Printer Friendly


Their Emerging Role As A Major Enhancement To System Performance

With the emergence of performance-demanding PC applications and functions, today's PC users are highly sensitive to overall system speed and performance. So much, in fact, that PC users evaluate each component within the system by the performance it delivers--the processor is known for its clock speed, the modem and CD-ROM by their transfer rates, and the motherboard for its bus rate. This is true for all components except the hard drive, where capacity remains the primary metric used when judging how the component measures up.

But the hard drive is just as critical an element in system speed. Any operation that involves moving large amounts of information on or off the disk will reveal the importance of a high-performance hard drive. To a large extent, the hard drive influences how fast the OS boots, how quickly applications launch, and the speed of loading large data or graphics files.


System responsiveness is generally defined as the speed at which the system receives and then completes a command or task (i.e., launching an application or booting an OS). Typically, system performance is thought to be a function of the processor, the faster the processor, the faster, more responsive the PC. But today's hard drive offerings, especially the new 7200RPM drives, provide performance benefits comparable to the benefits of a faster processor.

In system-level testing performed at Maxtor, two different upgrades to a Pentium II system were compared. In one upgrade, the system's 333MHz processor was replaced with a 400MHz processor. In the other upgrade, the original 10GB 5400RPM drive was replaced with a 10GB 7200RPM drive, while the processor remained at 333MHz. While the faster processor clearly provided a performance boost, when comparing the two systems' performance in real-world applications, the 7200RPM-equipped system provided even more performance benefit.

And the faster drive can often provide a performance advantage at a very reasonable cost when compared with the cost of other upgrades. We compared system performance versus processor price for the range of Intel Pentium II processors. A system-level benchmark, Business Winstone 98, was employed. The analysis indicated that on average a 1% improvement in system performance added $7.90 to the processor price. Using the same system-level benchmark, we then compared performance with drive spin speed. In this case, upgrading from a 10GB 5400RPM hard drive to a 10GB 7200RPM hard drive added system performance at a cost of only $0.65 per 1% of improvement. Even when measured at the system-level, rather than as raw data transfer rates, the faster hard drive provided meaningful and cost-effective performance improvements.


It may seem obvious to say that system performance does matter. More speed, after all, translates to a more responsive system. This means less waiting and greater productivity. But there are other reasons to be concerned about performance, reasons which are evolving as system capabilities evolve.

Certainly, one element is the increasing size and complexity of software: Windows 95 occupies 180MB to 195MB of disk space, while Windows 98 occupies between 195MB to 230MB of disk space for typical installations. Windows 2000 is expected to be significantly larger. As the operating system and associated applications packages grow, more speed is needed just to remain at a given system performance level. For system integrators concerned about Total Cost of Ownership, a higher performance system will ultimately pay dividends by enjoying a longer useful life.

In addition to larger OS and applications packages, we are seeing the overall demands on PC bandwidth increasing. More background operations including e-mail, virus protection, and entertainment are demanding more from the PC. Intel has proposed a model called "Constant Computing," wherein increasing the system bandwidth is a mechanism to boost the system's utility, increase reliability, and reduce overall maintenance costs. As we go forward, every part of the system will be called upon to improve throughput.


High-performance 7200RPM drives improve performance in several ways. The first is evident in the WinBench measurement, a popular means of evaluating system performance. In WinBench testing, a variety of popular applications are simulated and overall execution time is measured. A weighting factor is applied to the execution time for each application package. Measurements have shown that 7200RPM drives improve WinBench scores approximately 20% over equivalent capacity 5400RPM drives. Considering that WinBench is a measure of overall system throughput, as opposed to just drive throughput, this is an impressive result.

WinBench is a good tool for arriving at a summary number, but that number is just that: a summary. In evaluating the WinBench results, it's important to keep the actual intended application in mind. In particular, consider how a drive's basic performance elements will contribute to system performance in the application. These basic elements are seek time, rotational speed, channel rate, and caching capability.


In addition to these elements, newly introduced design features within hard drives today make them even more of a performance element. Some architectural features, once associated with server technology and inherently cost prohibitive in the desktop environment, are beginning to appear in ATA desktop disk drives.

Maxtor's DualWave multiprocessor architecture, as an example, features two processors: a digital signal processor (DSP) that controls disk functions (i.e. servo, reads and writes, ECC, etc.) and a second processor, a proprietary RISC controller that is dedicated to processing host commands. Traditionally, desktop ATA disk drives have used one microprocessor to handle seeking, error correction, host command processing, caching, reading, writing, and data flow. Each of these tasks requires a significant amount of processing power, which may decrease performance. By adding an independent host processor, the controller is constantly available to receive and execute host and buffer commands, thereby using the host's time more efficiently. Testing of this dual processor architecture has indicated a dramatic improvement in effective performance over previous generations of disk drive controllers, reducing host command processing time by 90%.


