Barracuda XT, Barracuda, Barracuda Green, DiamondMax and other desktop SATA and IDE drives

[barracuda378]

I purchased three ST1500DL003 drives from the same vendor at the same time.  All drives show firmware CC32.   Only one of them is showing the sustained transfer rate that I expect.

I have tried swapping SATA cables and motherboard ports, and the behavior tracks the drive, therefore this is not an issue of SATA cables or the SATA port it is attached to.   I also tried using the CC32 firmware update, but it said they are already the latest version and would not be changed.

Here are HD Tach results for the three drives:

ImageShack album

You can see that one of the drives peaks well over 140MB/s, while the other two peak at less than 130MB/s.  The drive is specified at 144MB/s sustained transfer.

Are the slower drives defective?  If they are not, how can I get them to be as fast as the other one, and meet specification?

[Cantbecanit]

A drive always tails off towards the end of the tests.

Yours drops to 60, my 3TB drops to 80, but the 3TB is a 7200rpm unit, yours is 5900rpm.

[barracuda378]

I am well aware of the falling speed curve of HDDs from outer to inner tracks.  The point is that two of the drives are not as fast as the other one, and that these are well below spec.  Please look at the charts again.

[fzabkar]

Short answer: I suspect that the performance difference is due to differing platter densities.


Long answer:

If you are willing to spend $10 on the following, then I believe a terminal report from the drive's RS232 serial interface may provide the answer.

Prolific USB to TTL Logic Cable (US$10):
http://www.serialstuff.com/products/Prolific-USB-to-TTL-Logic-Cable.html

You can follow the instructions at this URL ...

http://www.overclock.net/hard-drives-storage/457286-seagate-bricked-firmware-drive-fix-pics.html

... but don't send any of the commands to the drive. Instead I believe you need to send the Ctrl-L sign-on command.

The drive will respond as in this thread:
http://forum.hddguru.com/seagate-7200-t10899.html

Notice the two blocks of head characteristics:

Head 0, PhyCyls 000000-03C47E, LogCyls 000000-03C14E
Head 1, PhyCyls 000000-03AD26, LogCyls 000000-03AA06

The outer zones (Zn 00) of heads 0 and 1 have 0x0A3C (= 2620) and 0x0990 (= 2448) sectors per track, respectively.

The data rate is given by ...

data rate = sectors/track x bytes/sector x rotations/sec

The inner zones would have progressively lower transfer rates due to a reduced number of sectors per track. In your case the ratio of maximum to minimum transfer rates is about 2.4:1 which no doubt correlates with the ratio of the outer and inner track diameters.

You can confirm that the performance difference is not due to a difference in RPM by examining a HD Tune access time graph. The data points will be scattered over a band whose width corresponds to the time for one complete rotation. In the case of a 5900 RPM drive, this will be about 10 msec.

You can then confirm that the performance difference is not due to differences in physical sector sizes by examing the drive's response to an ATA Identify Device command. This response consists of 512 bytes of information which includes the drive's logical and physical sector sizes. An Advanced Format drive will have 4KB sectors as opposed to 512 bytes. According to WD, AF sectoring typically provides a performance gain of between 7% and 11%.

You can use the Text Copy function of CrystalDiskInfo to obtain the Identify Device Data, or you can just launch SeaTools and allow it to automatically generate a 512-byte 9VT16Lnn.ATA file for each drive. The name of the file is the drive's serial number. If you could upload these files, then we could examine them.

Assuming that all 3 drives are 4KB-sectored and spin at 5900 RPM, then the performance difference must be due to a different number of sectors per track. AISI, one way to explain this would be if the faster drive had 750GB platters while the slower drives used 600GB.

All other things being equal, a platter that has 4 times as many bits per square inch would have twice as many bits per inch. Therefore the relationship between platter density and transfer rate would be ...

rate A / rate B = sqrt (density A / density B)

If we take the transfer rates to be 143MB/s and 128MB/s, respectively, then ...

rate A / rate B = 143 / 128 = 1.117

sqrt (density A / density B) = sqrt (750 / 600) = 1.118

Therefore your results fit the hypothesis.

However, this calls into question the accuracy of the product manual:

Barracuda Green SATA Product Manual, Rev. C & Rev. A:
http://www.seagate.com/staticfiles/support/docs/manual/desktop/Barracuda%20Green/100649225c.pdf
http://www.seagate.com/staticfiles/support/disc/manuals/desktop/Barracuda%20Green/100649225a.pdf

Section 2.1 (Specification summary tables) indicates that the 2TB and 1.5TB models both have identical numbers of heads (6), discs (3), Bytes per sector (4096), Recording density, Track density, Areal density, Spindle speed, Internal data transfer rate, and Sustained data transfer rate OD.

AISI, if everything is the same except for the capacity, then the 1.5TB must be a short stroked version of the 2TB. However, this is not reflected in a lower seek time (both models have 12ms averages). Furthermore, short-stroking would also show up in the benchmark graph as a reduced max/min transfer ratio, but this is not the case. So this begs the question, are the specifications in error?

Yet another oddity is the fact that all 3 of your drives have identical serial numbers. AIUI, the first 3 characters reflect the manufacturing location (9 = KRATSG TK - Thailand), the second character reflects the platter type, and the third character indicates the number and type of heads. I could be wrong, though. In fact my hypothesis of a difference in platter density is at odds with the fact that the second character (V) is the same for all three drives. Stranger still, I found a reference to an ST2000DL003 with a serial number 5YD13MYA which also appears to be at odds with my hypothesis.

