Vol. 20, No. 7 • July/August 2000

• FEATURE • PC Systems for the Connected Desktop Networking in the 21st Century by Eric Flower

The Past Year in Review
Last year we began by revisiting James Martin’s 1990 prediction that, “The future belongs to organizations skilled in providing solutions for the desktop-computing environment, the database infrastructure and the network infrastructure” (PC Week, May 14, 1990, p. 74). How right he was. On March 27, 2000, Cisco Systems (network infrastructure equipment manufacturer) surpassed Microsoft (desktop-computing environment manufacturer) as the most valuable company in the world. On that date, number-one Cisco had a corporate capitalization of $555.4 billion, number-two Microsoft was valued at $541.6 billion, and number-six Oracle (database infrastructure) was valued at $249 billion. Although these capitalization figures date from before the mid-April 2000 stock market storms and the government threat to split Microsoft in two, these companies remain among the largest.

Forecasts of even broader future trends may be found in Visions: How Science Will Revolutionize the 21st Century (Bantam Books, 1998) by Michio Kaku. Starting from the simple methodology of asking his fellow scientists what they are doing in their research labs, Kaku goes on to outline revolutions in computing, molecular biology, and quantum physics. Look for a ubiquitous networking infrastructure that makes ubiquitous computing possible, “invisible” (embedded) computers, artificial intelligence that works, and eventually, machines that think.

“Prices in all market segments were either stable or up slightly due to RAM and CPU shortages for a good part of the year.”

While it may be a decade or two before computers become so small as to be invisible, for years it has been a truism that we could expect computers to become faster, smaller, and cheaper. This year they became faster, but that’s about it. Whether that speed was of any value in everyday office work, or more properly in everyday Microsoft Office work, is questionable.

PCs did get faster, and we were treated to the spectacle of a race between Intel and AMD to announce, produce, and then help vendors sell systems based on their 1-Gigahertz (1,000 MHz) CPUs. AMD won the publicity race and bragging rights by a day. But if you want one of these systems, be prepared to pay a premium—the CPU alone will cost as much as some complete systems that offer only marginally lower performance in the MS Office environment. More importantly, AMD’s Athlon chip line is as fast, or nearly as fast, as Intel’s Pentium in systems built for business use. For the first time, AMD isn’t playing catch-up with industry leader Intel.

Prices in all market segments were either stable or up slightly due to RAM and CPU shortages for a good part of the year. These RAM shortages were caused by loss of capacity (or fear of loss of capacity) due to earthquake damage in Taiwan as well as to increased demand. There were shortages of some of the more powerful central processors during the year as well. Even large companies like Dell and Gateway could not get as many processors as they would have liked at the end of 1999. As any Economics 101 student knows, decrease in supply or an increase in demand (all other things being equal) will raise prices. This year we saw both situations. Fortunately for buyers, RAM prices began to return to more normal levels, CPU shortages began to disappear, and prices began to come back down in the spring of 2000. In the quarter ending March 31, 2000, 30 million PCs were sold worldwide, up from 25 to 26 million in the same quarter last year.

Finally, we’ve all heard that the Department of Justice and the states party to its antitrust suit have asked that Microsoft be split into two companies: one to sell the Windows operating system, the other to sell application software. But how many of you have heard or read the following statement (definitely a candidate for the Microsoft quote of the year)? According to Andrew Brosseau, an analyst with S. G. Cowen who rates Microsoft a long-term buy, “When all is said and done, Microsoft is essentially a desktop software company, and in particular, a personal productivity software company rather than an enterprise software company like Sun or Oracle. Personal productivity is an old business. It’s got no more sex appeal than mainframe computing, and while not quite as mature, is nearly so” (Andrew Brosseau in “Microsoft’s Competition is Looking Attractive,” by Lawrence M. Fisher, The New York Times, April 30, 2000, National Edition, Money & Business Section, p. 12). Midway through the article in large bold type, this important question is highlighted: “Is a desktop-based model obsolete in a Web-based world?” It’s a critical question for Microsoft and for Intel, as well as for users of their products. Even now some users want greater bandwidth more than they want new software or a new system.
 

