Seagate Barracuda 7200.12 / LP vs. WD Caviar Black /Green in Linux

[Pepazdepa]

I would like to buy new drives for my Linux machine. The one thing I'm not sure is which hard drives to go. Generally, WD drives seems to perform better, despite the fact, they have lower density platters (320GB) Seagate 7200.12 and LP have 500GB platters, but in real world performance (ie copying files, multitasking), they usually loose with WD. But all these benches all over the web are made on Windows.

Is the situation on Linux different or the same? Which drives would you buy and why?

Barracuda 7200.12 or Caviar Black?
Barracuda LP or Caviar Green?

Thanks for any suggestions.

[frantaylor]

This chart may help a bit with your decision-making process:

http://www.coker.com.au/bonnie++/zcav/results.html

I have a bunch of the 7200.12 500 Gb drives. They perform just like the 1 Tb Seagate drive in the graph. In my RAID array they perform like champs. They are very quiet, and they use very little power, so they don't get very warm.

Shop around and you can get them for $50 each.

Make sure you get the CC35 firmware. All of the drives I've seen with CC34 firmware have died very early deaths. The CC35 drives are fine.

[Pepazdepa]

unfortunately those sequential read/write test have nothing to do with real world usage. Seagate traditionally have best sequential read/write numbers, but fails to compete with WD and Samsung when it comes to real usage like coping files.

Thats why I am asking real world tests.

[frantaylor]

unfortunately those sequential read/write test have nothing to do with real world usage. Seagate traditionally have best sequential read/write numbers, but fails to compete with WD and Samsung when it comes to real usage like coping files.

Thats why I am asking real world tests.

It does indeed depend on what you are doing. There are two issues: head latency and data transfer rate. If you are doing lots of random access, you should really consider using a SSD if possible. Drives with high rpms and low latency are great, but they are power-gobbling, heat-generating monstrosities that require special care in cooling.

Often one's data storage requirements can be split into two: a small data set that requires very fast access, and a larger one that is not accessed so often. It is well worth expending the effort to split your data up like this. In this case you are best served by a small fast drive and a large slow one working together.

It's been my observation that the choice of file system, and how it is tweaked, can have a bigger impact on performance than which hard drive vendor you choose. If you are using a motherboard SATA controller that doesn't support NCQ, you are not getting peak performance from your drives anyway. It's like obsessing over the V-8 engine that you then proceed to drop into a Pinto chassis.

My application calls for lots of sequential access. I use XFS and keep the image files defragmented, so I see close-to-hardware data transfer rates. Those big fat numbers in gkrellm make me smile.

You have to take reliability into account also. Personally I have had very bad luck with Samsung and WD drives. I have a big stack of dead hard drives on my shelf, and they are all either WD or Samsung. All my old Seagate drives work fine, even the really old 4 Gb IDE drives and the ancient 1 Gb 10000 RPM SCA drives from my old SPARCStations.

I know people who work in data centers managing multiple-terabyte datasets, and they only use Seagate drives. These are applications (healthcare) where lives are put in danger if the data is not available.

[graysky]

I have both a 7200.12 and a WD Black 1 TB drive on this system.

The WD1001FALS (WD Black 7200 RPM/32 Meg cache with 3x333 platters) using hdparm under Linux x86_64 is a bit slower than the Seagate drive. Here are the results, first the Seagate (results are averaged of three runs)

# hdparm -Tt /dev/sda
Timing cached reads:   14765 MB in  2.00 seconds = 7382.33 MB/sec
Timing buffered disk reads:  367 MB in  3.01 seconds = 121.82 MB/sec

Now the WD drive (results are averaged of three runs)

# hdparm -Tt /dev/sdb
Timing cached reads:   12991 MB in  2.00 seconds = 6495.33 MB/sec
Timing buffered disk reads:  308 MB in  3.01 seconds = 102.33 MB/sec

So roughly 14 % faster on the cached reads and 19 % faster on the buffered reads. Dunno what this means in terms of access times or the like.

The Seagate runs about 5 C cooler as well (2x500 G platters vs. 3x333 G platters in the WD).

[EarthMind]

I just got myself a Seagate Barracuda 7200.12 and I have to warn you that it's very load! It makes very load scratch noises when busy. And as for performance I don't have any complains.

To be honest, if I knew it would be this loud I'd definitely have chosen a different HDD.

[graysky]

I just got myself a Seagate Barracuda 7200.12 and I have to warn you that it's very load! It makes very load scratch noises when busy. And as for performance I don't have any complains.

To be honest, if I knew it would be this loud I'd definitely have chosen a different HDD.

That is the polar opposite of my experience with the drive. I cannot hear it running even when copying files. Perhaps you got a loud one or it is a symptom of something bad..? I have the 1 GB version btw. The WD black 1TB on the other hand is louder when read/writes are in progress, but silent when idle. Just my experiences.

[Panix]

Very interesting. I was considering the Seagate because the two platter design interested me. I have read that the WD Black is okay but not the best with power consumption and heat production. Not that it's alarmingly bad or anything, though. My two choices was the Seagate and the Hitachi (the newest gen. of 1000.B)
Hitachi Deskstar 7K1000.B 1TB.

I'd probably get the Hitachi since the price is right, well, of as right now.

[movieman]

There's supposed to be an issue with the Caviar Green drives parking the heads after a few seconds idling to save power, but I haven't noticed any problem with that on my system with 1TB drive. However, that's a combined MythTV backend, Zabbix and file server, so the disk is probably never idle.

[EarthMind]

I got the 1TB version and compared to my older Maxtor 40GB it's very load!