Pondering Windows and the Future of Unix

By Terry Wohlers

Published in Vol. 19, No. 2, February 1996 issue of CGW
Copyright 1996 by Terry T. Wohlers

For many years, Unix has been the undisputed platform of choice for high-end CAD applications. High-end software programs from Computervision, EDS Unigraphics, HP, IBM, Intergraph, SDRC, and PTC all run on Unix.

Recently, however, a growing number of these companies have begun porting their programs to Windows NT and/or Windows 95. That trend, along with the emergence of several impressive, new Windows-only-based solid-modeling programs, have suddenly turned Windows into a viable alternative to Unix at the high end of CAD.

Why has this happened? Why didn't it happen before? And what does it all mean for Unix?

One reason this is only just happening now is that, until recently, only Unix offered an environment for 32-bit application development. That meant software developers could develop CAD applications that included features that wouldn't have been possible using DOS or earlier versions of Windows.

In addition, hardware manufacturers such as HP, Silicon Graphics, and Sun Microsystems combined their hardware technology with Unix to offer workstations that provided unbeatable processing power, network performance, and graphics performance. Advanced 3D CAD software, particularly solid-modeling software, needed all of what these workstations offered, and more.

But Unix always came with a downside. Even Unix CAD software vendors will admit it's been expensive to create, sell, and support individual versions of software for each variation of Unix. Because each workstation vendor uses a special version of Unix that's customized for their particular hardware, software vendors must create a slightly different version of their software for each workstation. At the same time, CAD software vendors have had to endure the cost of employing people who are familiar with each hardware platform and each variation of Unix that they support. And whether CAD software buyers know it or not, that's an expense they end up paying for one way or another.

With the arrival of Windows NT and Windows 95, Unix has begun losing its advantages. The newest Windows programs offer not only a 32-bit environment but a graphical user interface that provides a welcome relief to Unix's command line-driven approach. (Even though some Unix workstation vendors have tried to protect users from Unix by offering their own GUIs, it's an effort that's been only partially successful.)

Equally important, the new Windows environment offers the attraction of a huge user base. Already, Microsoft has shipped more than twice the copies of Windows 95 than all variations of Unix combined. Near the end of 1995, Microsoft estimated that more than 100 million Windows installations (all versions) were in place worldwide.

With these kinds of numbers, developers can spread their development costs over a much bigger customer base. Also, it's easier and less expensive for them to get programmers and technicians who can and want to develop and support Windows products than it is to find and acquire Unix talent. That availability of Windows expertise is only likely to improve over time as the functionality and popularity of Windows applications expand.

Hardware Power for Windows

Finally, accompanying the arrival of Windows NT and Windows 95 has been the arrival of hardware platforms that, for the first time, can truly begin to rival Unix workstations in processing, networking, and graphics power. Intergraph, for example, claims its 200MHz Pentium Pro TDZ graphics workstations running Windows NT can outperform SGI's Indy and Indigo2 workstations, even with SGI's impressive High Impact graphics option. Yet the Intergraph systems cost thousands less.

Likewise, Digital Equipment Corporation is aggressively targeting the Windows NT marketplace, maintaining that most of its new opportunities in the workstation business will come from Windows NT. As for performance, according to the SPEC integer and floating-point performance figures published in the November 1995 issue of the Computer-Aided Design Report newsletter, a DEC NT-based workstation configured with a 266MHz 21064A Alpha chip can offer 4.6 to 14 times the price/performance value as that of similarly equipped Unix workstations from HP, IBM, SGI, and Sun.

Of course, Unix supporters argue that Unix offers a more mature environment, one that has been tested and improved for many years. Moreover, workstation vendors have had years to optimize their hardware to take full advantage of Unix. Other platforms, they'll tell you, simply cannot match the performance and tools available for Unix applications.

Finally, Unix supporters will point to the fact that Unix offers a 64-bit address space, which is not presently available under Windows. This gives software developers the option of developing applications that solve highly complex problems, such as the simulation of global warming.

But while 64 bits may be required for some very special applications, a 32-bit address space is more than adequate for most CAD applications. By the time CAD users and CAD software vendors really feel the need for 64-bit technology, it may well be that a 64-bit version of Windows will be available.

As for the point about the maturity of Unix, it's a valid one. But it's not an advantage that necessarily outweighs the advantages that Windows offers, such as its ease of use and its ability to put CAD programs onto the same platform that engineers currently use for their general-purpose office work. If Windows NT and Windows 95 can keep companies from having to buy two computers for their engineers, it'll be an advantage that'll be hard to resist. The money they could save in equipment purchases, upgrades, technical support, space, and training would be significant.

Of course, before Windows can really take off, CAD vendors will need to do a better job of supporting the operating system. Although some of the newer packages, such as TriSpectives from 3D/Eye and SolidWorks 95 from SolidWorks, were developed from the ground up to run on Windows, most of the Unix packages that have been ported over to Windows were not changed considerably to conform to Windows standards.

If high-end CAD software developers want to succeed in the Windows market, they will need to create Windows products that look and behave like other Windows products. Until they do, customers will not embrace the Windows version of their products and the transition to the Windows environment will appear slow. Some may blame this slow migration on the Windows platform itself, although the real problem may be with the poor effort put forth by the software vendors.

A Long, Slow Decline for Unix

Inevitably, though, Windows will continue to gain momentum, and I fully expect Unix vendors to experience some hard times over the next several years. Windows has already begun to eat into the Unix CAD market, although not yet to a large extent. But the Unix CAD market will surely erode further as 32-bit Windows applications mature.

But does that mean Unix as a platform for CAD applications will die eventually? Quite possibly. There will come a point in time that software developers can no longer justify the cost of developing and supporting individual Unix ports. With such a small volume, it won't make economic sense. But this won't happen soon, and this is why: customers of Unix hardware and software have invested large amounts of money in their systems and in training their people. Those who were a part of the decision to buy Unix will be slow to admit that Windows might offer a price/performance advantage over an operating environment they helped create.

Coupled with this is the influence of the Unix gurus who thrive on the mere existence of Unix. They will be the last to give it up. Without Unix, they might be without a job and without an OS to which they have devoted much of their life. Industry consultant Joel Orr once said, "Learning Unix is not a task, it's a career." These people are evidence of that.

Many years into the future, the choice will be obvious to buyers. They can go with the flow and spend less and get more, or they can continue to support the added expense and hassles associated with Unix. Over time, they won't have a choice. For CAD applications, Unix will suffer a slow and painful death.

CGW contributing editor Terry Wohlers is an industry consultant focusing on CAD/CAM, rapid prototyping, and reverse engineering technologies and applications.

Copyright 1996 by Terry T. Wohlers