A Candid Look at AMD to Celebrate Its 50th Anniversary

This week, AMD celebrated the 50th anniversary of its founding. Like Intel, AMD was conceived by Fairchild Semiconductor employees. Robert Noyce and Gordon Moore left that company in 1968 to found Intel. Jerry Sanders left Fairchild in 1969 and founded AMD.

I’ve covered many companies over the past 18 years, but none of them have inspired the kind of passion AMD does. People love a David-and-Goliath story with a scrappy underdog, and the AMD-versus-Intel narrative has delivered one for decades. Over the course of my career, I’ve seen AMD ascend to dizzying highs and suffer crushing defeats. Few companies have seen such reversals of fortune or survived being driven to the brink of collapse on multiple occasions.

AMD can fairly claim to be one of the handful of companies that have shaped the computing world in fundamental, transformative ways. In the late 1990s, Intel declared that the x86 architecture would not transition to 64-bit designs. Instead, a fundamentally new approach to designing CPUs, known as EPYC/IA-64 and launched with Itanium, would replace it. AMD, in contrast, announced that it would extend the x86 architecture into the 64-bit realm. Windows still refers to “AMD64” for its 64-bit architectural support for this reason, though the neutral label x86-64 was also adopted.

Itanium CPUs proved to be dreadful at running legacy software and extremely difficult to program and optimize. x86-64, in contrast, offered a smooth upgrade trajectory for consumers, as well as faster 32-bit performance than previous x86 processors. This was probably the biggest single ‘win’ for AMD over my entire career. A victory for Intel would have cut AMD off at the knees; widespread support for IA64 would have doomed its AMD64 initiative. The presence of a competitor forced Intel to add its own x86-64 support (Intel referred to this as EM64T before eventually renaming it Intel64).

In the early 2000s, it wasn’t unusual to see AMD fanboys declaring that Intel would soon be dead, replaced by a triumphant AMD. Instead, AMD would spend nearly a decade struggling to close an increasingly large performance gap. The smaller firm’s antitrust lawsuit against Intel was eventually settled (AMD received a $1.25B payment from Intel), but while that payment probably represented a fair calculation of the profit AMD might have earned on CPU sales during the time period the company alleged Intel had engaged in market tampering and illegal, abusive behavior, it did nothing to close the length of time it takes to develop and bring new technology to market. $1.25B or no, AMD was still running behind Intel.

The firm had enjoyed two distinct periods of strong competition with Intel: 1999 to early 2002 (the K7 era) and late 2003 to mid-2006 (the Athlon 64 and Athlon 64 X2 era). From mid-2006 forward, the Core 2 Duo opened a performance gap Chipzilla couldn’t close. The original Phenom was a hot-running disappointment, and while Phenom II was well-received, it offered strong competition for Core 2 Duo, not Intel’s then-new Nehalem.

By late 2012, it was clear AMD was on the existential ropes. AMD fans had looked to Bulldozer to restore AMD’s competitive standing and Bulldozer was a disaster. Relations with its new foundry partner, GlobalFoundries, were anything but good. If the Xbox One and PS4 hadn’t collectively been tremendous hits in 2013, AMD would almost certainly have died or been sold off for scraps. By Q3 2015, sales in the compute and graphics division had fallen to just $424M for AMD’s entire non-console business. For comparison, AMD just reported sales of $831M in that segment for Q1 2019. That’s a 1.95x improvement in under four years.

AMD made plenty of its own mistakes during the 2005-2012 time period that can’t be laid at Intel’s feet. It took on enormous debt to buy ATI, only to later admit that it had overvalued the company by a factor of two. Company employees have quietly admitted that in 2005-2006, AMD took its eye off Intel more than it ought to have done. While Prescott smoldered in its socket, Intel was preparing a new CPU core that hearkened back to the tremendously successful Pentium 3. When Conroe (65nm C2D) finally dropped in 2006, AMD didn’t have an effective answer for it. The company spent 11 years in the wilderness before bringing Ryzen to market in 2017, reinvigorating its own health and competitive standing in the process.

The Value of Competition

If AMD has demonstrated anything, it’s the tremendous value of having competition in semiconductors. Every time AMD entered a market, from 1999 forward, it meant Intel was forced to adjust its product positioning or price. AMD’s strong budget competition in the late 1990s with the K6 pushed Intel to develop the original Celeron. Prior to this, the company’s strategy was to fill needs in the lower-end computing market with older processors.

The original Celeron’s lack of L2 cache and poor performance in certain tests against AMD’s K6-2 led to the development of the Celeron “A” family — still celebrated as some of the best overclocking CPUs ever built. The K7 and 1GHz race forced Intel’s P3 prices downwards, while AMD’s entry into the server market in 2001 and its Opteron-fueled expansion in 2003 pushed Intel to cut prices on Xeon chips.

This pattern has repeated in the two years since Ryzen launched. Since 2017, Intel has added Hyper-Threading support to Pentium processors and added more cores to the Core i3, i5, and i7 families. CPU prices have come down significantly on chips with more than four cores. Three years ago, an eight-core Intel CPU cost $1,089.

Now, AMD is on the verge of another first. This summer, it will launch CPUs on 7nm technology, decisively beating Intel to volume production on a leading-edge process node for the first time in history. It is unclear what impact this will have on relative adoption rates for Intel and AMD processors, but even Intel CEO Bob Swan has acknowledged that his company faces a tougher competitive environment from AMD this year. Anyone wishing to predict that missing 10nm will be the death of Intel would do well to consult the historical record around Prescott and Core 2 Duo, but there are obvious chances for AMD to gain market share in critical segments where it wishes to do business.

I have seen AMD snatch victory from the jaws of defeat, and defeat from the jaws of victory. It’s changed the trajectory of the entire computing world, yet also managed to land squarely on its face in critical moments. The last two years under Lisa Su have been some of the best in its history as far as CPUs are concerned, but fans of AMD’s GPU business have had much less to sing about, particularly at the high end of the market. Consistent, sustained, full-throttle performance across all market segments is still a challenge for the company as it moves out of the resource-starved Bulldozer era and into what will hopefully be a time of greater stability as far as earnings and profits are concerned. We’ll know much more about how well the company has positioned itself for the future in a few months when 7nm GPUs and CPUs finally hit store shelves.

Happy 50th, AMD. Win or lose, rain or shine, it’s been one hell of an interesting ride.

Continue reading

AMD Buys FPGA developer Xilinx in $35 Billion Deal

The deal, which we discussed earlier this month, will give AMD access to new markets that it hasn't previously played in, including FPGAs and artificial intelligence.

NASA Created a Collection of Spooky Space Sounds for Halloween

NASA's latest data release turns signals from beyond Earth into spooky sounds that are sure to send a chill up your spine.

VIA Technologies, Zhaoxin Strengthen x86 CPU Development Ties

VIA and Zhaoxin are deepening their strategic partnership with additional IP transfers, intended to accelerate long-term product development.

AMD May Allow Custom RX 6900 XT Cards, Launch Stock May Be Limited

There are rumors that Nvidia may not be the only company facing production shortages this holiday season. High-end GPUs might just be very hard to find in general.