
Click for Mac Memory Specials!

|
in a Beige/Platinum G3 By Robert Bradley Published: 10/25/2000
The original system was a Platinum G3/266 Desktop Macintosh with 224 MB of memory,
a 5 MB/sec SCSI bus, a 16.6 MB/sec IDE bus, a CD-ROM, and a Zip drive. The SCSI
bus has the Zip drive, a tape drive, a 2 GB hard drive, and a scanner. The IDE
bus has the original 4 GB internal IDE drive. Two of the three PCI slots have
an Ethernet card and a 5 port USB card. (The SCSI 2 GB drive will be removed,
and the 4 GB IDE will be kept, partly to provide a backup bootable system. )
The Mac OS was upgraded to 9.0.4.
A Sonnett ATA/66 interface card was installed in the remaining PCI slot, and
an internal IBM 75GXP 46 GB, 7200 RPM drive was installed in the bay above the
Zip drive. The only additional purchase required was a sled for the drive. The
75GXP drive is ATA/100.
The Sonnett card came with a good descriptive installation manual with instructions
for a Desktop Mac, a Tower Mac, and the Blue & White/G4 graphite computers.
I downloaded the installation manual for the 75GXP from the IBM site; although
this was for a PC installation, it was still useful information. (Note: For Beige G3 owners replacing the original hard drive, see my 1998 Illustrated Beige G3 MT HD Upgrade Guide - Mike)
The installation went very smoothly. The formatting was done with Disk Setup
1.8.1. The Disk Setup Help said that the drive would run faster if divided into
partions. I created the maximum number of 8 partitions ranging from 3 GB to
10 GB. Having that many partitions makes backups easier to segment/manage.
The performance testing was done on the system partitions of the SCSI 2 GB
(SYS), the IDE 4 GB (G3-HD1), and the new 75GXP (G3-HD). I thought it would
be interesting to compare the ATA/66 bus with the ATA 16.5 bus and SCSI 5 MB/sec
bus to appreciate the tradeoff of add the Sonnett card.
I was surprised by the variation of performance among the various partitions
on the new drive. The speed decreased as the partition number increased. (Is
this related to track circumference?) This decrease was not monotonic, presumably
due to the variation in the sizes of the partitions. Steady state read speed
was between 22.41 and 36.07 Mb/sec, and write speed was between 17.79 and 31.97
Mb/sec.
The three partitions which were measured are all bootable partitions, but they
are on three separate disks: SYS (old SCSI), G3-HD1 (old IDE), and G3-HD (ATA/66
bus).
![]() At the low speed of the old SCSI bus, plus the scale of the plot, the write graph exactly overlays the read graph; the read/write speeds are identical. The bus is nominally 5 MB/sec.
![]()
The original IDE bus is nominally 16.5 MB/sec, and this partition is on the original IDE internal drive.
![]()
This is the first partition on the IBM 75GXP drive, and it had the fastest read and write times of all the partitions for this drive.
Real World MeasurementsFile and folder duplication tests were run for a 20 MB file and a 81 MB folder which contained about 1300 files. There may be some bias because, even formatted with Mac extended format, the minimum size file varied among the hard drives. SYS had files as small as 1K, for G3-HD1 the smallest file was 2K, and for G3-HD the smallest file was 4K. As I recall this has to do with the maximum number of files for HFS+; so the larger disk would have the largest minimum size file.
(Note: Remember the Beige G3's onboard SCSI is only a 5MB/Sec max interface. The onboard IDE is ATA/3 (16.6MB/sec max interface).-Mike)
All times in the table are in seconds.
There is some bias in the PhotoShop measurements because the files were on the test partition, but PhotoShop itself was on a different partition. There is also a bias caused by the decision as to where to allocate the scratch disk for PhotoSop. Allocating it to the same disk/partition as the file being edited slows down PhotoShop; allocating it to a different disk/partition puts less load on the disk/partition under test. For these tests, scratch disk area for PhotoShop was allocated to the partition being tested since that seemed to be the truest test of the drive.
The PhotoShop manual says that since PhotoShop copies its RAM image to scratch disk, it must have at least as much scratch disk as allocated RAM; it recommends 3 to 5 times the size of the file plus 10 MB. For the 62 MB file, that translates to 196 to 320 MB of scratch disk! However on one of the test runs, when I allocated 220 MB of scratch, it said that it did not have enough. If I gave it all of the scratch area it wanted for the 62 MB file, it took 306 MB (with PhotoShop allocated 120 MB of RAM). I ran tests with scratch assigned to the RAM disk, and tests with more RAM allocated to PhotoShop. RAM disk gained at most a 20% improvement in time. The most dramatic improvement was to give PhotoShop more RAM. Since the RAM disk takes memory that PhotoShop could use, it doesn't make much sense to use RAM disk.
The following two tables show PhotoShop results for two cases: (1) PhotoShop allocated 40 MB of RAM; and (2) PhotoShop allocated 120 MB of RAM. All times are in seconds. (Note: A general rule for optimal PShop RAM allocation is 4x to 5x the size of the image you will be working on. You can see if the swap file is being used in Photoshop by selecting "effeciency" as the option for the lower left edge of the image window. Less than 100% means the swap file is being using-Mike)
-Robert Bradley (Note the dramatic gains seen from allocating more RAM to Photoshop-Mike)
If you've installed/upgraded a Hard Drive, CD/DVD/DVD-RAM/Tape drive, or removable media drive in your Mac - please post an entry in the Mac Drive Compatibility Database. Include any performance comparison or compatibility info with your report.
Related Articles: Links to other IDE drive related articles. For general info/upgrade articles on Beige/Platinum G3s - see the Systems page and G3-ZONE.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
= Other Site Topic Areas =
|
||||||||||||||||||||||||||||||||||||||||||||||||||||
Your Source for the best in CPU/SCSI/VIDEO card reviews, daily news, and more! Disclaimer: Users must read and are bound by the Site Terms & Conditions of Use. | ||||||||||||||||||||||||||||||||||||||||||||||||||||