To obtain the benchmark results tabulated on the left, Storage Review
used a plain installation of Windows 95 OSR 2.1. We then patched the
default installation with Intel's
82371xB INF Update (also known as setupex.exe, needed to recognize
the newer 430TX and 440LX chipsets) and Microsoft's
Removable IDE Device Patch. Finally, under the test drive's entry
in System Properties' Device Manager we checked the "DMA"
box, enabling the OSR2 Bus Mastering drivers. The PIIX results were
obtained by taking this configuration and also installing Intel's
PIIX v3.01 drivers. IBM's Deskstar 5 6.4GB drive, posting the highest
scores of all the ATA drives tested, was selected as the test drive
for Bus Mastering effects.
The benchmark figures indicate that installing the PIIX drivers results
in a slight but negligible hit (1.2% decrease) in WinBench's Business
Disk WinMark 98 score. The reverse is true when looking at the High-End
WinMark 98, there is a slight increase (2.8%). Disk/Read Random Access
and Transfer Rates remained the same. However, Disk/Read CPU Utilization
jumps an astounding 677%, up to 35.2% from 5.2%.
It is often said by SCSI proponents that ATA drives cannot touch the
miniscule CPU utilization posted by SCSI drives. This is simply untrue.
The perception is likely perpetuated simply because most run tests using
PIIX vs SCSI instead of OSR2 vs SCSI. While SCSI still holds benefits
over ATA devices in many areas, CPU utilization is not one of them.
Its important to note that to achieve the low CPU utilization scores
(along with an increase in the sequential data transfer rates) that
the drive's individual DMA box must be enabled. Leaving the box unchecked
will result in CPU Utilization scores in the 80-90% range.