Two identical hard drives with different benchmark read performance

Hi

I have two WD Caviar Black 750 Gb (SATA 3) harddrives on an Asus P8Z68-V motherboard, connected to the SATA 3 sockets. The hard drives are identical, same hardware same firmware. One is manufactured in june 2011, the other one in july.

My problem is that my  hard drives perform differently on benchmark tests HD Tach and HD Tune. One of the drives performs as expected for a SATA 3 drive with avarage read of about 113 MB/s. The other one stays at 82 MB/s, a pperformance typical for SATA 2 drives. The “low-performing disk” is reported by HD Tune with avarage speed 70 MB/s, the high-performing one with avarage speed 97 MB/s. I have checked so there is no SATA 2 jumper installed on the drive. See screenshots of my results in HD Tach (there are similar results in HD Tune). Can you please help me with this and tell me what it can depend on? Is this a reason enough for RMA?

SATA6G_2.PNG

SATA6G_1.PNG

ISTM that the difference in the transfer rate may be due to differing platter densities. The rule of thumb appears to be that transfer rates increase in proportion to the square root of platter capacity, all other things being equal. For example, if we quadruple the number of bits per square inch, then this results in a doubling of the number of bits per inch (or bits per track). More bits per track equals more bytes per second.

Your results would suggest that the slower drive may have twice as many platters as the faster one:

(150 / 106)^2 = 2.002

In short, your slower drive is not broken, it’s just constructed differently.

Here are two HD Tune examples with a max transfer rate of 132 MB/s:

http://cfile30.uf.tistory.com/image/137E3E4C4E3779A82E7500
http://down.playwares.com/xe/files/attach/images/53/377/014/f2e38f093a4f57579cdb9a60f1063d6f.jpg

The above is consistent with a drive that has 500GB platters and 3 heads. Your own drives appear to have platter densities that are on either side of this spec.

BTW, I notice that the ratio of max:min transfer rates is about the same for both your drives, and for the additional examples.

150 / 70 = 2.14
106 / 49 = 2.16

This is consistent with a full stroke, or near enough.

One more thing I would check is the spread of data points in HD Tune’s access time graph. A 7200 RPM drive would have a band that is 8.33 msec wide (= time for one revolution), whereas a 5400 RPM drive would have an 11ms spread.

Any differences in the DCM (Drive Configuration Matrix) and DCX numbers on the label may reflect internal differences.

Thanks fzakbar! Is it really possible that two harddrives of the same model (WD7502AAEX, SATA 3, 7200 rpm) could be constructed differently? These two are bought within a couple of weeks from the same dealer, they are manufactured within one month, have same firmware installed and are identical on the paper. Here are the benchmarks from HD tune:

slower drive:

slower drive.PNG

Faster drive:

faster drive.PNG

As you can see the spread of data points for both drives is identical (I guess 8,33 ms) suggesting the rotaion speed is same.

Cheers

braco

The graph for the faster drive is a bit fuzzy, but I believe it is in line with the two 132 MB/s examples that I alluded to. Therefore I expect that the faster drive has two 500GB platters and 3 heads, while the slower drive may have three 250GB platters and 6 heads. The only other explanation I can think of is that the slower drive may compensate for its lower number of bits per track by increasing the number of tracks per inch so as to give it the same areal density (500GB per platter) as the faster drive.

I suggest you weigh your two drives.

Did you compare the DCM and DCX numbers on the labels?

Thank you fzabkar!

Yes. I have compared DCM and DCX numbers. The DCM number is the same for both drives. The DCX is different.

What does it mean?

Two 500 Gb platters add to 1000 Gb. Is it possible that those can be used in a 750 Gb drive?

They would be short stroked (from the factory) if you did indeed have 2 500GB platters in a 750GB drive. The inner-most portions of the platter, which have the slowest reads and writes, would not be used for data storage. It’s probably cheaper for WD to create 2 500GB platters than 3 250GB platters. Price discrimination is the reason they continue to offer a 750GB drive despite having to short stroke the 500GB platters.

exxas, AISI neither of the OP’s drives has been short stroked.

A 750GB drive can be manufactured from 2 fully stroked 500GB platters. A platter consists of 2 surfaces, each with a 250GB capacity. Therefore a 750GB drive would use 3 surfaces. If, instead, WD chose to use all 4 surfaces, then each would need to be short stroked. However, such a drive would have a higher minimum transfer rate than a fully stroked one. This is not supported by the benchmark results.

As for your observation regarding the economics of 3 platters versus 2, here is an example where a 500GB WD5001AALS HDD has been manufactured from 2 short stroked 500GB platters:

http://www.tomshardware.com/forum/261937-32-wd5001aals-00e3a0-possibly-500gb-platters-shortstroked

I believe that it has 3 heads and 3 active surfaces. That doesn’t make any economic sense to me, unless it is a refurbished unit that has a faulty head or bad media.

The reason I’m suggesting that the OP’s slower drive has 3 platters is because its transfer rate is what you would expect from a 250GB platter spinning at 7200 RPM.

The DCM is explained here (at least for older models):
http://forum.hddguru.com/western-digital-what-dcm-t6488.html

As for the DCX, I don’t know what it means, but it appears to related to the model number suffix in some way. FWIW, here are my observations:

http://community.wdc.com/t5/Desktop/Is-this-a-genuine-or-fake-WD-drive/m-p/225354#M6350
http://community.wdc.com/t5/Desktop/WD20EARS-00MVWB0-and-AAM-option-Supported-and-Not-Supported/m-p/222972#M6299