Performance
Along
with the benefits of using 1394 based hard drives, there are also a couple downsides.
Although 1394 is quoted as having a 400Mb/sec throughput, that's only a burst
transfer rate. Sustained transfer rates are in the 30MB/sec range. Drive access
times are also affected, higher latencies are inevitable when the drive has to
communicate across an additional bus and 2 controllers.
The
other performance bottleneck on Maxtor's 1394 drive is the 1394 to ATA bridge,
which only supports ATA33 for reads and PIO Mode 4 for writes. We
spoke to Maxtor about their reasoning behind the Symbios adapter, and their response
was the same as our research indicated. There are simply no IDE to 1394 adapters
which support modes faster than ATA33.
In
our SiSoft Sandra 2001 benchmarks the drive posts respectable numbers, slightly
above an those of a ATA66 drive. These results are quite surprising considering
that 1394 to IDE adapter
supports only ATA33 mode. This just goes to show that hard drives still have quite
a bit of performance bandwidth even on 'aging' standards like ATA33. With only
a single drive on the IDE chain there is little need for more bandwidth as the
fastest IDE drives are just beginning to push the 33MB/sec sustained transfer
rate.
Using
Winbench 99 we see the sequential read performance across the entire 80GB of the
Diamond Max 80 drive. Normally hard drives exhibit maximum performance on the
outer tracks, and diminishing on the inner. The DM80 on an internal ATA100 connection,
(green), is exactly the performance curve we have come to expect from rotation
based hard drives. However, under both WindowsME (red) and 2000 (yellow) we see
a flat performance line, a bottleneck limiting the maximum throughput. We first
identified the 1394 to IDE adapted as the problem, but the large performance increase
seen when using Windows2000 indicates that the MS Windows9x 1394 drivers are quite
poor. Windows 2000 turned
in numbers 34% faster than WindowsME, a testament to the maturity of the Windows
2000 Firewire drivers. Obviously if you are considering this drive, Win2K is highly
recommended.
Performance
is certainly respectable - a solid 25.5MB/sec on the outer tracks, falling to
17.5MB/sec on the inner tracks. Future driver updates will hopefully bring up
the Windows 9x performance up, and may give 2000 a small increase as well.
On
the individual Winbench99 tests, the 1394 external drive has a mixed showing.
In the Express 3.4 tests the 1394 drive is beaten by a 36% margin, while in the
FrontPage98 tests it tops its' internal sibling by 14%.
Maxtor's
Acoustic Management technology is also put to good use. In a nutshell, Acoustic
Management is accomplished by reducing the speed at which the actuator arm moves
back and forth across the drive, thereby reducing the familiar 'grinding' sound
during disc seeks. This is a unique situation for Maxtor where enabling AM has
few downsides. Usually, enabling AM has two major side effects - disk seek times
are greatly increased (around 50% or so), and sustained transfer rates also take
a small performance hit. In this case with the performance bottlenecks already
discussed, AM has nearly zero effect on the drive's performance.
The only spec affected by enabling AM is the disc access time. Because of the
performance bottlenecks, the hit to sustained transfer rate performance is not
seen. Unfortunately, the only way to change the default AM settings is to install
the drive internally. For this reason our testing was done with the default settings
as shipped.
Both
Winbench99 and Hd Tach report a 22.5ms access time, not fast but respectable.
Other reviews have measured the access time on the Diamond Max 80 internal drives
at 15.5ms, so the combined latency of the 1394 bus and AM translates to an additional
7ms on the access time. However, backup and general file servers are not dependent
on fast access times.
Both
noise and heat output are very acceptable. Even with the drive sitting on the
table it's hard to actually hear it operating. Heat is even less of a concern.
After hours of benchmarking the drive is only warm to the touch.
Conclusion
