Articles: Storage
 

Bookmark and Share

(38) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 ]

And what about reading and writing from the cache? Is there any effect from the drive’s support of 3GB/s SATA?

Well, there seem to be none, but we’ve forgot to remove the jumper that limits the speed of the drive to SATA-150. By default, the drives come with this jumper set for better compatibility with older mainboards and SATA controllers.

So, we remove the jumper and run the test once again (we increased the maximum data block size to achieve the highest speed):

And the result is impressive. The max speed of reading and writing from and to the cache is 255.5MB/s and 241.83MB/s, respectively!

Now what about NCQ technology? As usual, we checked it out with IOMeter: the disk was receiving a stream of requests to read random-address sectors. Every 10 minutes the depth of the requests queue was increased.

The drive proves to be slower with enabled NCQ than with disabled NCQ! As we know from our earlier tests (for details see article called Raptor X HDD from Western Digital: Only the "Naked" Truth... ), the Promise controller enables/disables NCQ quite correctly. But we shouldn’t rely on just one controller in such a tricky matter.

Let’s try to use Intel’s controller integrated in the ICH7R chip with AHCI support on and off.

So, the Barracuda 7200.10 looks somewhat better on the Intel controller (with the latest driver). Its performance isn’t brilliant, but at least not a disaster, either. We’ll perform NCQ tests on the Intel controller as on the “more compatible” one.

What else should we check out? Of course, we need to know how NCQ works on write requests. We ran five more patterns on the drive that differed in the share of write requests. Here are the results:

The graph with zero writes (i.e. 100% reads) looks normal, but the rest of them…

It looks like the drive doesn’t manage to put read requests into the queue and defer write requests as effectively as the WD1500ADFD did. So, the problem is on the Barracuda 7200.10 side. We’ll return to this problem later on when we’ll review the Barracuda ES drive.

But we’ve said enough about one drive. It’s time we introduced all the participants of our today’s tests to you.

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 ]

Discussion

Comments currently: 38
Discussion started: 10/31/06 07:49:21 PM
Latest comment: 04/18/08 03:55:08 PM

View comments

Add your Comment