Читать книгу The Race For A New Game Machine: - David Shippy - Страница 12

It is competition that drives us to higher levels of excellence and, therefore, to more opportunity. An accurate assessment of our competition’s capabilities is what enables us to refine the boundaries of our bold vision. We must make sure we shoot high enough.

JIM KAHLE, CHEKIB AKROUT, AND I relaxed on Kahle’s deck. A spectacular peach and purple sunset clung to the sky over the Texas Hill Country beyond Lake Austin. Just for a moment, the conversation lulled, got whisper quiet. Hummingbirds darted in and out of colorful flower beds, ruby throats glistening in the last of the sunlight. Bees buzzed in the trumpet vine. Heavy summer air and the sweet smell of gardenias wafted over us. I tipped back in my chair, propped my feet on the deck’s sun-weathered railing, and watched as a lone water-skier carved out one rooster tail after another on the placid lake below. Contentment flowed over me like warm honey…or maybe it was the Scotch. Very old, very good, single malt. Kahle’s favorite.

It was late summer in 2001 and after six months of startup work, things were going so well at the Design Center—the team was finally approaching critical mass, and a concept of the chip was coming together—that we decided to have our own little private celebration. “To success,” I toasted. We clinked glasses and smiled like silly old fools.

Darkness crept across the sky, pushing the last of the golden glow below the horizon. Kahle lit tiki torches and citronella candles to keep the mosquitoes at bay. Akrout was in particularly high spirits, and the more he drank, the more he unconsciously slipped French or Arabic words into the conversation—something Kahle and I affectionately called Chekib-speak. Every once in a while, something got lost in translation, and Akrout backed up to explain. After about the fifth time, I laughed so hard I nearly fell out of my chair. I guess we were pretty loud, because Mary, Kahle’s wife, came out to tell us to quiet down. The kids were in bed.

The silence might have lasted a full two seconds after she went back inside, then Akrout picked up right where he’d left off. He and Kahle took off on a tangent, fiercely debating some minute technical feature of the chip. No detail was too insignificant, for Akrout was such a geek at heart. Even as a successful vice president of one of the most powerful corporations in America, he could still hold his own against our top circuit designers. He loved technology; he was on fire about it.

Akrout could argue with Kahle all night long, but I didn’t want to squander this opportunity. We weren’t alone all that often. There was just one thing I wanted to hear from him. “What’s keeping you awake at night?” I asked, interrupting the friendly banter. “What’s the bottom line? The worst fear?”

Akrout replied without hesitation: “We don’t have any idea what Intel has up their sleeves, what they might bring out of the shadows in response to STI’s challenge. The home computing environment, not just games, is the ultimate target for all three STI partners, but Intel still dominates the PC market with over eighty-five percent of the market share. Intel also provided the chip for Microsoft’s Xbox, the most significant threat to the PlayStation line. We need to know what to expect next in their products.”

Intel was and still is, the number one semiconductor chip manufacturing company in the world. They had a yearly revenue of $40 billion, giving them both the dollars and the engineering talent to go head-to-head with any company in the industry.

We sipped our Scotch—had a nice buzz going—and discussed Intel’s potential for a while. Akrout was right. We didn’t really know our enemy anymore. The excitement in the Design Center was almost palpable, and my team was revving up for a serious race, but did they know where the finish line was? What could we offer that would beat Intel? That was really the question of the day. We had to know where our competition was heading.

I drove into the parking garage right behind Akrout the next morning. I wasn’t normally an early bird, but he was. I glanced at my watch just to double check. Yes, his routine changed, not mine. He usually arrived early enough to park his BMW sedan in the same convenient first-floor spot in a nearly empty garage, but today I followed him to the fourth level before we came upon some empty slots. Maybe he felt as rough as I did after our night with Kahle and got a late start. It made me feel better to think that he was no more Superman than I was.

Then I watched him climb out of his car. He was practically wrinkle-free, all crisp and neat and bright eyed. On second thought, I decided, he had probably already attended two business meetings by phone, sent twenty e-mails from home, and made a dozen phone calls while he drove into work. He grinned at me, and we walked together from the garage to the office building, though I had to pour on a little speed to keep up with his jaunty pace. He chattered cheerfully about the beautiful weather, and I grunted occasional responses. Obviously, the Scotch affected us differently. Maybe he was Superman after all.

Akrout always came in with that sunny smile on his face, patting the backs of the engineers he met along the way to his office, stopping to chat with someone by the coffee machine. Junior engineers were utterly stunned when Akrout called them by name and asked questions about their specific design work. His interest was genuine, and every engineer sensed the sincerity in his words.