The insatiable need for performance in high-end desktop systems and workstations requires new technologies and architectures in order to meet the demand. Today's 7200RPM drives in both benchmark and real world testing offer users tangible increases in system throughput. Improved electronics architectures also enhance the speed at which the drive is able to execute commands, providing PC consumers with additional system responsiveness.

The hard drive's contribution to overall system performance has typically been undervalued. But with the advent of high performance desktop drives, the hard drive has now emerged as a cost-effective contributor to substantial system performance improvement.

Jon Toor is the director of product marketing at Maxtor Corporation (Milpitas, CA).


This article makes an interesting argument for the performance increases that can be achieved through the upgrade to or initial selection of a faster hard drive. Clearly, a 7200RPM, or even better, a 10,000RPM drive will enable loading and saving of data and applications more rapidly than a 5400RPM drive.

The question of drive interface--UltraATA33 or UltraATA66, one of the flavors of SCSI, or Fibre Channel can also have an impact on the rate of data transfer between drive and processor, although today's drives cannot match the sustained burst speeds of the user interface. That is, a single drive cannot transfer a data stream at the 40Mbit/sec of Ultra2 SCSI, or the 66Mbit/sec of ULTRA ATA-66. In large servers with multiple drives in use, the extra bus bandwidth becomes more important than on systems with fewer spindles.

UltraATA66, which is becoming increasingly common on new ATA, drives, may finally rival UltraWide SCSI and possibly even challenge Wide Ultra2 for data transfer performance. For certain metrics that focus primarily on transferring data between drive and processor, the faster drives and higher performance data bus can clearly provide a significant level of improvement.

Technology improvements noted by Maxtor in its own drives can improve drive performance. It should be noted that other drive manufacturers are making changes in their drives that are also designed to improve performance and rellability.

As noted by Editor-at-Large Hal Glatzer on page 6 in the August issue of CTR, adding system RAM may be the most cost-effective upgrade available. It may not only be the most cost-effective, the effects may also have more impact on data movement and processing than adding a faster hard drive or a higher performance CPU. A system with a small amount of RAM will see more processing speed advantages from a faster hard drive than a system with a lot of RAM. The reason for this is fairly basic.

As Hal noted, RAM is the computer's 'scratchpad. 'When it's processing data, the processor moves the data to RAM--electronic memory, where it's stored while the processor performs its required tasks. If the memory requirements exceed the amount of RAM available, the data is written to virtual memory located on the hard drive. Clearly, swapping data onto and off of a fast hard drive will result in faster processing, since the latency between the processor event that moves the data onto virtual memory and the latency between the time the processor requests data from virtual memory and the time it receives it are critically important--the faster the drive, the shorter the latency, and the faster the processing.

Just as clearly, however, is the incontestable fact that data in RAM can be written or read orders of magnitude faster than it can be from even the fastest hard drive. A system used for many computing-intensive tasks can very likely benefit more from the addition of RAM (often available for less than $1 a megabyte), than from upgrading to a faster, larger hard disk drive. Conversely, systems that do a significant amount of file manipulations with data read from and written back to the hard drive; or with applications frequently loaded and closed, may derive a lot of benefit from a faster, larger, frequently defragmented hard drive.

Ideally, the best upgrade may involve both a larger, faster hard drive, and a good hunk of high speed RAM. Upgrading to a faster processor, while not necessarily the best upgrade from a price-performance standpoint, can also provide obvious advantages and would enable the server or desktop computer to further leverage the advantages achieved by adding RAM and a higher performance hard drive.

Mark Brownstein
COPYRIGHT 1999 West World Productions, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1999, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Technology Information
Author:TOOR, JON
Publication:Computer Technology Review
Date:Aug 1, 1999
Previous Article:Managing Storage On The San With Jini.
Next Article:Network Video Computing.

Related Articles
Does Size Really Matter Or Is It Performance That Counts?
Who's First? Does It Matter?
Packing The SAN.
Maxtor's 'Diamonds' Catch Google's Eye.
Serial attached SCSI or serial ATA hard disk drives: how to choose one or the other--or both. (Tape/Disk/Optical Storage).
Are you being served? Smaller, faster and more affordable, today's servers can help transform a district's network center.
Hitachi boosts speed for industrial-strength notebook hard drive.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters