Maxtor 91000D8 DiamondMax Plus 250010 GB Ultra-ATA Hard Disk Drive
Robert Bruce Thompson
|Thompson and I are doing a book for O'Reilly
&; Associates on "good enough hardware." Stay tuned. Thompson is the author
of several O'Reilly books. I have never seen a bad O'Reilly book, which is interesting.
Previous reports by Thompson:
Maxtor 91000D8 DiamondMax Plus 250010 GB Ultra-ATA Hard Disk Drive
February 15, 1999
Overall Grade: A+
Big, fast, cheap, quiet, and reliable. There's not much more you can ask from a disk drive. The Maxtor DiamondMax Plus 2500 10 GB, 9.5 ms, 7200 RPM disk drive is the fastest ATA drive we have ever tested. Its combination of large size, high speed, and low cost makes it the ideal drive for a stand-alone PC or network client, and a more than acceptable choice for small servers.
This is a quiet drive, which is particularly surprising given its 7,200 RPM rotation speed. Ordinarily, drives with high rotation speeds produce an audible high-pitched whine even when running at idle. Even with the cover off the case, however, this drive makes no more noise than typical 5,400 RPM drives. Maxtor rates it at 37 DBa average at idle, which subjectively appears to be true. Even seeks, although they are audible, produce no more than the subdued clatter typical of slower drives.
Although it has been reported elsewhere that this drive runs hot, we found no evidence of this during extensive testing. After running for more than a week in an internal 3.5" drive bay without a drive cooler, this drive was barely warm to the touch. Using other evaluation units of the same model yielded the same results. We can only conclude that reports of this drive running hot were based on a defective sample.
The evaluation Maxtor drives arrived in OEM packaging, multiple drives in a plain cardboard box with styrofoam separators. Maxtor did, however, include all components that ship with the retail package, including (a) a MaxBlast Plus Hard Drive Installation Utility diskette, (b) an IDE interface cable, (c) two 5.25" mounting brackets, (d) a package of mounting screws, and (e) an 11X17" folded installation guide. The installation guide illustrates jumper settings, describes how to install and configure the drive, and provides basic troubleshooting advice. We judge it adequate to allow even a novice to install and configure the drive successfully.
One minor quibble: although the geometry is labeled on the drive cover plate, jumper settings could be identified better. The only mention of jumper settings on the cover plate tells you to jumper J50 on for Master or off for Slave. Unfortunately, it is difficult to determine which is J50 by examining the drive itself. Although the circuit board appears to be labeled, surface mount components and solder points obscure those labels. Maxtor should emulate other drive manufacturers by including a graphic on the cover plate that illustrates jumper settings in relation to the IDE connector and the power supply connector.
Although we did not test it, Maxtor also manufactures a 3-platter, 7.5 GB version of this drive, the 90750D6, with otherwise identical specifications, whose street price as of 990211 is $197.
We installed the DiamondMax Plus 2500 91000D8 as the sole hard drive in our current Reference Platform. When we restarted the system, the BIOS automatically detected and configured the drive properly. Other than hard drive type and date/time, all CMOS Setup parameters were left at their default values.
We restarted the system again with the Windows NT 4.0 Server distribution Setup floppy in the drive and installed Windows NT Server 4.0 as a member server in the TTGNET domain, accepting all default settings. Setup reported the total drive capacity to be 8 GB, but this is an artifact of Windows NT and does not prevent you accessing the full amount of drive space after Windows NT is installed. We used Setup to create a 1 GB primary partition (actually, 1,028 MB) and selected the NTFS file system.
During Setup, when prompted to install the network adapter driver, we chose Have Disk and installed the drivers supplied with the LinkSys LNE100TX 100BaseT Ethernet card. We installed only TCP/IP transport, and configured the system to use DHCP to obtain its TCP/IP configuration information from a DHCP Server running on another system.
After Setup completed, we installed the Intel740 Video Accelerator drivers from the CD supplied with the EPoX EP-BXT motherboard, and then installed Service Pack 4 from the SP4 distribution CD. We then used Windows NT Disk Administrator to create an extended partition that comprised the full 8,511 MB of remaining space on the drive. [Windows NT reports the full drive capacity as 9,539 MB because it defines one MB as 1,048,576 bytes rather than 1,000,000. This translates to an actual drive capacity of 10,002,366,464 bytes, whence Maxtor's rated capacity of "10 GB".] We then created a logical E: volume within this extended partition, and formatted that volume as NTFS using the default Allocation Unit Size. Although using this relatively small system/boot partition may slightly improve the benchmark test results relative to those that would be produced on a partition of full drive size, we believe that this partitioning represents how a typical user is likely to configure the drive.
After restarting the system, we installed the Ziff-Davis WinBench 99 benchmark suite, installed the WinBench v1.1 update, and used Diskeeper Lite to defragment the C: volume. Before running the benchmark tests, we disabled the Always on top property for the Task Bar, and stopped the Protected Storage (Pstores.exe) service that is installed with Internet Explorer 4, as required by the benchmark test. We ran the disk-related benchmarks three times in this default configuration, and reported the average results from the three runs.
Windows NT does not install DMA (bus mastering) drivers by default, so we installed the Intel PIIX PCI Bus Master IDE Controller driver from the EPoX distribution CD and then re-ran the tests. As expected, the disk peformed much better using Ultra-ATA/33 transfers than the default PIO Mode 4 transfers. The data transfer rate tripled, and all other disk performance measures showed substantial increases, but the most dramatic difference was in CPU utilization. During the the PIO-4 transfers, the CPU was occupied 80.9% of the time handling the disk transfers. During the DMA-2 transfers, the CPU was occupied only 1.6% of the time. In other words, relative to DMA transfers, PIO mode transfers place about 50 times more load on the CPU, which might otherwise be doing useful work under a multitasking OS like Windows NT. The moral here is that you should install DMA drivers if your hardware supports them.
Copyright © 1999 by Robert Bruce Thompson. All Rights Reserved.
Send feedback to: firstname.lastname@example.org