From Computer Business Review, a sister publication
Life at high-end workstation vendor Silicon Graphics (SGI) is all about image. Everything from the artwork of its annual report to the fine lines of its computer casings radiates an obsession with cool, visual design. That coolness extends to the corporate interior at its Mountain View, California headquarters, where subtle lighting and color choice highlight modern art – 3D art, of course, such as sculptures and holograms – that reach outward from building walls.
By Tom Foremski
Ask those who work there: it is like hanging out at the latest cyber cafe. SGI people get to use the best, most colorful computers (no boring beige here) and the audience for their products are movie-makers and Web site creators, biotech molecule manipulators and aircraft designers. Even the conference rooms have names like Terminator 2, reminding them of the company’s glitzy connections. If they are there at Christmas – likely, since SGI has one of the lowest staff turnover rates in Silicon Valley – they might get a nice little bonus, like an $800 Swiss-made Tag Heuer watch, which was the bauble of choice in 1995. In many ways, SGI chief executive Ed McCracken is the perfect embodiment of the company culture. The 52 year old electronics engineer, from humble Iowa farming stock, dresses in immaculate suits, collects Lotus sports cars and hobnobs with the likes of Bill Clinton and Hollywood movie moguls like Steven Spielberg. There is no question that SGI is hot. Its visual workstations – Indy and Indigo – and its server and number crunching systems – Reality Station, Power Onyx and Power Challenge – offer a powerful blend of 3D graphics capabilities and supercomputer-like performance and are considered to be the Ferarris and Porsches of the computing world. The company philosophy is simple – to make and sell the very best in visual computing. This requires a business model which has become singularly unfashionable. While PC makers are turning into mere assemblers of ready-made standard boards, cards, disk drives and plastic, and while server companies have turned to commodity chips and off-the-shelf disk drives, Silicon Graphics stands apart.
Attention to detail
To consistently stay ahead of the pack, McCracken believes that SGI needs absolute control of almost every one of its system components. This is why the company owns its own semiconductor designs – through its MIPS Technologies subsidiary; why it designs its own graphics processors and video technology; why it writes its own visually-enhanced version of the Unix operating system; and why it provides lots of the visual design and multimedia production software used on its systems. The result is that SGI is possibly the industry’s most vertically-integrated systems vendor. Such attention to each detail of its systems is not just necessary, says McCracken, but the essence of the company’s success since it was formed by Stanford University professor Jim Clark and six of his students 14 years ago. But it takes a formidable commitment to R&D and a constant battle to stay ahead. With every product cycle SGI needs to scale new heights, keeping ahead of the competition who seek to bring to the masses the innovations which have made SGI systems special. For SGI, the key is innovation and that means spending a hefty 11% to 12% of revenues a year on R&D, equivalent to $250 million in fiscal 1995. As McCracken says: At SGI we’ll take that extra step, whether it is in terms of bus bandwidth, memory bandwidth [or whatever]. If somebody around here designs a product and it’s not the highest performing device in its category, their peers make fun of them. It’s a big thing in our culture that you’ve got to build that performance into it. It might cost more but for a lot of accounts [with visual computer requirements] the price/performance is very good, says McCracken. But it is not all plain sailing for SGI. The gap between SGI workstations and mass produced platforms may be narrowing and SGI may be fin
ding it harder to take that ‘extra step’. Technologies such as Intel’s Pentium Pro microprocessor, combined with Microsoft’s Windows NT, are emerging as inexpensive, capable workstation platforms which are reaching the peaks that only SGI workstations were scaling two to three years back. And closer to SGI in the workstation market, competitors with greater economies of scale such as Sun Microsystems, Hewlett-Packard, Digital Equipment and IBM continue to refine their systems to make them more competitive with SGI.
Niche market
At the same time, the visual computing high-end that SGI has dominated is changing fast. Some of the 3D visualization that has provided SGI with a very profitable niche market is now creeping into the mainstream. SGI has been forced to abandon the low end markets to continue scaling new peaks of innovation, leaving commodity markets to the players such as Sun and Compaq Computer. As an extension of its strategy, the acquisition of Cray Research has confirmed SGI as the world’s largest supercomputer company. SGI now faces a long difficult period of corporate and product integration to bring Cray’s supercomputer activities back into profitability. Elsewhere, SGI has had its image dented after its much-hyped efforts to push into interactive television have come to nothing. Some of these factors are beginning to appear in SGI’s numbers. Its ‘homegrown-wins’ approach has worked well in the past: sales rose 35% in 1994 and soared again by 45% to $2.23 billion in the year to 30 June 1995. But after that long period of sustained growth, the organization Business Week called a ‘gee-whiz company’ in a mid-1994 cover story, may be showing signs that it is going off the boil. Growth over the last two quarters has slowed first to 33% and then to a relatively anaemic 22%. Ed McCracken puts this down to the impact of the company’s transition to a new generation of workstations, plus a number of external factors – the US federal government’s budget impasse, slower economic growth in some European countries, and the ever deeper cuts in defense and scientific research spending. But some outsiders are starting to wonder if these figures are not indicative of more serious strategic problems. SGI’s response to the pressure remains unchanged from its beginnings. Innovation, innovation and more innovation. We think that’s our place, says McCracken. What we manage here is product transitions, we try to churn the market faster than our competition and we try to be there first with new paradigms. We try to understand what new applications will be enabled and we try to embed that into the system all the way down into the chip level. If we miss a generation we’re in trouble. There is some risk associated with that, but we manage that risk. It’s the kind of business model that SGI co-founder Jim Clark came to feel was a potential time bomb. Clark, who left to form Internet software firm Netscape Communications in 1994, disagreed with the company strategy. He feared that Silicon Graphics – and himself – would be stuck in high-end hell, the lucrative niche it has staked out. He feels that the low end of mass market computing is where the standards are set and where the real action is. The apex of the pyramid, where SGI lives, is constantly under threat of collapse as it is eaten away by the commodity base, he says. The overall workstation market grew by 43% in units shipped in 1995 to reach 1.4 million units, but as prices fell, total revenue rose more slowly by 24% to reach $17 billion, says market watcher IDC. But in what IDC calls the traditional workstation market of technical rather than commercial desktop workstations, revenues only grew by 12% and unit shipments were only up 4%. Nevertheless, IDC says SGI led the growth charts in that market ahead of Sun, IBM, HP and Digital. Is McCracken, who Clark recruited from Hewlett-Packard in 1984, worried? Oh yeah, we’re worried about everything. One of the characteristics of Silicon Valley is that we are all paranoid about everything all the time, he
says.
