Former Intel CEO Paul Otellini remembered

Former Intel CEO Paul Otellini died in his sleep on Monday at the age of 66. His tenure was marked by a significant comeback for the company, dealing with a number of business and technological issues, making the company a massive player in the data center but a fumbled opportunity for the mobile market. 

Unlike his predecessors, Otellini was not an engineer but had a MBA from the University of California, Berkeley. He joined Intel in 1974 right out of Berkeley and spent his entire career at Intel, eventually becoming chief operating officer in 2002 and CEO in 2005, a position he held until his retirement in 2012. 

“We are deeply saddened by Paul’s passing,” CEO Brian Krzanich said in a statement. “He was the relentless voice of the customer in a sea of engineers, and he taught us that we only win when we put the customer first.” 

Otellini’s time as CEO of Intel 

Otellini took over an Intel that was an unholy mess in 2005. Then again, he’d had practice in crisis management. In 1993, he took over the microprocessor products division and helped launch the Pentium processor. One year later, an obscure math bug was found in the chip, and the mainstream media went nuts, grossly overreacting and making the Pentium math bug sound like Skynet. And Otellini was the spear catcher for that mess. 

Under Otellini’s watch, Intel grew from $34 billion in annual sales to $53 billion by the time he left. But those numbers don’t reflect just what he had to wrestle with. Coming off the Craig Barrett era, Intel’s desktop product line was a confusing, chaotic mess. I remember not bothering to upgrade from a Pentium IV because I couldn’t make heads or tails of the product mix. 

On the server side, Intel was dismissive of AMD’s push into 64-bit x86 processors, saying its 64-bit strategy was Itanium. The chief limitation of 32-bit processors was they could access only 4GB of memory, and by the time Windows or Linux booted, half that memory was gone and there was very little room for apps or network connections. 

In 2005, Microsoft published tests on 64-bit Windows Server machines with 8GB of memory and said it was able to consolidate 250 32-bit servers running the MSN network down to 25 servers because the 8GB servers could handle many thousands more connections than the 4GB systems, and those servers ran AMD Opterons, not Itanium or Xeon. 

Before Intel knew what hit it, AMD had a 26 percent market share in servers running Opterons and almost one-third market share in desktops. AMD was first to market with dual core processors, the first with 64-bit and the first to move the memory controller to the die. 

Intel, meanwhile, was distracted because it spent a fortune on a bunch of communication chips that no one bought. To settle a lawsuit with Digital Equipment Corp., it bought an ARM family of processors called StrongARM, and with it the license to modify it. It had no coherent product strategy and didn’t even have a clear platform strategy. 

Otellini turns things around 

Otellini undid everything. He cleaned up the product line and launched the Core brands. He instituted the tick/tock strategy of a new microarchitecture followed by a die shrink that served the company well for a decade. He got rid of all the non-CPU products, like all the communications products, for pennies on the dollar. StrongARM was foolishly sold to Marvell, which turned it into a decent business, but in the long-term that proved a big mistake. 

Intel would get 64-bit religion, as well as multi-core and on-die memory controllers. It helped that AMD screwed up badly with the Barcelona processor. AMD tried to do too much at once and delayed the chip for a long time. AMD was never able to catch up and it fell behind competitively on both the desktop and server, while Intel regained its footing. It’s only this past year that AMD has come back to competitiveness with its Zen architecture that appears to be giving Intel fits. 

On the server side, Intel slowly dialed back the Itanium strategy and let loose the Xeon, which had been deliberately crippled so as to not compete with Itanium. Very quickly, Xeon got faster, more cores and, more important, more functions.

In 2010, Intel launched the Xeon 7000 series that added RAS (Reliability, Availability, Scalability) features previously seen only in the Itanium. Instead of the usual 2-socket servers, the 7000 series supported four- and even eight-socket servers, putting Intel in striking distance of massive RISC servers from IBM, HP and Sun Microsystems. The writing was on the wall for Itanium, along with RISC chips like Sparc and Power. 

The result? IDC puts Intel’s server/data center share at 99 percent. The rest is leftover Opterons, IBM Power and mainframes and maybe an odd Sparc box here and there. With the advent of 64-bit servers came servers with 16GB, 32GB and up to 128GB of memory. This gave rise to virtualization and cloud computing. You can say Intel is Patient Zero in the cloud computing revolution. Then again, AMD might disagree. 

Missing the rise of mobile

The one big technological mark on Otellini’s record was he really blew it with mobile. Beyond not seeing the value in StrongARM, in 2006 he passed on a request from Apple to make a special chip for its secret smartphone project — what became the iPhone. Once Intel realized its screw up, it hastily whittled down an x86 processor into the Atom processor in an attempt to compete with ARM. But despite a massive effort, the Atom never went anywhere.

There were rumors that Otellini was nudged into retirement because of this failure. He left Intel at age 62 when he had three more years before mandatory retirement. Since he retired in 2013, Otellini kept a pretty low profile, mentoring young people and being involved with several philanthropic organizations around San Francisco.

Kyrie eleison.