CNET también está disponible en español.

Ir a español

Don't show this again

Desktops

Return of the prodigal Sun

After a nine-year absence, Andy Bechtolsheim returns to Sun Microsystems, the server maker he co-founded in 1982.

SAN FRANCISCO--Nine years after leaving the server maker he co-founded in 1982, Andy Bechtolsheim is returning to Sun Microsystems.

Click here to Play

Scott McNealy, CEO, Sun Microsystems
At Sun, Bechtolsheim will pick up where he left off: designing computers. This time, he'll be working to share more technology between servers that use Sun's UltraSparc processors and those using Advanced Micro Devices' Opteron processor, he said in an interview after a radiant Chief Executive Scott McNealy welcomed the co-founder back during Sun's analyst conference here.

Not surprisingly, Sun executives were much more enthusiastic in talking about Bechtolsheim's return than about the 2003 departure of another co-founder, Bill Joy. "Andy is going to be a resource for every designer at Sun," McNealy said.

Bechtolsheim, 48, left Sun in 1995 to start Granite Systems, which built 1-gigabit-per-second networking technology and which Cisco acquired in 1996.

Everybody has a little more gray hair, including myself. But we got more efficient with age. The new era here is nonreligious.
He'll rejoin Sun through its planned acquisition of Kealia, a start-up that builds Opteron servers. Although Bechtolsheim co-founded Kealia in 2001, he didn't formally join it until December, when he left Cisco Systems.

"It's pretty cool to see two pioneers reunite," said Robert Frances Group analyst Ed Broderick, referring to the two executives. McNealy, he said, is buying not just Kealia's intellectual property but also "a confidant."

Although executives remained mum about Kealia's products and how they'll fit in at Sun, Bechtolsheim shared his thoughts about his sudden homecoming with CNET News.com.

Q: Why did you leave Sun?
A: Back then, I saw an opportunity in gigabit Ethernet. I founded a company Cisco ended up buying. Now, I see an opportunity in servers, of all things. I think that what has been missing to ignite the market is the next level of cost performance. As these things come to market, (customers) will upgrade.

What's different about Sun that you would rejoin the company?
Everybody has a little more gray hair, including myself. But we got more efficient with age. The new era here is nonreligious. There are incredible opportunities with Sparc-Solaris and with Linux-Opteron. It's not exclusionary. That's what made it attractive to me. In the past, Sun just sold UltraSparc systems, but customers, in the end, want other solutions.

What will your role be at Sun, and what will you do?
I'll be chief architect of the Volume Systems Products group. I like efficiency. I think that I can help by making things faster with a larger group. Neil Knox's Volume Systems Products group is one of the smallest at Sun. This is all about execution. There's going to be one chip coming after the next.

At Cisco, you were a networking technology executive, and now, Sun is bringing more networking into its gear, with its blade servers and with its planned acquisition of Nauticus. Will you continue that network work at Sun?
No. That was my previous job.

Will you be picking up the screwdrivers and soldering irons and building new servers?
I like to work with the engineers and will do so at Sun. Part of efficiency is that you don't have to reinvent the wheel. Besides continuing with Opteron work, I'll also closely work with other parts of Neil's organization. There's a great opportunity to accelerate Sparc.

You said today that Kealia's systems will be certified to run Windows as well as Linux and Solaris. Sun won't touch Windows servers with a 10-foot pole. What do you think of selling Windows servers?
It's really hard to add value to Windows. Whatever you add, Microsoft is going to take away from you.

Linux has opportunities. There are many ways to add value. There are lots of things Sun customers expect that are missing from Linux.

What is Opteron's performance? Is it easier to design servers with Opteron than other chips?
Opteron has the industry-leading benchmarks, except for floating-point mathematical calculations, for which Itanium is faster. If you look at SPECweb99 or CINT2000, Opteron is the fastest CPU. Advanced Micro Devices did really well.

What about Opteron system design?
It's fairly straightforward. But you still have to make the right choices, whether it's two-way (a two-processor server), four-way or eight-way.