![[H]ardOCP](/images/blank.gif){kind=small}
- Date:
- Monday , September 15, 2003
- Editor:
- Sean Quinn
- Google +1
SATA Roundup
Thinking about another upgrade and wondering if it's time to jump on the SATA bandwagon? We evaluate drives from Seagate, Maxtor, and Western Digital in a side by side comparison to help make your drive buying decision a little easier.
Benchmarks and Test System
Benchmarks
Hard disks are evolutionary creatures, not revolutionary like graphics cards. Designing new storage technologies takes an extremely long time. It’s no big surprise, then, that benchmarks that focus on storage are few and far between. Sites like StorageReview.com (One of our favorites for storage information.) have lead the way to developing new metrics for testing hard drives by utilizing “patterns” of activity, cohesively designed to emulate distinct styles of computing. For our review here, we are focusing on 3 separate tools: HDTach 2.61, Sandra 2003 Professional, and SCSI Toolbox’s SCSI DiskTools 4.01. Each of these benchmarks has a particular strength and weakness which we will highlight as we progress through our data. Following our objective reviews of the drives, I will add my more subjective thoughts covering noise, temperature, and day-to-day performance.
Test Platform
Mainboard: Abit NF-7S rev. 1.2
Processor: Athlon XP 1700+ @ 2100MHz (10.5x200MHz)
Memory: 2x256MB Crucial PC3200
Operating System: Windows XP SP1 with all available hotpatches
Host Drive: Western Digital Special Edition ATA-100 8MB cache 80GB
SATA Controller: Onboard Silicon Image Sil3112A SATA controller
SATA Driver Version: 1.0.0.32 dated 5/9/2003
SATA Controller BIOS version:
Let’s take a look at HDTach and the results we achieved there.
HDTach Read and Write Results
HDTach is an aging program that performs low-level analysis of a hard drive’s access time, writing speed, and reading speed. It appears to test the read and write capabilities of a drive by performing sequential reads/writes versus random reads/writes. While sequential read performance is key, it represents only half (some would argue even less) of the whole story. In the read graph above, it is very obvious that the Western Digital Raptor maintains a high standard of performance that the other drives are hard-pressed to keep up with. Average Reads, however, are neck-to-neck with Seagate’s 7200.7 coming in last place, and Maxtor’s DMPlus 9 splitting the difference.
In write performance, the same trend continues with the Raptor being the obvious performance champion. What’s interesting to note, however, is the 7200.7 pulling ahead of the DMPlus 9 in maximum writing bandwidth.
Sandra Pro 2003 Results
Sandra evolves almost yearly and is very adept at keeping abreast of changes in the industry. Unlike HDTach, Sandra requires drives to be formatted and then performs random and sequential read/write tests via the Windows File System. By using a sizable dummy file (generally around 511MB), Sandra can replicate data passing through the PCI bus versus speculating on what the theoretical transfer will be. The downside to this approach is that it relies on Windows XP to provide a consistent transfer medium. Only by taking multiple iterations and averaging the results will you receive a reasonably accurate data metric. NOTE: To keep my data in the same measurement criteria, I converted all Sandra scores into kilobytes per second. Sandra defaults to megabytes per second.
Examining Sandra’s read scores, we come across a similar situation to that of HDTach. Buffered Reads (synonymous with HDTach’s Max Read results) are extremely high, with the Raptor coming out strongly ahead. Things slow down as we progress to Sequential reads where all the drives are within several hundred kilobytes of each other. Random Reads, however, represent some fairly terrible scores as ALL drives fall below 20,000KB/s.
Right off the bat, you’ll see something terribly wrong. Instead of the Raptor having the highest score for Buffered Writes, it comes in last place by a margin of 70,000KB/s or so. I tested both Raptor drives in my possession and restarted, retested, restarted again, and could not seem to influence a change in the score at all. That being said, I believe that this particular data point is wrong and further benchmarking using STB Tools will bear this out. Also of note, we see the 7200.7 doing particularly well in the sequential writing portion of this benchmark, coming out ahead of the Raptor and DMPlus 9. All is restored to normal, however, when random writes are instituted.