
the amount of space that is available for user data, and so into the
transfer rate available for user data. Seagate reports the maximum
formatted internal transfer rate for the Cheetah X15 as 86 MB/sec. So,
using that as the single-track transfer rate and figuring in seek time,
the performance goes down to 72 MB/sec actually available to use.

 

Better objective analysis here:

http://storagereview.com/

Be sure to read the 'forum' for the users feedback!

The *.pdf is here:

http://www.maxtor.com/en/documentation/data_sheets/diamondmax_plus_9
_data_sheet.pdf

There appear to be two versions of the drive in the 'chain'. Newer disks
with the YAR41BW0 firmware, and older YAR41VW0 firmware versions. The
newer are "said to be" the faster/better!

Hope this helps!

BoroLad

 

Makes sense, I had to RMA four Seagate Barracuda IV's <IDE> that were
slow in Raid but OK in normal usage. Seagate sent me four new ones
with updated firmware and a note saying that the drives would be slower
than 'normal' Barracuda IV's if used on a non-raid controller. <shrug>

Bill

 

If you want some numbers, Sandra says for throughput:

WD 8M buffer 120G ATA-100 26.6 Mb

WD 8M buffer SATA Raptor 34.2 Mb

2 Seagate Baracuda IV 2M
buffer 80 G
Running Raid 0 45.8 Mb

Bill

 

Well, I don't feel so bad then. My IDE IBM is faster than the SATA Raptor!

 

The raptor is the ONLY SATA drive to beat IDE at the moment. Most of
the models selling both SATA and IDE models are actually slower on
SATA. The raptor is based on a SCSI drive and compares very well to
other 10K RPM SCSI drives in most tests. Most are slightly slower
than SCSI but it still manages to kill any other IDE/SATA drive out
there. You still get most of the quality of SCSI as well. The drive
is very solid and has a 5 year warranty. I ordered one for my system
drive. My data goes on a normal 7200RPM IDE drive.

Eric

 

Unfortunately, that doen't seem to be the case on my machine. If you
know of some newer drivers somewhere I'd be happy to download them and
see how well they work.

Bill

 

Why did you expect SATA drive to be faster than a standard ATA one?

No. Two entirely seperate controllers.

The speed of the drive depends on the speed of the drive, not the speed of
the controller.

OK, you have to remember that the limitation of getting the data off of the
media is mechanical. The "drive" as in the mechanics is kind of a seperate
issue to the controller they use to transfer that data.

For example, SCSI drives have been rotating at 10,000 RPM and 15,000 RPM for
years, double what IDE drives do. The WD Raptor is effectively one of those
SCSI drives but with a SATA interface. SATA provides some features over
normal ATA, a small increase in maximum transfer rate is one, but not the
primary concern at the moment, since most drives aren't able to reach
100MB/s sustained transfers.


Ben

 

Remember that the interface (IDE, SATA) has a speed rating that just
means, what is the maximum bandwidth available across this link? Just
about any of today's formats from UDMA 66 upwards have enough capac-
ity for very fast drives indeed, including these high-speed Raptors
and IBM's (though I have a feeling you're not reading the numbers
entirely correctly - the Raptor tops out every single performance
comparison I've seen anywhere... simply an awesome drive and I can't
wait til it's available bigger than the single-platter 36GB it is
now... but then 72 would be heaps for me and 2 of those things in a
RAID config. would be superb. They also have the lowest CPU utilisation
figures I've seen, even beating some pretty good SCSI systems - this
is another important factor for the net performance of your PC).

