While chip company business dealings and endless research into nanometer transistors can sometimes seem worlds apart, here are some consequences of how engineers' battle with Moore's Law directly affects the industry.
AMD is basking in the glow after convincing the investment community that its semiconductor business has made significant inroads against rival Intel, which has struggled in recent years to make good on promises to ramp up production of next-generation chipsets.
AMD shares soared to a 12-year high on Wednesday after several Wall Street analysts raised price targets, citing the company’s advances towards releasing new 7 nm chips that will have higher density than existing products for the first time in years.
Of Chip Fabrication and Market Projections
AMD late last month said it would shift its 7 nm business, including the Zen 2 CPU and the Navi GPU to TSMC (the Taiwan Semiconductor Manufacturing Co.), dealing a major blow to GlobalFoundries, which had been a longtime partner.
As a result, GlobalFoundries said it would place its 7 nm FinFet business on indefinite hold, slash jobs, and restructure the business, while shifting resources to its 14/12nm FinFet platform. CEO Tom Caufield said in the announcement that the vast majority of fabless customers were “looking to get more value out of each technology generation” and “fewer fabless clients designing into the outer limits of Moore’s Law.”
These decisions by chipmakers make changes in the market. Analysts at financial groups Jefferies and Cowen raised their price targets for AMD to $30 a share, while Bank of America Merrill Lynch later issued a new report raising its price target for AMD to $35 a share. AMD shares fell 2.35 percent to $27.84 on Thursday, following its 12-year closing high Wednesday at $28.51. Shares of AMD have risen more than 170 percent year-to-date.
By contrast, Intel has been plagued by delays involving its 10nm Cannon Lake processors, which are now on hold until holiday 2019.
Intel's group president of Manufacturing, Operations, and Sales, Stacy Smith, holding a10 nm Cannon Lake wafer in September, 2017. Image used courtesy of Intel.
“AMD has accomplished the reversal of fortune by intelligently building a chip that had multiple uses,” Kevin Krewell, principal analyst at Tirias Research said. “One die by itself makes an enthusiast PC processor, 4 die in a package make a killer server processor. AMD added more memory and I/O to its processors, while Intel was trying to push customers towards more expensive server processors.”
Krewell points out that AMD had the flexibility to use multiple foundries to build its processors, while the advantage Intel had by using its own fabs turned into a disadvantage as the company had difficulty moving beyond the 14nm node.
However other industry observers are pushing the pause button on declaring AMD the new standard bearer just yet.
Understanding the Stock Market vs Real-World Design Decisions
Alan Priestly, research director at Gartner, notes that Intel is still the much larger of the two x86 CPU vendors and says that while AMD’s Epyc appears to be gaining traction, the expectation by the end of 2018 is a 3-5 percent market share.
“AMD 7nm CPUs are 2019 and Intel’s 10nm are most likely 2020, but the debate is not really about process tech but workload performance as IT organizations do not buy servers on the basis of the CPU semiconductor process; they buy products that meet (or exceed) their workload demands,” he said.
He said if AMD’s utilization of the 7nm process and the related architectural enhancements of its Zen2 design result in much better performance than what Intel has in the market, AMD will gain business from Intel.
AMD CEO Dr. Lisa Su revealing their first 7nm GPU at Computex 2018. "We made a bet on 7nm," Su said. Screenshot used courtesy of AMD.
“If the performance advantages are marginal, AMD will have to rely on product differentiation (memory density, PCI channels, etc.) to win new business," according to Priestly.
While AMD’s performance (specifically those of its shares) has been exceptional, numerous claims that the company has Intel ‘on the ropes’ are lunacy in real-world terms,” according to Charles King, principal analyst at Pund-IT.
He points out that Intel’s quarterly revenues, at between $16 billion-$17 billion, are more than ten times that of AMD, earnings per share 93 cents to six cents and the company had ten times the amount of cash on hand.
AMD, Intel, and Everyone Else
The competition for more processing power goes well beyond these two competitors. In June 2017, IBM announced its Research Alliance partners, including Samsung and GlobalFoundries, had created a process to create 5-nanometer chips using silicon nanosheet transistors.
The new technology, announced less than two years since their 7 nm breakthrough, would boost density from 20 million to 30 million transistors on a fingernail-sized device.
A working chip with IBM's 7nm transistor. Image courtesy of IBM/Darryl Bautista
The processed used EUV (extreme ultraviolet) lithography to adjust the width of the nanosheets, something IBM said was not possible using the finFET process.
How closely do you watch the stock market for industry news? Do you believe the stock market is important to engineers to follow or is it far removed from a typical EE's job? Share your thoughts in the comments below.
No comments:
Post a Comment