A small team of researchers turned an opportunity into an award-winning product, helping to launch a workstation that provides the functionality of much more expensive machines with the convenience of a PC.
Applications Demanding real-time 3-D graphics - such as the special effects and animation software used by film and video professionals or the computer-aided design (CAD) software beloved by electrical and mechanical engineers - have traditionally required specialized engineering workstations. Powered by reduced-instruction-set computer (RISC) microprocessors running the Unix® operating system, these machines are typically supplemented by sophisticated 3-D-graphics engines, called accelerators.
The computational intensive tasks for which workstations are designed seemed beyond the practical capabilities of PCs. Yet, as so often happens in technology, the synergistic possibilities of unrelated events can often lead to surprising outcomes. In the last year or so, two quite independent developments have combined to change the face of an important part of the computing landscape.
First, in an effort to bring down the cost of their products, workstation manufacturers began to adopt an input/output interface standard for Pentium® PCs known as the Peripheral Component Interconnect (PCI). That meant, in theory at least, that the new graphics accelerators developed for RISC-Unix workstations could also plug into ordinary PCs.
At about the same time, Intel introduced the 32-bit Pentium Pro microprocessors, which - especially when coupled with Microsoft's 32-bit Windows NT® operating system - promised to boost PC performance in the range of low-end workstations.
Together, these developments added up to a new kind of PC. "The recipe for running high-performance applications and graphics on a Pentium processor, Windows NT platform became clear," says Randy Moulic, manager for professional workstations at IBM's Thomas J. Watson Research Center. "The pieces were there. We just needed to put them together."
But would customers be interested in a new kind of machine? Moulic and his team decided to find out. They soon learned that demand for workstations was poised to explode. "Most of our large customers - those with 2,000 or 3,000 Unix workstations - forecast their requirements to grow by five to 10 times by the year 2000," says team member Nick Dono.
Those same customers were also very concerned about their capacity to afford such growth. Quite aside from the $20,000 to $40,000 per workstation, there was the associated maintenance. "Their annual maintenance budget for 3,000 workstations is $10 million to $20 million or more," adds Bengt-Olaf Schneider. "The thought of multiplying that figure by five had them looking for another way."
What's more, every workstation had a PC sitting right next to it. "Our customers used a Unix machine to design with and a PC to document designs," says Moulic. Thus, the PC-based workstation that he and his team envisioned would not only cost much less - from $3,500 to $10,000 - it would replace two machines with one, slashing the costs of ownership.
Making A Case
The Research team began developing a business case, surveying potential customers to find out who would buy such a product and how many. They even calculated development and production costs. Then Moulic approached the Personal Systems Group (PSG) with the idea and an offer to help bring the machine to life. In the PC world, where new products hit the market every six months, Moulic knew that work had to begin at once without waiting for top-to-bottom consensus and fully committed funding.
Fortunately, Rod Adkins, general manager of client systems in the PSG, shared his sense of urgency, as did Satish Gupta, then Research vice president for technical planning and strategy, who has since moved to the PSG to manage the product development of the IntelliStation(TM) family. With support from Adkins and Gupta, Moulic and his team began a close collaboration with an equally dedicated group from the PSG.
As work got under way, the researchers found themselves donning both technical and marketing hats while shuttling between Yorktown and Raleigh, North Carolina, where their PSG collaborators were based. "It was a very close collaboration," says Greg Moore, the Intellistation development manager in PSG. "The Research group had the expertise in 3-D graphics, Unix, benchmarking and specifications, and we had the PC knowledge and the experience in making Intel-based products."
In addition to defining the overall architecture, the Research team played a crucial role in defining and developing the graphics subsystem, and working with vendors to troubleshoot bugs and wring out compatibility problems. The 3-D-graphics software they wrote, which served both to torture-test the design and to demonstrate the machine's exceptional performance, was first shown on a prototype IntelliStation at computer shows in summer 1996.
Recognizing IBM's achievement, Byte magazine tapped the IntelliStation for its Best of Spring Comdex award for 1997. The latest version, whose dual Pentium II® processors make it the most powerful IntelliStation to date, earned PC Week's Best of Comdex award at the fall show.
Today, the penetration of high-end PCs into the 3-D-graphics workstation world is well under way, and the competition is growing. The transformation to a PC-based world is not complete, however. There are a lot of Unix applications, especially on servers, notes Moore. "One of the big things Research is helping us with now is creating links between the Unix machines and the new workstation client systems, so that we can can continue the transition to a single platform for all applications."
In developing a vision for the new product, which became an entirely new brand for the PSG and IBM, Moulic and his team assumed roles that are becoming increasingly more important for all researchers. "Anticipating changes in the marketplace and ensuring that IBM has the products to meet them," says PSG's Gupta, "is essential to IBM's success."
Gil Bassak is a freelance technical journalist based in Ossining, New York.
More Information:
Graphic Displays of Power
Three-dimensional-graphics performance - the ability to rapidly create and display constantly changing 3-D images with accurate shading, texture and color - is what separates great workstations from merely good ones. But, unlike other performance characteristics, 3-D graphics has few quantitative benchmarks for comparing products. Rather, an experienced observer simply knows what to look for: image speed and quality.
And that's what visitors at the summer 1996 Siggraph saw. "People who knew the graphics field would comment that the performance was outstanding," recalls Chandra Narayanaswami, a member of the IntelliStation team at the Thomas J. Watson Research Center. "Many of the demonstrations instantly tell you 'Wow, that's fast.'"
But Clifford Pickover, a graphics expert on the Watson team, crafted the onscreen demonstrations to do more than just impress viewers. "The demonstrations also assess the workstation's overall speed as well as identify important graphics processing problems," he says.
Such graphical stress testing was vital for evaluating different vendors' graphics accelerator boards during development of the prototype. "Most conventional PC graphics demonstrations do flashy things on the screen, but they don't hammer the 3-D graphics parts of the machine the way Cliff's do," says Moulic. "We could quickly find color mismatches, color transition problems, as well as a host of other discrepancies requiring further investigation."
For more information: http://www.pc.ibm.com/us/intellistation/