Most days I saw little of Akrout. He spent his time framing the big picture, getting support from his peer executives, clearing away the barriers that threatened the Design Center. He assessed and reassessed the strengths and weaknesses of the STI partnerships. He dissected and studied the assumptions behind the business case that told us this venture made good financial sense. He queried experts on the future market potential for this breakthrough product. He placated customers. On top of that, he directed all chip development for Apple’s desktops and laptops, and Nintendo’s GameCube. A very busy man indeed.

We parted at the elevators on the third floor, but as Akrout stepped away, he grinned and said over his shoulder, “What is Intel thinking, and what are you going to do about it?” I nodded and kept walking toward my cubicle. It was too big a question to answer that early in the morning.

Akrout enjoyed a very close partnership with Apple, so he was keenly aware of how difficult it was to unseat Intel, the reigning king in the PC business. He stood at Apple’s side for years, loyally working with them to create the perfect product roadmap that would propel them ahead of the competition. It was hard, demanding work, and they had not won yet. Though Apple (with IBM chips) had at times outdistanced Intel in terms of processor performance, they were still behind in terms of raw chip speed. Akrout hoped he might finally have the right ingredients to put both Apple and the STI partners ahead of his old enemy. He was the eternal optimist, and I prayed we wouldn’t disappoint him.

To get an answer for his concerns about Intel, Akrout enlisted Dr. Peter Hofstee, one of IBM’s top research engineers, to explore the competition for our next-generation microprocessor and help define project goals to ensure a win. Hofstee’s claim to fame came from his involvement in inventing the first chip in the industry to break the one-gigahertz processor speed barrier, the ultimate Mt. Everest challenge to a chip designer pushing the leading edge of technology in the late 1990s. Hofstee, a brilliant researcher, obtained his Ph.D. in computer science from the California Institute of Technology, a school that produces some of the best computer engineers in the country. Originally from the Netherlands, he decided to stay in the United States after he finished his studies and even accepted a faculty position teaching computer science and chip design at CalTech from 1995 to 1996.

After sending Hofstee off to get the scoop on Intel, Akrout instructed Jim Kahle and me to define what we thought would be the most aggressive design parameters we could possibly achieve, and then to stretch beyond that. He wanted the STI product to beat all existing records, and then when Hofstee returned with his projections for Intel, we would see that our stretch design would trounce the enemy.

Kahle and I had confronted Intel during our days at the Somerset Design Center. Back then, a new computer architecture referred to as a RISC architecture was the kicker for our product, the thing that made our chip stand out among the competition. Instructions are the fundamental directions, the recipe, that tell the hardware what to do. We all believed a highly efficient RISC machine could be Apple’s wedge into Intel’s domination of the PC market. It was supposed to provide higher performance and higher frequency at a lower cost. Up until that time, Intel retained the top spot with their tried and true X86 Complex Instruction Set Computer (CISC) chips, the brains for virtually all PCs. The RISC approach relied on the fact that most software applications actually use only a small subset of basic instructions. Our work at Somerset focused on optimizing this reduced set of instructions to run faster than the older CISC architecture. The problem was that PC owners needed to port all their old legacy software programs from their old systems to their new ones, giving Intel a definite advantage with every upgrade.

In my first job assignment at IBM, I got a nice exposure to both the older CISC approach and the newer RISC approach while I cut my teeth on computer architecture and logic design. IBM hired me in June of 1985 in Endicott, New York, the birthplace of the company and much of the corporate tradition. In the 1960s and 1970s, IBM staked its claim to fame on the s/360 and s/370 mainframe computers designed and built in New York. However, in the 1980s, DEC and their popular VAX minicomputers started eating away at the low end of the mainframe market. Endicott’s Glendale Lab was responsible for delivering a crushing response to this threat with a product called the 9370.

The main central processing unit (CPU) in the 9370 was a complex CISC processor, because they needed something that was code compatible with the mainframe s/370. I landed the job of designing the specialized input/output processor (IOP) that handled all of the traffic in and out of the computer. This IOP was just a very simple RISC design, but I was thrilled to be working on an actual microprocessor core.

When I signed on with IBM, my buddies laughed at me and said, “Why do you want to go there? They won’t give a rookie like you any good design work.”

With confidence, I replied, “They will if I’m a good designer.”

How wrong my buddies were. IBM happened to be in an expansion stage just then and was willing to risk putting new hires into key roles. This product never saw the light of day, but it gave me valuable lessons in computer design.

I also learned to make homebrew and to snow ski while in Endicott. I sometimes think those skills have served me almost as well as my computer design experience. My cube mate in Endicott was Brice Feal. Brice was a zany bachelor with a wide variety of interests. He invited me over to his house after work and took me to his cellar, where several hundred bottles of beer filled the shelves. He blew the dust off one of them, cracked it open, and filled two frosty mugs.

