Three weeks after that secret meeting at the Gingerman, I found myself wearing an IBM badge…again. I would not have predicted it in my wildest imagination. When I resigned from IBM in 1999 to join a little startup company, I was certain I had left big company bureaucracy behind. The high-tech dot-com era was at its peak, and my pot of gold was calling. But Kahle convinced me otherwise. I walked back into IBM’s familiar pink palace on Burnet Road, very happy that I had not completely burned my bridges when I left.
Jim Kahle welcomed me back with a tour of his fancy facility. IBM signed over several floors of this high-rise to house his multicorporation team, and Kahle spared no expense in remodeling the space.
“Visiting executives sometimes criticize the plush surroundings, but the investment was for the team,” he said, as we strolled down a hallway lit with contemporary high-end wall sconces. “The modern uniformity bonds us together, levels the playing field, and rids all of us of petty jealousies over prime real estate.”
We walked past a handful of offices on our tour. One was Kahle’s, I knew. Kahle said the other four belonged to the IBM director, Chekib Akrout, his counterpart Sony and Toshiba directors, and the software leader. Everyone else on the team, including the managers, got a cubicle.
Kahle proudly ushered me past many empty cubicles to my own box, the best one in the building—a corner unit on the third floor with floor-to-ceiling windows on two sides. I turned my back to Kahle to hide my disappointment, looking instead out my new window at tall cedar trees swaying back and forth on the wooded lawn. The amount of money sunk into the modernization effort was impressive, but I would gladly have traded my new high-tech cubicle any day for an old, outdated windowless office with a door that closed. I didn’t tell Kahle that.
He pulled out a sliding whiteboard to demonstrate that I could increase the privacy in my cube, but that did nothing to reduce the distraction from the high noise level. No stranger to cubicle environments, I had always found it difficult to concentrate when I could so easily hear the phone conversations of my neighbors across the partition. Even quiet conversations among my fellow engineers disturbed my thought processes. Numerous break-out rooms lined the walls of the open floor plan, providing space for private conversations and small meetings, but I worried about the negative impact on the team with the heightened scrutiny and lack of privacy.
Kahle introduced me to my Japanese partners and a handful of IBMers as we continued our tour. My stomach clenched. So few people. Who was going to help me do all that design work Kahle promised?
The second and third floors sported large break rooms with snazzy new-age green glass walls laced with zigzagging copper tubes. Modern kitchen appliances and café-style furniture matched the look.
Kahle and I returned to his office. File cabinets lined one wall, while photos of previous chip designs, various project plaques, and patent awards adorned another. I called it the “I LOVE ME” wall. Every engineer has one. His desk was a familiar mess, cluttered with stacks of technical papers and journals. We sat at his modular, natural-maple desk while he gave me a short history lesson on the Design Center.
“My first discussions with Ken Kutaragi occurred just a few months ago in a hotel in Roppongi, Japan, the location of IBM Japan’s head office. I offered him every chip option we had for a potential PlayStation 3 processor. The idea I liked the best and pushed the hardest was a derivative of the Power4,” Kahle said.
“Oh, man, that would’ve been sweet!” I said. Kahle and I had invested a great deal of our own blood, sweat, and tears into the development of the Power4, and we would have been so proud to watch this baby take over a high-volume consumer market like the PlayStation 3. This highly successful server chip was IBM’s first microprocessor to break the one-gigahertz clock barrier. That long-standing performance obstacle had once seemed insurmountable. From the early 1980s—when Intel’s first 8086 PC microprocessors ran in the low megahertz range—to the turn of the century, it had taken nearly twenty years of evolution to break this barrier.
“Unfortunately, that idea didn’t capture Kutaragi’s interest.” Kahle pointed to a list of products on his whiteboard. A maze of notes and arrows swirled around the barely decipherable column of names. “So I proposed all these other IBM PowerPC derivatives, including the line of embedded cores developed by the team in Raleigh, North Carolina. By the end of the Roppongi trip, I had pitched the entire spectrum of IBM’s processor cores, from the very simple and small up to the very large and complex.”
I studied the list for a moment. “So Kutaragi rejected everything. Did he give you some idea of what he does want?” I was beginning to worry that Kutaragi had some pie-in-the-sky dream that wasn’t achievable in real hardware.
“Oh, yes, he did.” The excitement in Kahle’s voice ramped up. “He challenged us to create something new to leapfrog Intel’s technology, something like a supercomputer-on-a-chip. This is Kutaragi’s bold vision. His chip will be the heart and soul of a bleeding-edge gaming console. He insists on both multigigahertz frequency and very high floating-point mathematical computation capability.”
Floating-point units are included in many microprocessors, but they involve some very complex circuitry with a high transistor count, which translates into costly silicon real estate. Due to this complexity and size, they create much greater challenges for achieving high frequencies.
I nodded my head, nearly salivating over the opportunity to invent something new. This supercomputer-on-a-chip would provide high precision as the chip adds, subtracts, or multiplies very large decimal numbers. “I agree that’s a deadly and difficult to beat combination. Tough to build, though,” I said.
“You bet,” Kahle answered. “A typical game application uses millions of floating-point computations to create animated graphics. Higher precision means that the processor can calculate the physics involved in moving, bending, jumping, crushing, colliding, bouncing, and so on, with a higher degree of accuracy, and thus provide more fluid character movement in crisp, realistic scenes.”
That made a lot of sense to me. I already knew that millions of computations determine every pixel position in every scene that flashes across my computer or TV screen. The faster the position of the final pixel is calculated, the more lifelike and fluid that game becomes.
Kahle understood very well what Kutaragi wanted, but in order to get IBM to commit to a major new processor development effort, I knew he would have to engage in and win a major turf battle with IBM’s Server Group, our former team. I was sure the folks that had worked with Kahle and me on the Power4 processor would fight fiercely to own this processor and would push for a Power4 derivative as the base design. “What do our friends in the Server Group have to say about all this?” I asked.
“Plenty. They argued that one of their own homegrown server chips or a derivative thereof could just as easily fulfill Sony’s game machine requirements and IBM’s requirements for new broadband product development. Why invest millions of dollars into something new when off-the-shelf processors fill the bill?”
“Can’t argue with that line of thinking,” I said.
“True, but I’ve already been down that path with Kutaragi, and that idea won’t fly. He doesn’t want what they have. As much as I dread it, a futile and time-consuming turf battle might possibly be the only way to make progress and clinch a deal with Kutaragi. Even with a corporate approval in my pocket, I have no doubt there will continue to be bloody battles over which in-house team—Server Group or Microelectronics Division—will win the job.”
I sympathized with Kahle’s reluctance to enter into a turf war because I carried my own scars from previous processor wars within IBM. In the mid-1980s,