A Net-Based Environment
Our work has become increasingly, or perhaps more accurately, entirely Net-based. While you may define the range of your network however you like, today we all live in a networked infrastructure that is more or less secure, reliable, fast, and ubiquitous. Client PCs view or work with content provided over the network, workstations create rich multimedia files for network delivery, and servers distribute those files. Where you are in this network hierarchy will determine the tasks you have to do in your job, which in turn will determine what kind of software, operating system, and computer hardware you’ll need to accomplish those tasks.

For most of us, Microsoft Office is where we live, and its foundation is either Windows 95, 98, NT, or 2000. Microsoft’s soon-to-be-released Windows ME (Millennium Edition) is an attempt to clearly define, split, or fragment (depending on your view) the Microsoft operating system market between consumer/home and business/office systems. It is the successor to Windows 98 SE, and it will be the last operating system in the Windows 95/98 line. To emphasize its consumer/home roots, it lacks some key inter-networking components found in Windows 95 and Windows 98.

Microsoft wants networked office workers to upgrade to Windows 2000 Professional. Key features include reliability, remote manageability, security, Plug and Play, and a familiar interface. According to Microsoft, minimum Windows 2000 hardware would include a 133-MHz Pentium with 64 MB RAM, but “more memory generally improves responsiveness.” Microsoft Office, where the work gets done, has even more modest system requirements. Minimum hardware is only a 75-MHz Pentium for the Professional version or a 166-MHz Pentium for the Premium version. Thirty-two megabytes of RAM are required, 64 MB for NT. Keep up with Windows 2000 at http://www.microsoft.com/windows2000/default.asp and with Office 2000 at http://www.microsoft.com/office/features/ofc2000tour/Sysreqs.htm.
 

The PC System Design Guide
Anyone who has used Windows 2000 to run the Microsoft Office Suite knows these minimum recommendations are a prescription for complaint. If the standard is, “Will it run?” the answer is yes. If the standard is, “Will it run acceptably for most users?” the answer is an emphatic no. And while Microsoft wants to advertise that the software will run on as large a number of legacy PCs as possible, these minimum recommendations do most users a disservice.

CNET News.com staff writer Michael Kanellos captured the Windows 2000 power controversy nicely in his article “Intel: Faster Chips Needed to Run Windows 2000,” posted February 16, 2000 at http://news.cnet.com/news/0-1003-200-1551163.html. Kanellos noted that even Microsoft’s partner in the “Wintel” alliance, Intel, disagrees when it comes to the processor power needed to run Windows 2000. “Most of today’s newest software will run on previous versions of our processors, but for best performance—including Windows 2000—you’ll have the best experience or performance on a Pentium III or Pentium III Xeon processor,” Intel spokesperson Christine Chartier said. Craig Beilinson, Microsoft’s lead product manager for Windows 2000 responded, “According to independent tests ... Windows 2000 Professional is up to 39 percent faster than Windows 95, 30 percent faster than Windows 98 and up to 24 percent faster than Windows NT Workstation 4.0 using current hardware configured with 64 MB of memory or higher. We’re thrilled to be providing faster performance with Windows 2000 Professional across the board.”

The truth is probably somewhere in the middle, since Intel wants to sell new systems built around its processors while Microsoft wants to sell as many copies of Windows 2000 as possible. Two disinterested parties side with Intel. The Gartner Group recommends at least a Pentium II-based system with 128 MB of memory. “Competitive Systems Analysts, another research group, recently conducted a study testing various Windows 2000 scenarios across both single- and dual-processor PCs from Dell. They concluded that PCs will need the equivalent of a 200-MHz processor upgrade to run Windows 2000 at the same performance level as Windows NT 4.” (All of the quotes on this controversy are from the Kanellos article. It’s worth a look.)

