IBM: Let the Chips Fall Where They May
Page 1 of 1
Sure, watchers can delight in the microprocessor battles between Intel Corp. and Advanced Micro Devices (AMD), both of which are trying to top each other for the fastest, most reliable chip for mobile devices, but there are other equally interesting competitive schemes afoot -- try IBM Corp.'s push to catch up with Sun Microsystems Inc. in the UNIX server market.
Aside from pecking away at Sun's market share with new, regular releases from its eServer product line (low-end, midrange, and high-end servers) IBM Thursday revealed that it has crafted a new weapon to power servers -- a chip with two processors. Big Blue calls this its Power4 Gigaprocessor; it incorporates IBM's copper and silicon-on-insulator technology to operate, as one could guess from its name, in excess of one gigahertz.
But as everyone knows, it's one thing to have cool, new technology at your disposal. Big Blue said a Power4 chip delivers bandwidth from Level 2 cache to the processors in excess of 100 GB per second, which is comparable to downloading 20 full length DVDs in a single second. That said, it's another entirely to sell people on the idea of the new scheme, which IBM has apparently done Thursday -- though not yet in the U.S. where the market for servers is arguably the most expansive.
No, Big Blue has secured Munich, Germany's equivalent of, say, MIT, by catching the eye of the Max Planck Society for the Advancement of the Sciences, an organization that researches natural science, social science and the arts and humanities. The non-profit Society wants to build a supercomputer and it will pay IBM an unspecified, multi-million-dollar sum for it's services and new technology.
While it may not be a Deep Blue (it doesn't have a name yet), it will be driven by eServer Power4 technology, and those involved anticipate that it will be the fastest non-classified supercomputer in Europe when the finishing touches are put on it in 2002. It will be designed to achieve a peak performance of 3.8 trillion calculations per second.
It will be used in polymer research, solid state physics and theoretical chemistry, fusion research, astrophysics and biochemistry. But specifically, it will be used to pick out atomic instabilities in the helical structure of proteins that are believed to be responsible for Mad Cow (Creutzfeld-Jakob) and Alzheimer's diseases.
IBM's eServer General Manager Rod Adkins, the man largely credited for turning around IBM's server division in the last few years along with Dr. John Kelly, said this is the next generation of supercomputers for the company following Deep Blue and to ASCI White.
Indeed, the new computer will feature "distributed switch interconnect architecture," which is a new method of connecting devices to boost scalability by combining bus and switch architectures.
For those who are bullish on IBM, such as UBS Warburg, the Planck Society deal news is even more encouraging. In a recent, engaging report titled "Draining the Server: The Empire Strikes Back!" the investment firm said a shift in the server industry makes IBM the top contender to overtake Sun in the UNIX market, which when one considers the case history of IBM's past server follies, makes the notion of such an accomplishment impressive. (In the mid-90s, IBM made a strategic mistake in UNIX. Management focused investments on making UNIX systems as desktop alternatives to Windows. Because of this, IBM completely missed the investment window for high-performance UNIX servers -- a dropped ball that Sun was able to pick up and run with time and time again.)
So, how did Power4 come to fruition? The Warburg report said a key reason for IBM's slow climb to the top from the bottom rung is that management launched a $250 million microprocessor development program in 1998 to develop an advanced microprocessor (Power4) which was designed to leapfrog the Sun UltraSPARC 5 and Intel McKinley microprocessor programs by late 2001.
From where Warburg stands, IBM is on schedule to introduce systems with the Power4 microprocessor in the fourth quarter and the Planck Society deal announced Thursday is a harbinger of things to come. In the report, the unnamed analysts are decidedly confident in IBM's strategy, almost so much as to be championing Big Blue outright. But the analysts have spent time with IBM and have reasons for why IBM is qualified to charge forward to grab enterprise server market share by big fistfuls.
"We have been recommending IBM largely because of the attractiveness of its financial model (large and growing recurring profit model, diversified revenue sources, financial controls, easy compares, mainframe product cycle, cash flow, etc.)," the report said. "But we also see IBM positioned to leverage the fundamental shifts in the 'Draining of the Server' and re-establish leadership in enterprise servers."
The report went on to paint a bleak picture for what it calls those server competitors in the "second-tier." This includes Hewlett-Packard Co. and Compaq, which it said are not in viable positions.
The report concluded with a reason for why IBM may overtake Sun in the enterprise server arena:
"The question now is what matters more for UNIX market leadership -- technology or market share and the advantage share provides in attracting a broader application software portfolio. Normally we would bet on the market leader and the natural application barrier to entry. But IBM's footprint in enterprise servers is huge and extends beyond UNIX to include a $10 billion mainframe revenue stream and a $3 billion AS/400-iSeries revenue base. If Sun continues to execute poorly in server technology, IBM may very well overtake SUN for UNIX server leadership over the next 2-3 years."
Despite the gaudy numbers bandied about by UBS Warburg, the recurring theme in the IBM vs. Sun server war is not necessarily that IBM's products are better or faster, but that they get them out when they say they will and have stuck to their schedules. Warburg reported that Sun is in the middle of a delayed transition to the UltraSPARC III, and as of May 9, had not shipped a single 900 Mhz copper mpu based system.
IBM's real gauntlet to Sun then seems to be, "Can you meet your deadlines?"