“Give it a try,” he said with a smile, and we clinked our mugs together in a silent toast.

It was the smoothest, best beer I’d ever tasted. “Wow, Brice!” I exclaimed. “Where can I buy this stuff?”

He replied, “You can’t get this in stores. I make it.”

His particular brand of beer was really a barley wine with a smooth flavor and high alcohol content. I was hooked, so Brice taught me all of the tricks and soon I was brewing my own beer.

Brice also introduced me to another passion—snow skiing. There was a local ski resort called Greek Peak. Many Friday afternoons we skipped out of work early and hit the slopes. The ski resort lit the runs with spotlights, so we skied until very late at night.

RISC computer design, homebrewing, and skiing. Life wasn’t too bad in Endicott for a young engineer. However, I heard about a new development project at the IBM site in Austin, Texas, and the central processor was RISC rather than CISC. From everything I knew, this seemed like the way to go. Exposed to both methods, I could see that the RISC approach would deliver simpler, higher performance hardware. If I stayed in Endicott, the sexier computer design assignments would be on a messy CISC design. I wanted an opportunity to create streamlined fast microprocessors using the RISC techniques.

So in 1989, I went to the office of my third-line manager, Bobby Dunbar, and told him I was going to Austin to work on a RISC microprocessor. Bobby was just a good ol’ boy, content to ride the success of the s/370 computers he’d come to know and love. He propped his boots on the desk and laughed at me. “Nothing will ever come of that RISC architecture,” he said.

His predictions did not prove true. Today, the highest volume chips produced at IBM and at Freescale (Motorola’s spin-off) carry the PowerPC RISC architecture. PowerPC is the architecture of choice at IBM for everything from game chips to supercomputer server chips. The PowerPC and Intel’s X86 are the two primary architectures that stood the test of time. I knew a good thing when I saw it.

Intel stayed with their proven architecture, but they adopted virtually all the RISC techniques developed by our Somerset team. They employed a brute force approach to microprocessor design. They applied a team of thousands of engineers to streamline the instructions, and then to optimize and tweak those X86 microprocessors until they could offer as good or better performance at higher frequencies than we could with our designs.

Intel capitalized on parallel processing techniques invented by IBM and other companies, such as superscalar and out-of-order processing. A superscalar design gains efficiencies by having multiple parallel execution units that operate in parallel on groups of instructions. It was like the difference between having multiple checkout lines at the grocery store versus a single checkout line. An out-of-order design gains efficiencies by scheduling instructions when they are ready to execute. This meant when even one long instruction stalled, other shorter instructions could be routed around it. Idle time is wasted time.

Intel remained the only game in town when it came to the PC, in spite of our best efforts at Somerset. Of course, it didn’t help that IBM experienced various software failures around the same time.

This early defeat at the hands of Intel was in the back of my mind as I walked beside the snazzy glass wall that separated cube-city from the STI war room, the place where Kahle held his daily architecture meetings. Another engineer beat me into the room and snagged the last available Ethernet port for his laptop. There were never enough outlets for everyone at the table, and the security team had recently disabled the wireless capabilities in the building as a defensive security measure. I sat near the back of the room and opened my laptop, intending to work offline while we waited for the meeting to start. But a swirl of thoughts about Intel grabbed my attention. I knew Intel wasn’t sitting on its hands; their people were working, just as we were, to push the limits of technology. They had thousands of engineers working on their next chip, while we had a few dozen. I knew several former IBMers, smart guys, who moved to Intel, and I knew they were inventing cool new stuff for our enemy. How could we compete?

Kahle waited for a quorum to gather, then stood and explained Hofstee’s mission and our job. He said, “You have to be paranoid when it comes to beating Intel. Basically, we need to attack with multiple weapons, because just having a higher frequency will not be enough to make Intel’s customers switch. This calls for an extraordinary new design offering an order-of-magnitude improvement in performance.”

It wasn’t quite a battle cry, but it generated the right discussions. We batted around various strategies, scribbled ideas on the board, and argued about competing technologies, all the while lacing our language with words like frequency, throughput, process, and performance. Also, there was a new term in the industry to describe the raw frequency of a processor. It was called “fanout-of-four” (FO4), which described the number of gate delays in each pipeline stage or, more specifically, the number of simple inverter gates connected in series, each having a fanout or load of four gates connected to them. A smaller FO4 gate delay translates to a faster frequency. This new term provided us with a way to describe and compare processor speeds across multiple manufacturing technologies. Therefore, when a processor design migrated from, say, a 90 nanometer manufacturing process to the newer generation 65 nanometer technology, the FO4 gate delay would stay the same while the frequency (gigahertz) could increase. The Power4 processor I worked on had a 24 FO4 gate delay, which translated into a 1.1 gigahertz clock speed. That was the fastest in IBM. Intel had the current speed record in 2001 with an 18 FO4 gate delay, which translated into 1.5 gigahertz.

