(This is a re-post of a post on Sega-16.)
The following is a translation of the final part in a six-part series of articles on the development of the Hitachi SH microcontroller. The article originally appeared in the September 22, 1997 issue of Nikkei Electronics magazine and was subsequently posted on the Renesas website. An archive of the original can be found here.
Fortune and Misfortune are Intertwined; Overcoming Hardship to Reach the World Stage
At last, an evaluation model of the SH microcontroller had been completed. All that remained was to wait for the start of mass production. However, another obstacle blocked the way: almost no users had decided to adopt the SH. The only way to raise widespread awareness in one go was to find a large-scale user to adopt the chip. The target was the home console market, where hundreds of thousands of units could be shipped per month. However, rival makers were several steps ahead. On the tiniest sliver of hope, the SH was promoted to Sega Enterprises.
Something was missing. That was all that Shunpei Kawasaki could think as the SH evaluation chip was finally completed. They had built a RISC microcontroller using an original architecture, and with just two people had put together an instruction set whose performance satisfied even Kawasaki. “The research you’re doing is no good.” Those words said to him six years before had strengthened Kawasaki’s resolve to finish the development of the RISC microcontroller. In that time, just about all of the things he had vowed to do had been accomplished. And yet, something vital was missing: the new microcontroller had no users.
“We’re very interested. We’d love to test it out,” many potential customers said. However, restricted by their project plans, nothing ever came about. Kawasaki could keenly feel the expectations of the Hitachi leadership for the SH. “To be honest, they started to become impatient,” he recalled. No matter how much thought went into the SH, if it didn’t have any users, it was no more than a simple piece of silicon. What could be done?
A Slim Chance
Kawasaki was at a loss for what to do next, but then he heard some encouraging news from one of the sales managers. Hitachi had been promoting the PA-RISC architecture-based PA-10 processor to Sega for inclusion in its next home console, but Sega had just turned it down (the PA-10’s performance is about 10 MIPS). In an effort to redeem themselves, the sales manager wanted to pitch the SH to Sega.
However, the chances of success were slim. Sega was reportedly close to finalizing its choice of NEC’s RISC microcontroller (at that time, Sega was considering about ten different microprocessors and microcontrollers, including the NEC V810 and the Motorola 68040). Nevertheless, there was a remote possibility of success. If the chip was included in a home console, they could expect monthly orders in the hundreds of thousands. This was their big chance to boost the reputation of the SH. They decided to approach the situation as if they had nothing to lose.
In the early summer of 1992, Kawasaki, together with his boss Shirou Baba, sales manager Yasuhiro Ueda, and engineer Junichi Tatesaki, made their way to Sega’s headquarters right next to Haneda Airport. “It has to be a quick victory,” Kawasaki felt. “We have to end it in one shot.” More than was usual for him, he elaborately prepared and practiced his presentation over and over. The long-awaited day had finally arrived.
A Cafeteria Negotiation
The four entered Sega determined to succeed. However, they were told that there were no conference rooms available, and of all places were then led to the company cafeteria. “Employees were sitting in groups around us, chatting away while drinking tea,” Kawasaki recalled.
They met with Kazuhiko Hamada, who was a section manager from Hardware R&D in charge of the hardware design for Sega’s next console (later known as the Saturn). However, let down by the unexpected surroundings, the team from Hitachi failed to find their rhythm as the first day of discussions came to a close. “He told us that it was interesting,” Kawasaki said, but the meeting ended without a clear indication of any real enthusiasm.
Following that first meeting, the team from Hitachi met several more times with Sega, who eventually began to discuss possible conditions for adopting the chip. By some possibility, it seemed that Sega had become interested, and the hopes of the Hitachi team grew.
Some of Sega’s requirements included increasing the performance of the multiplier, including a synchronous DRAM interface circuit, and presenting the results of a hypothetical benchmark for game software.
With their newly assigned tasks, Kawasaki and the sales staff immediately went to work running performance simulations on the new additions. They frequently traveled back and forth to Sega’s offices as the results became available. It seemed that everything was going well. At one point, they were convinced that Sega would adopt the SH.
However, Sega’s attitude remained indecisive. Would they really adopt it? Based on his previous experiences, Kawasaki felt that the situation was bad. “If the other side doesn’t show strong interest after the third meeting,” he said, “then things usually turn out bad.”
Kawasaki’s Second Tough Decision
Meanwhile, in September 1992*, the mass production of the SH-1 had begun. A commemoration ceremony was held in a room in Hitachi’s Musashi factory to celebrate the first shipment.
*[note: the text says 1993 here, but this is almost certainly a typo based on the dates in the rest of the narrative]
It was unusual for Hitachi to hold this kind of event. When the development of the SH first began, it was met with the comment, “Where are we ever going to be able to sell this?” However, as development proceeded, it was designated as a “Special R&D Project” and began to attract strong interest throughout the company. For the development group, the company’s expectations had swelled so much as to almost be a burden.
However, they still had not acquired a large buyer. “We were shipping a few thousands chips per month,” Kawasaki recalled. “All that effort to make the SH, and it wasn’t selling at all. It was a shock.”
As Kawasaki despaired, one day a letter arrived. The sender was the American company GO Corp., an OS maker who worked with portable telecommunications equipment. Inside the letter was a job offer. GO had prepared a special position for Kawasaki as supervisor of technological development in Japan. The proposed salary was astonishingly high.
The SH-1 wasn’t selling. The negotiations with Sega had come to a standstill. And above all, Kawasaki was dissatisfied with his treatment at Hitachi. “I felt that my work wasn’t appreciated by the company,” he said. He decided to accept GO’s job offer.
