Early Sun Microsystems—You bet the company

(1983, age 36)

By the spring of 1983, I had just finished the Tower 16/32 project for NCR. I was the Senior Systems Architect for that project, which meant that I was responsible for a design that would meet NCR’s stringent performance and operational requirements. That project was based on the Motorola 68k microprocessor, which was a cheaper, faster, and better design than anything NCR had built before. It was intended for business applications so reliability was important. I was proud of the “Tower,” as we called it, and of the small, dedicated team of about 15 people that built it.

I was aware that there were other companies trying to build computer systems based on microprocessor technology. One of them was Sun Microystems out in California. What they were doing intrigued me. It was another small team. They wanted to build a workstation for scientific and engineering applications. I took a look at the list of people then at Sun and knew immediately that that’s where I needed to be. Here is the list (click to enlarge):

SUN, 1982

I knew some the people on the list because the world of microprocessor-based designs was a small community. I agreed to fly out and meet them in Silicon Valley at the end of March. During the visit it became clear to them that, though they had a workstation design that worked, they weren’t sure if it could be produced in volume. What would it take to build a company around the Sun design? I had just completed doing just that for NCR with the Tower. They said they knew they needed me and weren’t going to let me leave the building without a job offer. Mine was an easy decision and I accepted on the spot.

Two weeks later I would be at COMDEX wearing an NCR trade-show badge for the last time as we introduced the Tower 16/32 as a general purpose computer. It would go on to become the biggest money-maker in the history of NCR.1

trade show badges

Two weeks after that, I would be at another trade show, the National Computer Conference, wearing a Sun Microsystems badge. The decision had been made and my career was now tied to a small company occupying the top floor of one building in Silicon Valley.

Sun had built its first product, the Sun-100, from Andy’s original processor design. That design worked if all the critical integrated circuits were near the center of their operating parameters, if the power supply voltages were nominal, if the temperature of the board was controlled within a narrow range, and if so many other things were just right. It worked in a lab environment, but as that design was migrated into the Sun-120, its unmanufacturability for the real world became apparent. Sun simply couldn’t produce the 120 or the server version, the 170.

Scott McNealy, the CEO of Sun, came to me privately and we discussed how serious the problem was and what could be done about it. We decided to bring the entire company together in one room and announce a “do or die” get-well plan. I recorded that meeting. In it, Scott says:

The situation we are facing is that the world knows we have the 120 and the 170. They’re not really interested in the 100 and the 150 anymore. In fact after this month we have three or four units to ship next month of the 100s and the 150s plus some OEM stuff that’s out over time, so basically this month after we ship the seventy-five units or so we are pretty much done shipping if we can’t get the 120 going.

The biggest fear was that the world would decide that we could not produce a commercial product. Andy had come from Stanford, Bill Joy from Berkeley. What did graduate students know about making a commercial product? There were other companies out there that the OEMs and end users could go to if Sun lost their orders. Scott continued:

We’re betting the company on this product that’s sitting on the floor right there throwing out parity errors and disk errors and that sort of thing.

It was clear that it was a technical problem and that the solution, if there could be one, would fall to me and my new engineering team. When asked who would be in charge of what area, Scott replied: “I’ll be in charge of extracurriculars like pizza and making phone calls… Roger is in basically in charge of the techincal. He’s making the technical calls…” At that point Vinod Khosla “V.K.” switched on the hot chocolate vending machine.

A question was then asked: “Is the logical division hardware, software and then manufacturing?” Eric Schmidt responded: “I’m taking my marching orders from Roger.” At that point Scott made the full scope of responsibility clear: “I would like to put hardware and software … reporting to Roger.”

For a few weeks, everyone at Sun was focused, day and night, on one purpose: building a workstation that could be produced in volume and that would operate flawlessly over a range of real-world parameters. They had hired me believing I could help them make a commercially viable product out of a research project. By this time, I had already brought in my most trusted engineer from NCR, Jim Lockwood. We were looking at the existing design carefully. On an analysis I requested, Jim attached a sticky note to one page that read: “Column address setup violation is so bad that I am unsure of my numbers. Please independently derive this one. JML” Column address setup is a timing relationship measured in billionths of a second deep in the heart of a computer that has to be just right, every time, millions of times every second for the computer not to crash. We found several situations like this in the original design and made corrections quickly and effectively.

Sun moved from being an interesting university experiment to being a real company.2 But it wasn’t the last time I would find myself in the middle of a “bet the company” situation requiring a technical solution. We solved the internal challenge with the design of the Sun workstation. Soon an external challenge would be faced in what would be a winner-take-all battle with our biggest competitor, Apollo Computers.

Supplemental Material

“A Lot, plus or minus 10%”

As Manager of Hardware Engineering, for any new product I would hold a design review. Peter Costello had designed a new video card for the Sun workstation. Part of the design review was to convince other engineers that the design was manufacturable. I invited Bill Joy to these reviews even though his expertise was primarily in software.

“All you need is two bits”

In the early days of Sun Microsystems, people designed computers. Later, computers would design computers. But back then, decisions made in design reviews were implemented in circuitry on printed circuit boards. We had to understand it, design it, and make it work. In this tape, Bill Joy questions a design as being over-complicated.

“Not so pretty news…”

Jim Lockwood is explaining a timing design for a memory management unit. Pam King joins us with news about integrated circuits we need for our computer design. Pam could always find a way to get things done. She was valuable then and later would be even more mission-critical to me on the High Performance Workstation project that would come later.

“All I need … is answers”

In the early days, there were fewer than 100 of us in the company.3 It was the most amazing ongoing collaboration imaginable. It’s what made being in a startup so exciting. Every day was mission-critical. Everyone had to rely on everyone. We worked hard, but we also had fun. In this tape, after faced with an Engineering Change Order request, I speak to Alison Shanks who is working tirelessly on documentation for the Sun 170.

A few dumb looks…

Alison had an M.A. from Stanford. She was a very intelligent, very capable young lady. So many people were over-qualified for what they were doing but it didn’t matter. We were all in it together and we all did whatever it took. In this email snippet, she refers to sticking those “funny little bugs into boards.” Those were integrated circuits, the building blocks of the early computers. In the earliest days, when a batch of printed circuit boards would come in, we would gather everyone from upper management to administrative assistants into a conference room to insert the integrated circuits, the “funny little bugs,” into the boards.

  1. From Wikipedia: In 1982, NCR became involved in open systems architecture. Its first such system was the UNIX-powered TOWER 16/32, the success of which (approximately 100,000 were sold) established NCR as a pioneer in bringing industry standards and open systems architecture to the computer market.

  2. Sun was successful and so were many of the company founders whose voices are heard on the audio clips. As of 2015, Scott McNealy is worth $1 billion, Bill Joy $1.5 billion, Vinod Khosla $1.6 billion, and Eric Schmidt is worth $9 billion.

  3. The kind of person that made Sun work when there were a handful of us wasn’t necessarily a good fit as the company grew. By 2008, Sun had 34,900 employees and most of the early contributors had moved on, as had I. Though we were gone from Sun, Scott didn’t forget us: we learned that the first 100 employees of Sun, including me, were always guaranteed a job with the company if we needed it.