The pressure was on.

The need for low power really tied our hands. For the compact cost-conscious PlayStation 3, achieving the frequency of a PC with the reduced power budget of the game console would be a huge challenge. Game consoles are smaller than PCs and have less capacity to keep the chips cool, and games are very compute-intensive functions that tend to max out the processor usage. Higher power on the PlayStation 3 would lead to more costly thermal control techniques like fans and heat sinks, and the costs for those components were very hard to reduce over time. Kahle explained that Kutaragi’s aggressive cost-cutting strategy proved to be a huge money maker for Sony on previous products, so of course that would be the plan for this product too.

The Sony architect, Takeski Yamazaki, said in broken English, “Seventy-five watts is the highest power the console can physically tolerate.” Heads nodded in agreement, all Sony engineers.

I was skeptical. Most of the server chips and PCs I had worked on in the past were well over this 75 watt budget. We set our sights on designing the fastest microprocessor in the world, but could we still do it knowing that we faced this ridiculously low power budget? We didn’t know yet what raw frequency (measured in gigahertz) would describe the top speed of our microprocessor, but now we knew that, at least for the game console application, it would be constrained by this maximum operating power. What would happen if we failed to meet the power goal? Major malfunction resulting in either a hang or an automatic shutdown. Or picture little Johnny game-player running to Mom when his game console burned a hole in his desk. Or worse, console meltdown. The images weren’t pretty. The faster a chip runs, the more heat it generates, so to avoid a meltdown, we have to either remove all that heat, or run it at a slower speed. Two terms that don’t normally go hand in hand are high speed and low power. Seventy-five watts was going to be really tough.

Still, I wanted to believe Kahle when he encouraged the team as we adjourned: “Guys, I don’t know how we’re going to get there, but we’re going to do it.”

Dr. Hofstee was finally ready to present his competitive analysis to Akrout and the team. I had previewed the data during the course of his research, and I was anxious to hear the most recent stuff. He had studied the trends for chip frequencies, primarily dictated by Intel. Detailed graphs showed his predictions of what current and future technologies could achieve. Part science and part science fiction, his analysis was a combination of what the physics of technology could deliver and what smart engineering could achieve. Intel was always the benchmark. There was no other game in town.

I had done some comparison studies in the past, much like Hofstee’s task, so I was well aware of what a struggle it was to compare the different microprocessor designs (apples and oranges) on the market. Each one, shrouded in secrecy, manufactured in a different fabrication facility, used a different process. Details were scarce. As much as the efficiency of the design, the silicon manufacturing technology also determined the achievable frequency for the chip. It defined the minimum size of the transistor, the fundamental switching device in a design. Transistors became smaller and smaller as technologies evolved, and smaller meant faster.

We gathered in Akrout’s executive conference room with its subdued sage green against natural maple, smoky shaded windows, automatically dimming recessed lighting, and luxuriously upholstered chairs that swiveled, tilted, adjusted, and rocked into no less than twenty-four different positions. Hofstee prepared to present his work. Jim Kahle and I walked in together, both dressed in sandals and shorts in celebration of the last days of summer. I lounged in the back of the room as usual and tilted my chair against the wall. Kahle moved to the front of the room and sat across the table from Akrout, who was already warming up the crowd. He was in prime form, chatting with each attendee as he or she entered the room, laughing with Kahle, greeting those who were attending by phone. About twelve other technical leads and a few managers sat at the table or on either side of the room against the walls. Jim Warnock, a newly appointed Distinguished Engineer who had recently joined Akrout’s staff, flew down from IBM’s Research Division in Yorktown just for this meeting. All the attendees were IBMers and all were male except for Mickie Phipps, Kathy Papermaster, and Linda Van Grinsven, three of the managers who were now responsible for the multicorporation design teams.

Akrout stood and put a hand on Hofstee’s shoulder. “I asked Peter to research our competition, and he is ready to present his findings. This is my first time to see his data, too. What do we have to do to beat Intel? Where do we set the bar? Listen closely, for you are the ones,” he paused to point around the room at us, “who will determine whether we succeed or fail at this endeavor.” Through a wiring hub in the center of the conference table, Hofstee connected his laptop to the top-of-the-line projector suspended from the ceiling above the table. He looked more like a college student than the veteran engineer he was. Tall and lanky, perfectly straight sandy-red hair cut fashionably long, an open expressive face. He rubbed long-fingered hands together and clicked the button to pull up his introductory slide. Years of experience standing before engineering students gave him confidence and style. He carefully worked his way through a series of charts and graphs, clearly and methodically building the case to support his conclusions.

