Articles: Storage
 

Bookmark and Share

(17) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 ]

We checked the first hypothesis by ear. You can hear the noise produced by SerialATA Barracuda 7200.7 during seek operations much better than the noise of its ATA fellow. Moreover, the latter generates some “unstable” and “clear” sound while the SerialATA drive sounds are very “rich” and “persistent”, which is very similar to the noise of Quantum drives, so dear to me :) This fact may be a proof of the second point. As for the supposition about the influence of the firmware version, it didn’t get any proof this time.

Which of the three points do you like most? I assume that both above described ways contributed to the positioning speed of Seagate Barracuda 7200.7 Serial ATA, because the average access time they demonstrated is beyond all praise and yields a little bit only to some IBM/Hitachi products.

During random writes the gap between ATA and Serial ATA Barracuda 7200.7 drives gets really huge, and Barracuda Serial ATA V proves even faster than all the others. Once again the comparative analysis shows that barracuda 7200.7 have a lot of room for further improvement.

Since all contemporary hard drives cache write requests and perform “background” writing, the average write access time gives us to understand how efficient the lazy write algorithms are. If we divide the average access time during reads by the average access time during writes we will obtain the efficiency coefficient:

Again we observe a very clear system: the HDDs split into several groups. The previous generation of Barracuda drives is notably ahead, having demonstrated relatively efficient lazy writing (compared with the competitors’ products). And Seagate Barracuda 7200.7 can hardly boast anything here: if the SerialATA versions still manage to show some acceptable results, the classical modifications suffered a total fiasco, and even a large buffer doesn’t save them from failure.

In our Samsung SpinPoint PL40 HDD Review we have already mentioned a strange peculiarity of Seagate Barracuda 7200.7 when some test results contradict each other. Despite pretty low results obtained in this IOMeter benchmark they still perform quite fast in SoundForge test from WinBench 99 test package, which also requires efficient lazy writing. Where does this contradiction come from?

Of course, we don’t have any definite answer, but dare suppose that it has to do with the type of oncoming requests. Our experience shows that the latest Seagate Barracuda generations are not optimized for multiple concurrent reads and do not cache the data already read once. In other words, having read a certain data block and transferred it to the PC, Seagate drives immediately forget about it. As a result, any non-sequential data requests interfere with the read ahead algorithms referring to the previous address. No multiple segmentation (although the specs claim it is there), the HDD copes well with only one data stream. If we suppose that the situation is absolutely the same during writing (we didn’t have the chnace yet to check this out), then everything becomes absolutely clear.

In SoundForge test the HDD is addressed more or less linearly, and in IOMeter the requests are random. As a result, lazy writing works fine in the first case, and in the second case the buffer is constantly re-associated. Do you know what the consequences of this situation are? The more fragmented gets the data, the lower gets the HDD performance…

The analysis of the HDD performance with different firmware versions by ATA Barracuda 7200.7 drives indicates that the differences between them are minimal.

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 ]

Discussion

Comments currently: 17
Discussion started: 10/26/03 03:32:21 AM
Latest comment: 07/23/07 08:10:44 PM

View comments

Add your Comment