The money needed to play in the highest-performance space has forced everyone but IBM to the sidelines.

Once upon a time, there were many major computer makers in the world, and they developed and marketed expensive high end computational systems, spent a lot of money developing cool technology, and had impressive manufacturing operations.

Remember Siemens, the BUNCH (Burroughs, Univac or Sperry/Univac, NCR, Control Data and Honeywell), Digital Equipment (DEC), ICL and CII Honeywell Bull? Remember supercomputer makers Cray, Cray Research and Steven Chen’s SSI?

Sure, Hitachi, Fujitsu and NEC are still around. Cray still makes a few supercomputers. In what we used to call the workstation space, a few companies remain. Some part of Silicon Graphics still exists in the Silicon Valley, at least in name. Oracle purchased Sun Microsystems, a major server maker. Looking a little deeper into the server space, Hewlett-Packard is also a major player, and if PC servers are included, Dell counts too. However, if we look at the high end computing business that we used to call the mainframe, IBM may be the last one standing.

This year, IBM introduced its zEnterprise 196, claimed the most powerful commercial IBM system ever. According to Big Blue, the zEnterprise System represents one of the most significant updates in nearly 50 years of IBM mainframes and incorporates new technology. It contains the world’s fastest, most powerful microprocessors, the core server boasting 96 microprocessors running at 5.2 GHz, capable of executing more than 50 billion instructions per second (BIPS). The z196 processor is a four-core chip that contains 1.4 billion transistors on a 512-square mm die. IBM reports the die was designed by engineers at its Poughkeepsie, NY, labs and manufactured using 45 nm SOI processor technology in its 300 mm fab in East Fishkill, NY. Contributions for the processor development also are reported from IBM labs in Austin, TX, Germany, Israel and India. By the way, the zEnterprise System represents an R&D investment of more than $1.5 billion, as well as more than three years of collaboration with some of IBM’s top clients around the world.

Why could IBM be the last mainframe company standing? Few, if any, companies have the resources to continue to play in this high-performance game. Even NEC exited Japan’s Next Generation Supercomputer Project, focused on building a hybrid scalar/vector massively parallel system. The $1.2 billion program (when launched in September 2006), sponsored by the Japanese government, included NEC, Fujitsu and Hitachi. While NEC said in 2009 when it pulled out of the government project it would continue to develop its vector supercomputers, the company has been struggling financially and is merging its semiconductor business with Renesas.
NEC’s last supercomputer, the Earth Simulator, was a cluster of 5,120 of its SC series of vector supercomputers and was rated at 35.8 teraflops. Even though I saw some very impressive computations when I visited it in Yokohama, Japan, several years ago, the Earth Simulator cost $350 million to build, consumed 5 MW of power, and occupied 34,000 sq. ft. of floor space.1 The power cables alone were amazing. Maybe NEC has even built its last SX supercomputer. Fujitsu’s financial position probably does not allow it to make the massive investments it once did when it was a mainframe powerhouse. Today, Hitachi is tied in with IBM’s technology and may be considered more a marketing arm than technology developer.

In the past 25 years, I have toured most of the major mainframe and supercomputer makers’ manufacturing lines. When I call IBM the Last of the Mainframe Makers, I am serious. It really could be. Who else can afford these massive investments? It is questionable if even IBM can. Maybe only governments will be able to spend the financial resources required for advanced computing. Should we watch for China Computer, Inc.? After all, it did spend the money to play in the Space Race, orbiting a manned vehicle around the earth.

References

1. Timothy Prickett Morgan, “NEC Abandons Japan’s Next-Gen Supercomputer,” The Register, May 14, 2009.

E. Jan Vardaman is president of TechSearch International (techsearchinc.com); jan@techsearchinc.com. Her column appears bimonthly.