Akrout sat forward in his chair, intensely focused on the data Hofstee presented. Occasionally, he pointed to the screen and asked in his shortened version of English, “Why that?”

Questions were raised, interruptions were tolerated. Hofstee captured our complete attention. It was a topic of extreme importance to each one of us, as the conclusion of this meeting could very well dictate our workload for the next two to three years.

“When the STI project started in 2001,” Hofstee said, “the best of breed in the industry was the Intel Pentium4 microprocessor. It topped out at about 1.5 gigahertz in a high-end PC. That design has eighteen FO4 gate delays in the basic pipeline. However, an inner integer core in the Pentium4 executes at nine FO4, or twice that speed!” He illustrated this point with a detailed diagram showing how Intel’s design frequencies improved year to year. “Based on this data,” Hofstee concluded, “Intel could very well produce a microprocessor in the 2005 timeframe that could achieve nine to ten FO4 and over four gigahertz, so to be competitive in this timeframe, we need to match that frequency in our seventy-five-watt power budget. We need a ten FO4, four gigahertz frequency!”

The room exploded. It seemed impossible!

I took Hofstee to task over a few of his basic assumptions, and he gave reasonable and believable explanations. Others argued back and forth about his predictions for Intel’s ability to make such rapid improvements in the frequency of their chips.

There was much skepticism and vocal opposition from Warnock. “I am very certain four gigahertz will push the chip power well beyond the seventy-five-watt power target, but even so, I doubt we can ever come close to achieving the frequency goal anyway. You’ll only demoralize the team if you present them with this unachievable goal.”

Other naysayers joined in, voicing opinions that bespoke worry about the team, or about the accuracy of Hofstee’s predictions, or the lunacy of reaching for the stars. The din in the conference room grew louder and louder.

Hofstee spoke clearly and loudly over the roar of discussion: “Guys, guys!” He held out his hands, patting them up and down like a priest bestowing blessings on his congregation. The roar dropped down a notch. “Intel already knows how to do this! They currently have a three-gigahertz fixed-point unit in their Pentium4. If they can do it, we can.”

Kahle immediately agreed with Hofstee. “I’ve fought the ’dark side’ before and lost,” he said, referring to Intel, “and the last thing I want to do is to come up short again.” He pounded his fist on the table to let us know his decision was final.

Kahle let us debate the point for a while, but we all knew it was pointless to argue. We settled on 10 FO4, four gigahertz at 75 watts as the mind-boggling goal for the STI chip. This meant a quantum leap in frequency with only a fraction of the power compared to the best-of-breed, Intel-based PCs. These were such far-reaching targets—like projecting into outer space—that we could only trust they would be enough to push us ahead of Intel. I sincerely hoped the chance to grapple with Goliath would motivate my team to overachieve and slay the giant.

It was 2001, and no one even knew if a four-gigahertz clock speed was physically possible. To start with, it required ultraefficient circuits with no excess fluff or nice-to-haves. We had to invent a new animal that ran like a cheetah, roared like a lion, and ate like a kitten.

We ended the meeting with concerns weighing heavily on our minds, but before we left, we all stood together and vowed to do whatever it took to achieve the goals. Akrout looked each one of us in the eye, shook our hands, and encouraged us: “Before Roger Bannister broke the four-minute mile barrier, nobody thought it was possible. We can do this.” His bold vision for the team and his confidence in us won us over.

The bar was set, a multicore four-gigahertz microprocessor chip with a power budget under 75 watts! In comparison, the fastest Intel chip at the time was a single core Pentium4 microprocessor running at 1.5 gigahertz. The best IBM microprocessor in production was the Power4 microprocessor running at 1.1 gigahertz. It was an aggressive goal that would push us to innovate on all fronts and would require an extraordinary effort from each engineer. We would have to quadruple the speed in a single generation of microprocessor design and also meet a much lower power budget.

Our efforts to top the 1 gigahertz mark on the Power4 microprocessor had been a real struggle. We had spent nearly a year in high-level design defining the high-frequency microarchitecture. During the implementation phase, we tweaked timing paths until we were blue in the face. For that design, we didn’t even consider a power budget. These new goals made my head hurt just thinking about them.

Now that we understood the competition, we were able to share Kutaragi’s bold vision for the future of this little game machine.