If Xbench is the only test you run between the
Maxline 250GB and the Seagate 7200.10 250GB then you
will be convinced that the 7200.10 is slower as a
result of the slower random write speeds that Xbench
displays for the 7200.10.
Xbench isn't the only test I ran. Disk tester showed the exact same lack of random write performance with the Seagates. In fact, it was only after the disk tester runs that I decided to put some energy behind sourcing some MaxLine drives (they were hard to get by over here).
However, once you realize that Xbench really is a
very poor test for hard drives, as it does not
reflect actual hard drive performance with video. You
start to look at other tests.
No benchmark is ever going to accurately predict how a piece of hard- or software will perform for a particular user. That's the nature of benchmarks. Most aim at covering common usage patterns, with varying success. Your example of video use seems a rather specialised application, however.
Even though I do deal with video files (typically < 10GB), I still don't fall into the huge file/sequential write scenario. Simply because I don't have a dedicated volume for video files. The same volume also contains all my other data and in fact the OS. Besides, there is hardly ever a situation for me where a video file gets written out as fast as the disk can take it. Creating it usually takes much more time than writing.
If you test a striped RAID set of 7200.10 250GB hard
drives with file duplication you will find that the
Seagate 7200.10 outperforms the Maxline by 10-15%.
That may well be. However, random disk IO is a much more relevant scenario for me. With random write we're not talking about a 10-15% margin, we're looking at the Seagates providing 10%
of the MaxLines' write speed at small block sizes and 60% at larger block sizes.
While the Maxtor wins with Xbench it is actually
slower when testing with DiskTester or duplicating a
group of files. This specific speed capability makes
the Seagate 7200.10 striped RAID set ideal of working
with large video files where large block performance
is needed much more than random write performance.
Again, I don't dispute that they shine on specialised applications. I just don't happen to have a dedicated video volume.
In fact, in real world usage I have not found a
single test where a Seagate 7200.10 striped RAID set
was beat by a Maxline striped RAID set by any
considerable margin.
I have. After switching from Barracuda ES to MaxLine drives the interactive performance of Aperture (including importing files, generating thumbnails, refreshing large pages full of thumbnails, flicking through images in full-screen, making adjustments, etc) increased substantially. I presume this is because those operations rely heavily on small amounts of data (adjustments, metadata, image and thumbnail caches,...) to be written a lot, and all over the Aperture library.
Cheers
Steffen.
Overall I believe you can do better, especially in a RAID set. The thing is, the Barracudas do perform extremely well when you have large, sequential, read and writes. Problem is, they're not especially good at the smaller random stuff.
