That was Mohammad Ali talking about his epic boxing match against Joe Frazier in 1974, known as the Thrilla in Manila, in which the two great fighters attacked and mauled each other for 14 rounds before Ali came up victorious.
Writing for Sports Illustrated, Mark Kram, in his inimitable post-match reportage wrote: “A couple of miles away in the bedroom of a villa, the man who has always demanded answers of Ali, has trailed the champion like a timber wolf, lay in semi-darkness. ‘Man, I hit him with punches that’d bring down the walls of a city,’ said Frazier. ‘Lawdy, Lawdy, he’s a great champion.”
Just three years before that, in 1971, Frazier had become the first man to knock out Ali at the Madison Square match.
In the semiconductor chip-making industry, Intel has been the Ali to AMD’s Joe Frazier since 1968. For most of the last 50 years, Intel has continuously hammered AMD or Advanced Micro Devices. It is a miracle that AMD has even managed to survive the pummelling. But in the last four years, AMD has staged a comeback.
CEO Lisa Su deserves much of the credit for AMD’s turnaround. When she took charge a decade ago, AMD’s stock was languishing at around $3; today, it trades at $125 per share. In 2014, its share in the data-centre chip market was miniscule; today, AMD has a better margin growth and is second only to Nvidia in graphics processing units (GPUs). Intel, meanwhile, is trading at book value and is an acquisition target as well.
Discover the stories of your interest
ETtechSo, what has Lisa done which proved so effective? “What I noticed was when she killed a bunch of stuff and really focused the company on its core strengths. She killed older architectures like Bulldozer and focused more on Zen. And then she focused on long-term road maps under Zen to provide consistency and confidence on the road map to stakeholders. Those two things seem to be the architectural advantages that AMD has now doubled down on,” says Hemant Mohapatra, a former AMD engineer, and now a partner at Lightspeed, a venture capital fund.
Siamese twins? Well, almost
Intel and AMD are almost the same age. Intel was founded in 1968, AMD a year later. Intel was founded by Silicon Valley engineering royalty of those days, the stuff of legends now. Robert Noyce (co-inventor of the integrated circuit), Gordon Moore and Arthur Rock. Noyce and Moore had worked with William Shockley, co-inventor of
the transistor, the fundamental building block of any computing device today.
AMD, meanwhile, was started by Jerry Sanders, a salesperson and engineer. It also had some of the best chip engineers who had worked at no less a company than Fairchild Semiconductors, the mothership that launched most US microprocessor companies because of the engineers who learnt their trade at the company.
Until 1978, nothing very significant happened at either company. The microchip design and manufacturing development were still very nascent. It all changed when Intel launched the 8086 microprocessor, which could handle calculations faster. It also had an ability — pre-fetching — to access programming instructions ahead of time. That’s exactly what IBM, the computing pioneer, needed to run the new computer applications like spreadsheets and word processors.
In the coming years, Intel’s chip and Microsoft’s Windows combined to put a lock on the PC market that would not be broken for nearly three decades. “Intel has been historically dominant, and we attribute that to their prior dominance in semiconductor process technology. Intel had the best-in-class manufacturing technology for the
exclusive use of their CPU products that usually meant their CPUs were the best in terms of performance,” says Myron Xie, an analyst at Semianalysis, a semiconductor- research outfit.
Soon enough, Intel had 90% of the market share of computer microprocessors and it is a surprise that AMD managed to survive on low-margin, lower-end PC manufacturers. “It (Intel) never really tried to put AMD out of business. When you have 70% to 90% market share, you can put your competitor out of business, right? You just cut your prices, but Intel knew that if they did that, they would be in Microsoft’s position with the US government and antitrust regulators would come to break them up. So, they had to, they almost had to, let AMD live,” says Dean Takahashi, lead writer for GamesBeat at VentureBeat, who has tracked the semiconductor industry closely.
This meant AMD went “as low as 10% market share, though it has now gone back well into the 30% market share”, Takahashi shares.
ETtechA different tack
It is a well-known fact that challenging a market leader with more than 70% of the market share is incredibly difficult.
To try and break this cycle, AMD took enormous engineering risks and tried to push the boundaries of hardware innovation. It developed chips that had multiple cores — CPUs — and which supported multi-threading, essentially allowing different parts of a software program to be processed simultaneously because of multiple cores.
Unfortunately for AMD, such complex architecture created more problems than it solved. A chip architecture called the Bulldozer (2011) was the clearest example of this mess; Bulldozer wasn’t really optimised for Windows.
And it wasn’t as if AMD’s engineers weren’t capable. They were some of the sharpest in the business. “Jim Keller was amazing. He’s now the founder of Tenstorrent. Mike Papermaster was amazing. Ben Sander was amazing. So was Dave Christie, one of the smartest silicon-design engineers in Silicon Valley,” says Mohapatra.
A few of them were so committed to AMD that they could not see the company perish. “Another important engineer is Mike Clark, who was (and still remains) the Zen lead architect and has been with AMD for life. In terms of Jim Keller, it seems that he rejoined AMD in 2012 to lay the foundation for the Zen architecture because of AMD’s struggles and a desire to help turn AMD’s fortunes around, since Keller has said that he wanted to save the company from bankruptcy,” says Xie.
Intel’s mistakes
Since we covered one sporting metaphor in the form of boxing’s greatest to explain Intel’s dominance, here’s another to explain AMD’s resurgence. Anyone who’s seen a champion like Rafael Nadal play against a lower-ranked player knows that in order for the lower-ranked challenger to pull off an upset, Nadal needs to make a few mistakes.
Something similar happened with Intel with the Titanium chip. While AMD’s chip with similar specs worked with applications already in the market, Intel’s could not. Intel also passed up the smartphone opportunity because that market did not have the margins that high-end servers offered. And lastly, it could not anticipate the rapid rise of cloud and then artificial intelligence (AI), both of which needed GPU class of chips.
To its good fortune, AMD had made an acquisition of ATI Technologies in 2006 that gave it a presence in the GPU market. After Su took over, she made sure of two things. “The other thing she did was to really diversify away from Global Foundries (chip manufacturer), which lagged behind TSMC. Finally, AMD found a way to make the best of the ATI Technologies acquisition. It took 10 years, but it finally got there,” says Mohapatra.
Su was the one who got the company to focus on the Zen chip architecture. “The Zen architecture was a game changer for AMD’s fortunes and returned the company to being competitive with Intel’s products after a long period of time,” says Xie.
Different dream, same reality
More things change, the more they remain the same – kind of. Lisa Su has managed to turn things around, but AMD is still the number two player in the GPU market. Nvidia is the dominant market leader.
And like Intel did through the 1980s and the 1990s, Nvidia is at the absolute cutting edge of GPU chip architecture. AMD will not want to fall behind too much so that it is boxed into low-margin customer segments like the last time.
The technology sector does not offer too many second chances to those who cannot keep up with the market leader.
ETtech