What this suggests is that Microsoft might be better off featuring the hardware recommendations of the PC 99 System Design Guide. The guide, co-authored by Microsoft and Intel with input from the rest of the industry, “... provide(s) guidelines for designing PC systems that will result in an optimal user experience with typical Windows-based applications running under either Microsoft Windows 98 or Windows NT Workstation operating systems.” Surely an optimal experience is preferable to a minimal experience. The 99 Guide calls for the following hardware (among other things) to be found in a typical office desktop computer:

• 300 MHz or greater Pentium or equivalent CPU
• 128-KB Level 2 cache
• 64 MB RAM

• 667 MHz or greater processor
• 128-KB cache
• 64 MB RAM; 128 MB RAM for Windows 2000 systems
• USB required; PCI, SCSI optional; ISA prohibited
• CD or DVD (DVD preferred)

• Specific types of systems that will run either Windows 2000 or Windows Millennium Edition operating systems
• Systems adopting the goals of the Easy PC Initiative, including requirements for legacy-free PC systems
• Devices supported under Windows ME and Windows 2000, including graphics and video device capabilities, digital media, storage, networking and communications, and other devices

Here we have a golden opportunity. Microsoft and Intel have described the requirements for an optimal computing experience in 2001 and the years following, but the critical parts of the systems they describe are available today at a reasonable cost. The PC 2001 System Design Guide draft has given us an outline that we can use now to specify systems that won’t be obsolete a year or 18 months after we buy them.
 

Intel and AMD Product Lines
There are three market segments in the desktop arena—value, mainstream, and performance/enthusiast. Intel and AMD compete in each. We’ll ignore free and inexpensive PCs because they typically are underpowered and frequently turn out to be neither free nor inexpensive. (See “Free PCs: The Price You Really Pay,” by Chris Lindquist in the April 2000 issue of PC World.)

The value market consists of systems selling for less than $1,000 built around the Intel Celeron or AMD K6-2 with 3DNow! Technology processors. AMD recently announced another addition to this market—the Duron line. The mainstream market ranges from $1,300 to $1,700, more or less, and is built around the Intel Pentium III and the AMD Athlon lines. The performance/enthusiast market is concerned only with performance, and cost is of little consequence in the price/performance ratio. These machines include the very fastest Pentium IIIs, Pentium III Xeons, and AMD Athlons. The systems we are most interested in are the value PCs built around the Intel Celeron or AMD K6-2 processors and the mainstream PCs built around the Pentium III and Athlon processors.

In late April, Celeron processors were available at 600-, 566-, 533-, 500-, and 466-MHz speeds with 16-KB instruction and 16-KB data Level 1 (L1) cache (32 KB total) and 128 KB of integrated on-die Level 2 (L2) cache. Cache is high-speed memory that gives the CPU more rapid access to program instructions and to data. Level 1 cache is cache memory built into the processor. Level 2 cache may be part of the processor, in which case it is called “on-die,” “on core,” or “integrated,” or it may be external and physically separate. In general, more cache is better and more on-die cache is better yet. The competing AMD K6-2 with 3DNow! Technology ran at 450, 475, 500, 533, and 550 MHz with a 32-KB instruction and 32-KB data L1 cache (64 KB total) and support for 512 KB of L2 cache. The new AMD Duron processors will feature integrated, full-speed L2 cache and will be targeted at the value desktop market.

When the Pentium III was released, it was touted as the first Intel processor designed specifically to power a new Internet experience. The Pentium III includes 70 instructions to enhance imaging, 3-D, streaming audio and video, and speech recognition applications beyond MMX technology. There are now two Pentium III families; be sure you know which you are buying. The older Katmai design runs at 450, 500, 550, and 600 MHz and is built on 0.25-micron technology. It includes a 16-KB instruction and 16-KB data L1 cache (32 KB total) and an external 512-KB half processor speed Level 2 cache. The newer design Coppermine CPUs run at 600-MHz to 1-GHz speeds and are built on 0.18-micron technology. They include 16-KB instruction and 16-KB data L1 cache (32 KB total) and 256 KB of integrated on-die L2 cache. A chart showing the relative performance of the processors in the Pentium III line is at http://www.intel.com/procs/perf/icomp/brief/brief.htm.

The oddly named Athlon is AMD’s product for the mainstream desktop market. The Athlon runs at 600, 650, 700, 750, 800, 850, 900, and 950 MHz and 1GHz. It includes 128 KB of L1 cache and support for 512 KB to 8 MB of L2 cache.