The next Apple?
With vertical integration the key to its competitive edge, SGI bears some striking resemblance to a high-profile neighbor. Like SGI, Apple Computer dominates specific markets with premium priced products; it has kept a tight grip on its homegrown technology, steering clear of ‘inferior’ standards; and it has faced the hordes of Microsoft/Intel systems which continue to scale upwards toward ever higher performance and ubiquitousness. While that formula worked for a long time at Apple, the story of the company’s fall from grace in 1995 must keep McCracken’s level of paranoia high. To prosper, SGI is under pressure to continue to find areas where it can innovate, and new markets that need its cutting edge technology, says Hambrecht & Quist analyst Doug Van Dorsten. SGI has a choice of staying in the high end, high profit margin sector or trying to cut prices in exchange for market share. It’s exactly the same choice Apple faced, except that Apple did not add enough differentiation while SGI still does. McCracken concedes that SGI has similarities to Apple, but he points to the positive aspects. As with Apple users, a strong brand loyalty among SGI users borders on cult status. McCracken also sees a fundamental gulf between life at Apple and SGI. There are some real differences in the culture of the two companies. SGI has a real hard edge to it. We don’t just sit there and talk about what we are going to do. People are very strictly evaluated on whether they get the job done. That was never true at Apple; Apple was always more of a philosopher’s company. Controlling the underlying chip technology is an essential part of SGI’s strategy. It means the company builds support for its high- end graphics processing deep within the system architecture. This is why SGI has shunned any involvement with the Windows NT bandwagon, preferring its own version of Unix which sews specialist support for real-time processing and 3D graphical imaging right into the fabric of the system. It is also why it has had its own chip design labs since acquiring MIPS Technologies in 1992. The benefits it gets from that ownership have not left SGI blind to the fact that it needs to make MIPS a volume chip in order to sustain the huge costs of developing enhancements and next generation silicon. While MIPS processors are part of all SGI products, real volume will come from MIPS being used in the low-end consumer market through deals with Sony for inclusion in PlayStation and with Nintendo as part of its forthcoming Ultra-64 system. There is also a potentially important partnership with Time Warner to use MIPS chips in interactive TV set-top boxes. The chips also sit at the heart of commercial Unix servers from Siemens-Nixdorf, and its subsidiary Pyramid Tech- nology, as well as systems from NEC and Toshiba. Newly-acquired Cray will also eventually switch to MIPS-based systems. Analysts applaud SGI’s efforts to guarantee a wider audience for the MIPS chip. Importantly, these deals help generate the high manufacturing volumes which drive new versions of the MIPS microprocessor architecture. MIPS is facing stiff competition from Digital’s Alpha, and most importantly, from the Intel-HP alliance to develop the ‘Merced’ microprocessor, the next generation after Intel’s x86 and Hewlett-Packard’s PA-RISC lines. Merced, which the companies’ plan to ship three years from now, will use so- called ‘very long instruction word’ technology which potentially offers superior performance to current RISC designs.
Performance race
SGI is working on its Merced competitor, says McCracken, which will not use VLIW technology but should keep up in the performance race. We will have something new that will be out before [Merced], he says. Thanks to sales of 5 million MIPS microprocessors last year and a projected volume of 20 million this year, SGI’s semiconductor manufacturing partners, NEC and Toshiba will be able to invest the billions of dollars necessary for building the chip manufacturing process technologies
. SGI will also need to spend $100 million for the design of new processors. While volume ramps up, MIPS has still scope to increase performance. MIPS is a very good chip, observes Linley Gwennap, editor of the Microprocessor Report. And SGI probably has about three years to continue innovating on MIPS before it might face more serious challenges. As proof of that, SGI introduced its latest R5000 and R10000 MIPS microprocessors late last year, performance engines that give the company ownership of world’s fastest commercial chip design. But the lead is temporary. MIPS will be leapfrogged later this year when Digital and HP introduce their latest designs. In part II tomorrow, CBR examines just where SGI’s weaknesses lie, why Hollywood can’t get enough of it’s technology, and the method behind the Mountain View company’s recent acquisition strategy.