Three years before, Kawasaki had made the choice to switch companies to Apple Computer Corp., but he gave up the idea before telling anyone else. This time was different. Without delay, Kawasaki informed his boss and coworkers that he was accepting the offer from GO.
An Unexpected Call
Kawasaki had one month remaining at Hitachi. At that point, he was no longer meeting with potential customers, and Sega was no exception. Kawasaki’s coworkers knew that he was soon leaving the company.
One day, Kawasaki was unexpectedly summoned by Tatsushi Itou, the assistant manager of the Hitachi Semiconductor Design and Development Center. At that time, Kawasaki did not know Itou very well and was not sure why he was being summoned.
When Kawasaki arrived, Itou began by saying, “I want to improve your treatment at Hitachi.” It seemed that Itou had seen through to the true cause of Kawasaki’s dissatisfaction. Itou told Kawasaki that he valued all of the work Kawasaki had done, and that Kawasaki was vital to the future of Hitachi’s microcontroller business. “I didn’t think there was anybody like that at Hitachi,” Kawasaki said. The unexpected talk with Itou left a deep impression on Kawasaki, and he changed his mind about leaving the company just in time.
An Abrupt End
In the fall of 1992, Kawasaki, feeling fresh after letting go of the idea of changing jobs, once again visited Sega. He brought with him a graph comparing the performance of the SH with other companies’ microcontrollers, and he was determined to not give up until everything was settled.
Kawasaki received his answer sooner than expected. As soon as he met with Hamada, he was told simply, “Oh, that’s OK. We’ve already decided to go with the SH.” Kawasaki felt the energy drain from his body as the long negotiations came to an abrupt end. It had been decided. The SH would be included in Sega’s next home console.
After the decision by Sega to adopt the SH, the full-scale development of the second SH chip, the SH-2, began in January 1993. Jun Hasegawa, who had just returned from studying at Carnegie Melon University in the U.S., joined the SH development group.
However, more than a few people who had taken part in the SH-1’s development were confused upon hearing of the SH-2. Up to that point, Hitachi had been making a new microcontroller about once every four years. The SH-1, however, had only been in mass production for a few months, so many were perplexed about why the next microcontroller was already being developed. “There was a lot of ‘What?! Are we really making the next one already?’” Takaki Noguchi recalled.
Once development began, however, it proceeded quickly. After having made the 16-bit multiplier for the SH-1, Noguchi and a coworker spent two months rebuilding it as the 32-bit multiplier for the SH-2. They spent their days off designing the synchronous DRAM interface circuit. “From May to July 1993, we worked without taking a single break,” Noguchi said.
A Request from the Research Lab
One day, the SH development group received a written message from Nobuyoshi Doumen, the head of the System Development Lab. It read, “Did anything ever come of the multiprocessor function that we discussed before?”
A few months prior, Doumen had requested the inclusion of a multiprocessor function in the SH-2. At that time, the lab was researching the development of a digital assistant for reading news articles code-named “Yajiuma Shimbun” and wanted to use the SH-2 in a multiprocessor configuration. However, the SH group did not want to include the function just for a research system and had been pretending to overlook the request.
When the second request came, they were unable to continue to ignore it. They were indebted to Doumen for a previous problem with the number of registers in the SH that he had helped them to overcome. They decided to include a rather simple circuit that would allow the SH-2 to be used in a multiprocessor configuration. “In my mind, I thought that certainly nobody would ever use that function,” Kawasaki recalled.
The Secret to Improved Performance
In the summer of 1993, a slight incident occurred. Sega stated that the performance of the SH-2 (25 MIPS) was insufficient for a next-generation home console (this was right around the time when Sega’s biggest rival Nintendo announced that they were including a 64-bit CPU as well as a graphics processor jointly developed with Silicon Graphics, Inc. in their next console, the Nintendo 64). They wanted to increase the performance of the SH-2 by raising the frequency. However, to do that, it would be necessary to re-examine the chip design, and the SH development group did not have the time remaining to do that.
The decision of what to do was left for the top-level meeting between Hitachi and Sega executives that took place in Hakone in September 1993. The SH group had prepared a secret plan to resolve the “performance improvement problem.” Their solution was stated as follows: “If we use the multiprocessor function that is included in the SH-2, we can operate two SH-2s linked together. This should satisfy the request for higher performance.” Nobody had expected that the multiprocessor function, which they had been reluctant to include, would prove to be this useful.
In this way, Sega’s next console, the Sega Saturn, came to be equipped with two SH-2s.
Advancing to Second in the World
In June 1994, mass production of the SH-2 began.
By July, the monthly production output had reached 200,000 chips. Of course, these would supply the Sega Saturn, which would be released in the fall of that year.
By March 1997, a total of 7.56 million Saturn units had been manufactured. In other words, approximately 15 million SH-2 chips had been made for the Saturn. With this, the SH had suddenly risen to the second-most-selling high performance RISC microcontroller, and its name became known throughout the world.
However, in terms of sales, the outcome was by no means favorable. Even though the Saturn was equipped with two SH-2s, that did not mean that profits simply doubled. In fact, the profit per unit decreased. Due to being the first sale, they were unable to increase the profits to a satisfactory level.
The only solution was to decrease costs as much as possible. From April 1994, before the start of mass production, work had begun on reducing the size of the SH-2. After that, a special chip was made for the Saturn that combined two SH-2s into one. This effort paid off, and profits finally rose to a healthy level.
The Real Challenge Lies Ahead
More time is required before the SH will achieve permanent status as “the world’s microcontroller.” In order to increase sales, it will be necessary to quickly grab more users beyond the Saturn. Some success has already begun to appear. The SH-3, announced as the successor to the SH-2 in 1996, is being included in various personal digital assistants running on Microsoft Corp.’s Windows CE OS. However, as of now, it has not yet become a big hit.
By Hiroki Eda