Two articles appearing in the January and April 2000 issues of Computer Shopper comprise the best summary report on the current state of the processor. In January (p. 79) columnist Michael Slater, who is also executive editor of Microprocessor Report, covered Intel’s line in “Sorting Out the Pentium III Proliferation.” Read it at http://www.zdnet.com/computershopper/stories/reviews/0,7171,2403907,00.html. In April (p. 80) he wrote “A Plethora of PC Processors” covering the rest of the market. That article is at http://www.zdnet.com/computershopper/stories/reviews/0,7171,2454555,00.html.

Look for information on current and future Intel processors at http://www.intel.com/design/processor/index.htm. See the AMD home page at http://www.amd.com for links to information about its processors and future plans.

The Office Desktop Market
In Tables 1 and 2, Axis Systems, Inc. is used as a proxy for the value and mainstream office markets. Axis is a second-tier direct vendor with a full product line that has done quite well lately. Its products appeared in the Top 10 Power, Midrange, and Budget PC lists in the April 2000 issue of PC World. (Note this is not an endorsement of Axis Systems or its products; rather it was selected because it has Intel and AMD-based lines at every price point.) Comparable systems from first-tier companies may be a little or a lot more expensive while similar systems from no-name, third-tier, white-box companies will be a little cheaper.

Table 1 lists the cost and key features of two of the many multimedia Intel-based systems offered by Axis. Table 2 lists the cost and key features of two of the many multimedia AMD-based systems offered by Axis Systems. All prices include an accelerated graphics port (AGP) video adapter, a 17-inch monitor, a sound card, and speakers. I prepared Tables 1 and 2 on May 1, 2000, from data I collected on Axis Systems’ home page (http://www.axissys.com). Check there for up-to-date configurations and prices.

Don’t forget to add shipping to these prices. You’ll also need application software, system and anti-virus utilities, an uninterruptible power supply (UPS), and a network card. Other purchasing considerations include product reliability and vendor service.
 

Tomorrow’s Specs Today?

Table 3 is taken from page three of the Executive Summary of the Industry Review Draft version 0.7 of the PC 2001 System Design Guide dated March 30, 2000. It provides a summary “of the requirements for the generic PC system, often a desktop computer, with variations for mobile and high performance systems, often called workstations. Mobile computers and workstations are based on the PC system requirements.”

Perhaps what’s most exciting about Table 3 is that requirements for 2001 may be met today at reasonable cost. Compare the requirements in Table 3 with the mainstream Intel and Athlon-based systems offered by Axis Systems in Tables 1 and 2. For $1,518 (Intel-based) or for $1,488 (AMD-based), we can purchase systems “that will result in an optimal user experience with typical Windows-based applications running under either the Microsoft Windows Millennium Edition or Windows 2000 Professional or later operating systems.” In other words, we are buying a system that will perform well today and into the foreseeable future if we follow the specifications outlined in Table 3. Tracking this table and others like it as the Design Guide goes from draft to final document will be one key to building systems that will have relatively long, useful lives.

If you’re on a budget or work with a shorter time horizon, you will be well-served by keeping the recommendations in the PC 99 System Design Guide and today’s marketplace realities in mind. The fact is that almost any currently available system will meet or exceed the recommended 300-MHz CPU with 128 KB of L2 cache and 64 MB of RAM called for in the PC 99 Guide. The critical improvement would be to add more memory to a total of 128 MB of RAM. In short, those on a budget should buy the fastest Intel Celeron or AMD K6-2-based system with as much RAM as they can afford.

If you have a longer time horizon, you may want to spend more and take this opportunity to specify systems that may add another 12 to 18 months to their useful lives for only marginally more money. We know what those systems likely will require from the early drafts of the 2001 PC System Design Guide. If you have the funds, buy systems with 667 MHz or faster AMD Athlon or the newer Intel Pentium III processors, and be sure they have at least 128 MB of RAM. But stay away from the bleeding edge; the better buys remain a step or two back.
 

