|
Test
System
Because
there's some confusion over which IDE drivers offer
the best performance for NVIDIA's nForce2 platform
I decided to swop it out for our regular Albatron
KT333 board.
Test
Setup
Albatron KX400-Pro (KT333) Motherboard (Kindly
Supplied By The
Overclocking Store)
Maxtor 80GB ATA133 HDD (Kindly Supplied By The
Overclocking Store)
2x256MB Corsair (TWINX3200LL) Memory (Kindly Supplied
By Corsair)
AMD AthlonXP 2200+ CPU (Kindly Supplied By AMD)
Connect3D Radeon 9700 Pro (Catalyst 3.1) (Kindly Supplied
By The
Overclocking Store)
Windows XP Pro + SP1
SiSoft
Sandra 2003
It
may be a synthetic benchmark but it could be argued
that a contrived real-world file read or write is
synthetic in these situations.
Western Digital WD2000-BB (200GB)
Maxtor DiamondMax D740X (80GB)
Failing
any kind unexpected anomaly I think it's fair to say
the WD shows its pedigree under this benchmark, besting
the reference score quoted for dual 120GB ATA-66,
7200RPM drives in a RAID0 configuration. This is by
far the highest score I've seen from a single drive
to date and prompted me to retest several times. Despite
its performance edge the WD2000-BB shoes a weakness
in both its buffered write and its buffered read speeds.
Sequential writes on the other hand are way ahead
and account largely for the superior score.
HDDSpeed
2.1
HDDSpeed
is a DOS based hard drive benchmarking utility that
gives a host of text based performance results, the
important ones of which are quoted below.
| Average
Seek Time |
4.1ms
|
| Maximal
Seek Time |
5.2ms
|
| Average
Access Time |
9.2ms
|
| Track
To Track Seek Time |
1.4ms
|
| Average
Linear Speed |
24.7MB/sec
|
| Max
Cached Read |
63.5MB/sec
|
| Speed
Index |
2765
|
Western
Digital WD2000-BB (200GB)
| Average
Seek Time |
5.2ms
|
| Maximal
Seek Time |
10.2ms
|
| Average
Access Time |
9.9ms
|
| Track
To Track Seek Time |
2.5ms
|
| Average
Linear Speed |
40.6MB/sec
|
| Max
Cached Read |
91.0MB/sec
|
| Speed
Index |
4193
|
Maxtor
DiamondMax D740X (80GB)
Once
more we see the WD's apparent disadvantage when it
comes to cached performance and its apparent edge
in access and seek times with track to track times
especially good. These are significantly faster than
the claimed seek times. Average linear speeds here
are pretty dismal which reduces the drive's speed
index.
At
this point it's worth reminding ourselves that seek
time and access time are not the same thing. Theoretically,
seek time is the time taken for the head to move
from the track the current data is on to the track
that the next required piece of data is on and doesn't
take into account the time it takes for the platter
to rotate the required data to the head. A 7200RPM
hard disk takes around 8.3ms to rotate a platter
through one revolution, so if the head arrives at
the required track say 0.1ms too late to read the
data it will take a further 8.3ms before it comes
round again. Obviously there's an equally good chance
that the head will arrive at the track say 0.1ms
early and it would then take just 0.1ms for the
data to reach the head. To simplify calculations
we tend to average the rotational time (rotational
latency) and work on the time it takes for the platter
to perform a half revolution, thus 4.2ms. Actually
if we're going to get picky,
Access Time = Command Overhead Time + Seek Time
+ Settle Time + Latency
or if we make a calculation based on WD's specs
that's
Access Time = 0.3ms + 8.9ms + 0.1ms(typical)
+ 4.2ms = 13.5ms
For reads and
Access Time = 0.3ms + 10.9ms + 0.1ms(typical)
+ 4.2ms = 15.5ms
for writes.
This
may sound incredibly fast but just to put the cumbersome
hard drive in perspective an Intel 3GHz CPU will
theoretically perform around 12.6 million calculations
in the time it takes a hard drive platter to rotate
once at 7200RPM! Scary!
|