One other observation is that a 1.5TB drive with a data density of 750GB per platter would only need 2 platters, not 3. Can you measure any differences in the weights of your 3 drives?

[barracuda378]

fzabkar, that is an interesting possibility.  I have a limited window for returning these, so I will have to limit the analysis I do; I don't think I'll try the RS232 thing.  Later I can try weighing the drives. 

I believe that "9VT16L" is an extended model number, not a serial number.  At least that is the way it is reported by software utilities.  The actual serial numbers are an eight character alphanumeric.  The fast drive and one of the slow ones' differ by only the last two characters; the second slow drive differs in the last four.

You raise an interesting question:  how do the 1.5TB drives differ from the 2TB drives?  An unpleasant possibility comes to mind:  if these are short-stroked 2TB drives, perhaps these are 2TB drives that had physical defects on the outer or inner edge of the platters, and they set a 1.5TB range to avoid these.  If the drive is short-stroked toward the outside then it is faster; if toward the inside it is slower.  If this is the case, then the slow drives are defective by design, and their STR spec is a lie, and this is unacceptable.  I really hope this is not the case, but it would explain why the 1.5TB model exists and is priced less than the 2TB model, if they are physically identical.

(I am attaching a more precise STR chart from a full surface scan of the drives.)

[fzabkar]

Sorry, I goofed with the serial numbers. :-(

I should have written that "the first character reflects the manufacturing location, the second character reflects the platter type, and the third character indicates the number and type of heads". The remaining 5 characters are unique to your drive. Hence, these first 3 characters are not super secret. Are they "5YD"?

FYI, here is my database of Seagate serial numbers:
http://www.users.on.net/~fzabkar/HDD/seagate_sernum.txt

Here is a thread which discusses short-stroking a drive to achieve performance gains:
http://forum.hddguru.com/modifying-the-seagate-5tb-hard-drive-t16073.html

It has two examples of WD drives that have been short-stroked at the factory.

That said, your latest transfer rate graphs prove that your drives are operating with a full stroke. I say this because you can see that each drive has 17 performance steps corresponding to 17 zones. This is identical to what we saw in the terminal report that I alluded to in my previous post.

Instead it now seems to me that the differences may be due to variations in head characteristics. That is, a better performing head would allow a platter to be recorded at higher bit densities, resulting in a higher transfer rate.

See the following article for an explanation of Variable Bits Per Inch (VBPI).

HDD from inside: Tracks and Zones. How hard it can be?
http://hddscan.com/doc/HDD_Tracks_and_Zones.html

The maximum sustained transfer rates in your latest graphs are 125MB/s, 120MB/s, and 141MB/s. It could be that all are due to head variations, but it could also be that the faster drive has higher density platters due to a generational technology difference.

BTW, here is how a short-stroked 3TB Barracuda XT performs:
http://i1080.photobucket.com/albums/j326/cant3/24-June-2011_13-42.png

In order to circumvent the 2TB MBR limitation, the 3TB drive has been split into two physical drives (2TiB and 746GB) by means of a driver. As far as HD Tune knows, it is testing a 2TB drive. Therefore one would expect that the reduced stroke would result in a reduced access time, but this does not appear to happen.

Here is a HD Tune benchmark for an uncut 3TB drive:
http://img854.imageshack.us/img854/6857/goflexdesk3tbusb30hdtun.png

[barracuda378]

Yes, the serial numbers begin 5YD.  What is the purpose of keeping these secret anyway?  I would have included them but I observed that they were removed from other posts by a moderator.

Thank you for your help with this.  You know that 17 zones must be across the full width of the platter?  It cannot be over a restricted portion of a platter?

I shall read through the material you referenced, but for pragmatic summary:

Do any of the scenarios you propose allow for the slow drives to be made faster?

If not, would you agree with the statement that these drives are below spec and therefore defective?

[fzabkar]

I can't answer your question as to whether Seagate would cram the same 17 zones into a short-stroked platter. However, the WD examples that I alluded to previously do not appear to do it this way. That is, the widths of each zone remain the same, resulting in a reduced number of zones when the drive is short-stroked.

As for speeding up the drive, I don't believe this is possible, at least not in the field. The speed limitation appears to be due to differences in the number of bits per track. That said, I wouldn't conclude that the drives are defective, but clearly they don't perform to the published specification.

I'm willing to bet that your Green drives have relaxed specifications because their normal usage is in external enclosures. In such applications the performance would be throtted by the USB or Firewire interface, rather than the drive itself. I suspect that similar transfer rate variations would not be observed in Seagate's 7200 RPM drives.

[barracuda378]

UPDATE: I returned the two slower drives to the vendor, and received replacements. One of the drives tests like the fast one, second down in chart.gif above, and the other tests slower, like the first drive in chart.gif. Also, I contacted Seagate support, and got an email asking for further information. I responded to this on July 1st, but I have yet to hear back.

[Cantbecanit]

If the two new drives are doing the same as the old this suggests either a driver or a bios issue with your motherboard.

Latest chipset drivers especially the USB ones.

Latest bios and flash it

Check MB maker site for known issues

If you are on Intel, get the RST driver as well.

Knowing what you are using them with would help as well.