Looking at What’s to Come
AMD plans to introduce differentiated versions of the Athlon processor based on three new processor cores called Thunderbird, Spitfire, and Mustang. Thunderbird will be a high-performance version of the Athlon processor with integrated, full-speed L2 cache targeted at the performance workstation and desktop markets. Spitfire will be a value version of the Athlon processor with 128 KB of integrated, full-speed L2 cache targeted at the value desktop market. (The Spitfire code name is the Duron product line.) Mustang will be an enhanced version of Thunderbird featuring a reduced core size, lower power requirements, and large, full-speed, on-die L2 cache. Multiple versions are planned for the high-performance server/workstation, value and high-performance desktop, and mobile markets. See http://www1.amd.com/products/cpg/athlon/result/1,1651,788,00.html for a full report.

Intel has published an “Intel Processor Roadmap” (http://www.intel.com/design/processor/future/roadmap.htm) in which the company describes how it will increase the speed of its performance and mainstream Pentium III-based desktops to 700 MHz and above, and how the value desktops will be based on 500-plus-MHz Celeron processors. The Roadmap page also links to descriptions of major future developments in Intel 32-bit architecture (also called IA-32 architecture).

Intel’s next push will be into 64-bit computing. For years we’ve read about the Merced chip, and it may become a reality under its new name, Itanium. “Scheduled for mid-2000, Intel’s Itanium brand processor will deliver world class performance, scalability, availability, reliability, and full IA-32 binary compatibility for high-end servers and workstations. The IA-64 processor family will co-exist with future IA-32 processors to enable a complete range of Intel Architecture solutions.” For more about Itanium, see http://developer.intel.com/design/ia-64.
 

We Have a Clearer Vision
It’s not often that we are presented with a clear vision of the future, but in the PC 2001 System Design Guide drafts, we have one. If we follow the drafts through to final publication and then on through the inevitable addenda, we’ll have a good idea of where industry leaders Microsoft and Intel want to take us. If we couple that knowledge with an understanding of our organization’s mission and how work is accomplished in the organization, system specification should be relatively  easy. Predicting  computing trends and directions a decade or two ahead is difficult; predicting what the connected office desktop PC system will look like later this year and next is not. And more to the point, some users may want or need faster network connections, not new software or new systems.

Forecasting the longer term is more challenging. In January 2000 Bill Gates stepped down as Microsoft CEO and became chairman and chief software architect to concentrate on strategic technology and corporate direction. Number-two man Steve Ballmer took over as CEO. “‘In the coming years, Microsoft will focus on assembling an “Internet-based platform” set of applications, tools, and services that leverage and maximize Web use,’ the duo said. ‘The initiative will revolve around sharing every possible data type on the Internet,’ Ballmer said” (“Bill Gates Steps Aside as Microsoft’s CEO,” Bob Trott and Michael Lattig, January 17, 2000, http://www.infoworld.com/articles/hn/xml/00/01/17/000117hnmci.xml). (If this strategy sounds familiar, you’ve heard it before; Sun Microsystems, Inc. has said “The Network is the Computer” since 1982. See Sun’s corporate vision at http://www.sun.com/corporateoverview/who/vision.html.)

In mid-April, Intel announced three new switches to provide gigabit Ethernet data transfer over existing Category 5 cable. In late April, Intel chairman Andy Grove was “characterizing Intel’s current state as being ‘in the middle of a fairly sizable strategic transformation.’ Intel essentially intends to put its chips in every step of the Internet, from PCs to data centers to handheld devices to network communications equipment. ‘We have realized that the Internet runs on silicon, and our strategy is to follow the bits,’ Grove said” (“Intel Officials Detail Corporate Strategies,” Dan Briody, April 27, 2000, http://www.infoworld.com/articles/pi/xml/00/04/27/000427piintelstrat.xml). With gigabit Ethernet over Category 5 cable a reality and industry leaders concentrating on Net-based strategies, the first of Michio Kaku’s visions for the 21st century—ubiquitous networking supporting ubiquitous computing—has been achieved.
 
 

Eric Flower is the library director at the University of Hawaii–West Oahu. His e-mail address is flower@hawaii.edu.