So anyway, you have drives that can deliver a certain amount of data
per second, and interfaces that can be up to several times faster
than this maximum rate. You only need to consider the higher-speed
interfaces if there is more than one of these fast drives on them -
for instance, 2 Raptors both going fast will use up most but not
quite all the bandwidth available on UATA 133. Putting a fatter pipe
on the drive doesn't speed the drive up, as in today's systems the
drive interface isn't limiting the transfer rates you're getting...
the drive itself on any new-ish interface will not be limited by that
pipe, and conversely, a faster pipe won't give you any boost in itself.
That said, there's potential for more efficient protocols to be im-
plemented with new formats, but I haven't heard that anything of that
sort applies with SATA vs. UDMA

Those figures we've seen in this thread were mostly peak rates, and
of course, no drive delivers this sort of rate all the time. It's
probably the case that even UDMA 66 will rarely limit what you're
getting, given the *average* throughput of the drives, although it'd
be a shame to be shackling one of these speed-demons at all when fast
controllers are available cheap as an upgrade. It's especially when
you have two drives on the one channel you start to need extra
speed, and then how often do you really have both drives under load
simultaneously, unless you're in certain RAID modes (like below)?

Me, I can't wait to RAID up a couple of Raptors and run them with
striping! I deal with some big files including multitrack audio and
DV so this would help me keep everything real-time... the low demands
on the CPU are a big bonus for this work too, although RAID causes
that figure to jump quite a lot. Still, with these drives, the end
result is still a lot better than any other ATA-based drive I've
heard of. At Tom's Hardware, there's some comparisons of these and
other drives and useful background info. about the figures. It's
plain that there are some mighty fine drives on the way and this WD
is the first of a new breed. When other companies bring out their
answers to this drive, the market will be looking very good indeed.

 

"My IBM P-ATA", "My Maxtor SATA" - you're still talking about the interface
as if it's the drive. Which IBM? Which Maxtor? What rotional speed? What
size Cache? What average seek? The interface has nothing to do with any of
those factors, which all affect the performance of the drive. THE INTERFACE
DOES NOT DETERMINE THE SPEED OF THE DRIVE!

Based on a single drive that happens to use that interface?

I didn't realise all WD drives have all had exactly the same performance.

Clearly you still have not seperated the interface from the drive, when you
learn to do that you may be able to look objectively at different drives and
the interfaces they have. At that point you may even have a chance of
making an informed decision.

Ben

 

Same here, except it's performance isn't that spectacular on my pc. It
may be a driver issue, although I'm running the latest < that I know
of, 1.0.0.33 > drivers. You have tried this drive on your own machine,
haven't you? Numbers I supplied were from Sandra 2003.3.9.44. I believe
they reflect throughput and not burst transfer.

This pc has 7 hard drives for a total of 476GBs.

4- 80GBs Seagate Baracuda IVs
1- 120GB WD Caviar
1- 36GB WD Raptor

Not counting USB 1.1/2.0, and Firewire removable drives.

Bill

<snip>

 

Same brand, same series, same number of platters, same access times,
etc? Same RPM and same cache size does not imply equivalence.

"Nothing to do" is perhaps too strongly worded, but very little to do.
The limiting factors are the number of bits on a single track and the
RPM. The interface can be infinitely fast and the drive still won't be
able to move data faster than the number of bits on a track divided by
the time it takes to move that track past the head.

There is a notion that all sorts of advanced techniques are used by
modern drives to overcome that limitation, that notion is false. With
very rare exception, contemporary drives have one head per track and
read one head at a time. They don't have two heads on a single track
and they don't read all the tracks simultaneously, so the limitation is
the ability of the mechanical parts to move bits past the head.

ATA/33 can be a limitation for many contemporary drives. Very few can
fill an ATA/100 pipe for even a single revolution (as soon as seek time
is factored in the data transfer rate goes way down), and none can fill
an ATA/133 pipe. With SCSI, the bandwidth is better utilized as there
can be up to fifteen devices sharing a channel--with IDE it is at most
two and with SATA only one device is allowed on a channel.

 

All drives are the same as I stated..the only difference is the INTERFACE.
All 7200 rpm, all 8 meg cache.

So since the interface has nothing to do with the speed, you are stating
that there is no difference between ATA/33, ATA/133, SCSI, and SATA. Please
explain.

Yep. Especially since I have the fastest SATA drive on the market, other
than the Raptor. I need more than 36 gigs though..that would hold only 2
files for me at times.

I didn't either. Where did you read that?