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

[theoix]

Hey guys

I recently purchased a new ST2000DM001 from Amazon.

I installed it in my computer 3 days ago and I am seeing some ridiculous SMART readings.

Here's a screenshot of CrystalDiskInfo

How is this even possible?

Total Host Reads is 419.09 TB and Total Host Writes is 3.34 PB in mere 49 hours?

All I did was copy approximately 200 GB of data from my old hard drive to this new drive.

I did run various diagnostic tools and benchmarks on the drive. Could that be the cause? But those numbers still seem very unrealistic.

Secondly, isn't 117 for Read Error Rate kind of high?

It was actually 113 earlier today but it has been slowly climbing to 117.

Also, Speedfan reports only 92% fitness and performance for this drive.

"The overall fitness for this drive is 92%.
The overall performance for this drive is 92%."

I think 92% is really low for a new drive consideirng the fact that my 4 year old drive is still at 90%.

I am thinking about just returning this drive and getting a WD Black.

[fzabkar]

Firstly, I would ignore SpeedFan's and CrystalDiskInfo's interpretations of your SMART data. There is no standard governing how these data are presented, so the format can vary between manufacturers, and even between models from the same manufacturer.

As for your observations regarding the Read Error Rate, I have attempted to make sense of this counterintuitive attribute in the following article:
http://www.users.on.net/~fzabkar/HDD/Seagate_SER_RRER_HEC.html

I suspect that the actual values of Total Host Writes and Total Host Reads attributes are probably stored in the lower 32 bits of the 48-bit raw data.

If so, then Total Host Writes = 0x7403BDB4, which amounts to 928GB:

http://www.google.com/search?q=0x7403BDB4+x+512+bytes+in+gigabytes

... and Total Host Reads = 0x8C26C128, which amounts to 1.09TB:

http://www.google.com/search?q=0x8C26C128+x+512+bytes+in+terabytes

[theoix]

Well, I know you are obviously trying to defend seagate but if you should take a look at this person's CrystalDiskInfo reading

http://hardforum.com/showthread.php?t=1699872

He has the exact same hard drive as I do.

However, his Read Error Rate is only 111. Mine started at 113 and has climbed to 117.

His hard drive's total read and writes are not even a fraction of mine. There is something very abnormal about my hard drive's total reads and writes.

Moreover, I am very concerned about my hard drive's Command Timeout. I read that it should be at 0 or very close to 0.

[fzabkar]

I'm not defending Seagate, I'm merely stating the facts as I see them.

With respect, you have a fundamental misunderstanding of SMART data.

See http://en.wikipedia.org/wiki/S.M.A.R.T.

The Read Error Rate figures of 111, 113 and 117 that you allude to are actually good results. In fact your figure of 117 is the best of the three. IIUC, the ideal value would be 120. These data represent the "normalised" value of the RRER attribute. As such, they can be considered to be a health score, with higher numbers being better.

As for the Command Timeout, your drive has recorded 1 (the value in the lowest 16 bits, or 4 nibbles). IMHO, that's nothing to be concerned about.

I don't believe there is anything "abnormal" about your HDD's read and writes. In fact the other thread you allude to has similar numbers for the Host Reads and Writes:

003C_116918EF Total LBAs Written
0002_011B3316 Total LBAs Read

To make you feel more comfortable about the numbers of reads and writes, let's calculate the maximum possible amount of data that could have been transferred based on the Seek Error Rate.

The lower 32 bits of the SER reflect the total number of seeks. Let's assume that the drive reads or writes the entire track after each seek, and that these accesses occur at the outermost zone where the data transfer rate is fastest. Let's also assume that no sector is read or written twice, in which case all data transfers entail accessing the platters rather than cache memory.

Your serial number begins with W1E, which means that your drive is the faster 2-platter version, with a maximum sustained data transfer rate of 210MB/s. At 7200RPM the time for one revolution is 8.33msec. This means that the data density is ...

210 MB/s x 8.33 msec per track = 1.75 MB per track

Total number of seeks = 0x54D73 = 347507

Therefore, maximum amount of data transferred = 347507 x 1.75 = 608GB

This figure is still a lot less than the results I arrived at by utilising the lower 32 bits, so maybe these data need to be interpreted in some other way. Or perhaps much of your data were retrieved from cache, in which case the result could be correct. If you really want to accurately test this attribute, I would perform a complete zero-fill of the entire drive and compare the number of writes before and after. The difference should reflect the total number of sectors on the drive's label.

[theoix]

Ok. I am definitely less concerned after reading your explanation. Thank you.

However, I still dont see how the total host read/writes in the other thread I linked are similar to mine.

Power On Hours : 28 hours
  Power On Count : 8 count
      Host Reads : 4104 GB
     Host Writes : 123019 GB

Compared to

  Power On Hours : 59 hours
  Power On Count : 11 count
      Host Reads : 480429 GB
     Host Writes : 3546021 GB

I do see that my hard drive's power on hours is 21 hours longer but the host reads and writes are still unproportionally higher.

Also, are you sure that the command timeout is nothing to worry about? I read elsewhere that even 1 is dangerous.

Do you mind explaining to me how you derived at 1 for Command Timeout when the raw value is 10001? I would like to know how to calculate it so that I can do it myself in the future.

[fzabkar]

HDD manufacturers don't make their SMART information publicly available. Therefore most of what you read is either speculation, or has been discerned from experimentation. Moreover, since there is no official standard, each manufacturer may define the SMART attributes in their own way.

That said, the raw data for HDD attributes occupy 48 bits (= 12 nibbles). These 48 bits may be composed of several components. For example, Seagate's Seek Error Rate attribute stores the total number of seeks in the lower 32 bits, and the total number of seek errors in the upper 16 bits.

I believe, from examining other SMART data, that the Command Timeout data are arranged in 3 groups of 16 bits. So your result can be viewed as follows:

0000_0001_0001

Therefore the number of timeouts is 0x0001, not 0x10001.

Whether this value is dangerous is a matter of opinion. All I can say is that it wouldn't bother me.

To better illustrate the above, look at the Airflow Temperature and Temperature attributes. In fact these attributes are actually one and the same, and they probably use the same sensor. "Airflow Temperature" is better described as "Temperature Difference from 100".

ie Airflow Temperature = 100 - Temperature

Clearly a temperature of 70C is bogus, but 100 - 70 = 30C is correct.

Now look at the raw data:

00_1A_00_00_00_1E

00_00_1F_1D_00_1E

In this case the data make sense only when arranged in groups of 8 bits.

In hexadecimal, 0x1E = 30, so that's the current temperature in degrees C.

The other values (0x1A = 26C, 0x1D = 29C, 0x1F = 31C) probably represent the maximum and minimum temperatures over the lifetime of the drive, or perhaps during the current power cycle.

As for that other thread, I say that the host reads/writes are similar to yours in that they are similarly nonsensical, and that's because they incorrectly assume that the uppermost 16 bits are significant.

If instead we just take the bottom 32 bits, then we have ...

Total LBAs Written = 0x116918EF = 139GB
Total LBAs Read = 0x011B3316 = 8.8GB