Is 6 GB/s a hoax?

WD Internal Drives Desktop & Mobile Drives
1

Hello,

I recently bought two Caviar blue HDDS, model WD5000AAKX and have them in a raid 0 config.

Using the MSI 880GMA-e53 motherboard. Ports 7and 8 are for 6 GB/s.

Using the Marvell 88SE9128 controller.

Using HD Tune I hit 225MB/s max.

It appears to me that there should be no jumpers used because I am not using the features they represent.

Downloaded software from MSI for the sata 3 feature.

Why am I not getting 6 or near 6 GB/s on my test results?

So far WD tech support has been disappointing.

I hope someone can offer me real help.

Thank you,

Sabbath

2

I believe you mean 6Gb/s NOT 6GB/s i.e. bits NOT Bytes!

3

The HD Tune result of 225MB/s appears to be the maximum sustained data rate to and from the platters.

6 Gb/s (= 600 MB/s) is the SATA interface speed.

What does HD Tune report for the burst rate? This figure represents the data rate into and out of the drive’s cache memory. It should be much closer to the SATA rate.

4

Hi and thanks for your reply.

When I get home from work I’ll test again and post the result.

Sabbath

5

Hi fzabkar,

Could you plz interpret these numbers?

Ave: 178.5 MB/sec

Access time: 15.5 ms

Burst rate: 2268.2 MB/sec

Thank you,

Sabbath

6

The average data rate is the average “sustained” data rate. It represents the transfer rate to and from the hard drive’s platters. The maximum transfer rate will be at the outermost zone, and the minimum rate at the innermost zone. This is because the outermost zone has the most number of bytes per track.

Data transfer rate (bytes per second) = bytes/track x revolutions/sec

The access time is the time required for the read head to read the target sector. The head needs to seek to the target track, and then wait for the target sector to pass underneath. Therefore the average access time is equal to the average seek time (approximately one half stroke), plus the average rotational latency, plus the OS overhead. A 7200 RPM drive has an average latency of 4 msec (one rotation requires 8.33 ms). If you examine the spread of data points in your access time graph, you will see that it spans about 8 msec, ie the latency of one full rotation.

As for the burst rate results, I confess that I don’t have a satisfactory explanation. AISI, two striped 6Gb/s SATA drives would have a maximum data rate of 2 x 600 MB/s, ie 1200 MB/s. However, your result is nearly twice that figure. Either HD Tune is miscalculating the data rate, or perhaps the result reflects a read cache in the OS.

BTW, does the read performance curve have a long flat section at the beginning, or does it have a smooth downward slope?

7

Hello and thank you for the detailed post.

When I test it is generaly more straight and starts to peak and valley then about 2 thirds through it starts to drop off.

I downloaded HD tune pro and it sees my sata 3s as sata 3s but my test early on dips near the bottom then back up to around 200 and down again. Not for the duration of the test but around 3 times. Wierd. Anyway, after work today I’ll run it again then post a screen shot.

It does show like a 2.5% CPU usgae. I like that. The raid controller isn’t so CPU dependent.

Sabbath