Home > News content

50 years of the Three Kingdoms history of the chip industry: Intel's loneliness, AMD Renaissance, and NVIDIA in the new era of AI

via:CnBeta     time:2018/9/17 22:11:34     readed:618

This is particularly evident in share prices and market capitalization. In just a few years, Nvidia has achieved the legend of 10-fold shares, rising 16.4 times in five years, 36.1% so far this year (YTD), and slightly less than AMD's 5-year gain (6.8 times), but this year's performance, even in all technology stocks, has been one-sided (177.6%) compared with Intel's 5-year rise of only 0.98 times, YTD. Even the yield of -2.73%.

The most important reason why AMD and Nvidia have performed so well in the past two and a half years is that they have seized the keys of the great age. The ability to compute and image manipulate has led to higher valuation flexibility for both companies in cloud computing and AI, the strongest driver of the surge of the two "graphics cards" companies this time around, as mentioned in the Wall Street story "Crazy AMD: Anti-Super Intel, Bite Invida!"

Just as Amazon has seized the key of the era of retail business, WAL-MART has "sunk"; Ali andtencentSeize the key of the IM era, Baidu "sink". In the face of more than 10 years of industry opportunities, the rise of AMD and Nvidia, and the "downfall" of Intel, is forming a clear trend of high and low misalignment, and is likely to be just beginning.

Yahoo Finance of Nvidia, Intel and AMD over the past five years (Deep Blue AMD; Light Blue Intel; Purple Line Nvidia)

Yahoo Finance (Deep Blue AMD; Light Blue Intel; Purple Line Nvidia)

01 back to the original: 1968-1985 Intel ascended the throne, AMD sat second chairs.

Intel, founded in 1968, and AMD, founded in 1969, are co-founders of Fairchild Semiconductor.

Intel's founders include Gordon Moore's proponent of Moore's Law. Robert of Moore (Gordon Moore), "the Father of Integrated Circuits". Noyce (Robert Noyce), and AMD were founded by (Jerry Sanders). Sanders, a sales executive at Fairboy.

The different backgrounds of the founders determine that the former emphasizes technology and R&D (which naturally appeals to capital and industry's best talent) and the latter emphasizes different genes for marketing. It took "5 million minutes to pull in $50,000" (Saunders recalls) for AMD to follow the technological pace of industry leaders, be honest learners and replicators, stay on top of second-largest suppliers at low prices and cost-effectiveness, rather than "5 million minutes to pull in $5 million" for Venture Capital Intel. Set.

The age of 02 CPU: the eldest brother has a big fight.

(1) 1985-1995 years Intel and AMD ripped off lawsuit;

Intel AMD shares 1985-1995 (Yahoo Finance) (deep blue line AMD; light blue line Intel)

Intel's tendency to eat meat and AMD, the "second largest supplier," to drink soup in thighs continued for 20 years, until the two families were caught in the 1987-1995 Century Litigation War.

In 1982, Intel was forced to cooperate with AMD and signed a five-year technical cooperation agreement to authorize AMD to produce its own X86 Series in order to win the IMB PC outsourcing order.processor. However, Intel soon discovered that the AMD that technology was licensed to "technology is missing" threatened its position, and in 1985 it unilaterally tore up the technical cooperation agreement. AMD kicked itself out to make the most advanced 386 processor at the time, and AMD just launched its own AMD386 processor.

At this time, AMD, is not a poor and white enough to have thighs, IBM's list of Intel and AMD meant to make an important bucket of gold, at least enough to fight a barrel of gold, so AMD will Intel sued the court, the two giants began as long as eight years of protracted litigation.

In 1995, although AMD won the lawsuit, it was dragged by Intel's "malice" in 386 of the technology era. By the time the lawsuit was settled and AMD could legally sell its own AMD 386 series, Intel had already launched the 586 Pentium series of processors that hoisted the 386 as early as 1993.

In the third years since its establishment, NVIDIA, the rising star, has just launched the first multimedia accelerator NV1. "Small Transparency" Nvidia is also a chip company, but clearly put forward its vision is to develop a 3D video game chip dedicated to bring more realistic visual experience to the player.

(2) 1995-2005 lawsuit finished Intel and AMD frontal rigid technology; transformation of GPU

After burning tens of millions of dollars in litigation fees and being dragged down by Intel's procrastination tactics, AMD is determined to compete head-on with Intel in research and development, and is no longer so passively subjected to technology.

Intel AMD shares 1995-2005 (Yahoo Finance) (deep blue line AMD; light blue line Intel)

Thus, from 1997 to 2005, the semiconductor market appeared AMD and Intel (you) foot (chase) phase (I) to cheat fast iteration phenomenon.

  • In 1997, Intel stole chickens without eating rice: to crack down on competitors (mainly AMD) who relied on Socket 7-compatible sockets to solve their own fake processor-Intel platform incompatibility problems, it launched incompatible Slot1 sockets and Pentium II, which adopted the Slot 1 socket standard; this step was a "military adventure" (Forcing consumers to replace the fake motherboard, otherwise unable to upgrade the CPU to Pentium II), AMD "see the move" as a good opportunity to grab market share, taking advantage of the introduction of Socket 7, cheaper than the Pentium II K6 framework. The sale of K6 provides AMD with a positive financial and market base for Intel.

  • In 1999, Intel silently turned the page on incompatibility, with the new socket standard Socket 370 compatible with Socket 7 and the release of the latest Pentium III processor

  • In 1999, AMD stared at the Pentium III, launching a more powerful Athlon CPU, the first time in history to surpass Intel in product technology.

  • In March 2000, AMD first released the first 1GHz CPU speed dragon 1000.

  • By the end of 2000, Intel launched 1.4 and 1.5GHz's Pentium 4. But quickly discovered by consumers, Pentium 4 is not as good as AMD's speed dragon series, but also expensive.

  • In 2003, AMD released speed dragon 64. Intel continued to respond to 3.8GHz's Pentium 4 with higher frequency.

  • In 2004, Intel was forced to start.DevelopmentProducts for the standard speed dragon 64

  • In 2005, Intel grabbed a dual core CPU Pentium D. A week later, AMD released dual core speed dragon 64*2. Melon consumers find that the Dragon continues to beat Pentium Dual Core.

(3) in the past few years, eldest son has been on the way to climb the pit.

In 1996, after three years of groping, young Invid shifted its focus to graphics processors. Then it began to go smoothly from 1996 to 2000. Invid's work in the graphics processor market began to run smoothly, and the entire game factory placed orders one after another. In 1999, its revenue was broken 150 million US dollars and smoothly listed on NASDAQ. Graphics processing, 3D technology in this subdivision industry in addition to ATI is invincible.

However, Lin Dong said that it was there. In the next 2000-2003 or three years, NVIDIA hit the hard time.

  • Failed to meet the deadline.MicrosoftAXboxWith the order of graphic chips, Microsoft turned back to the embrace of ATI.

  • In the future, the flagship product Geforce series will not stop exploding quality accidents at this time. After many spontaneous combustion failures, Yingwei CEO Jason Huang likes to mention the title of "two bombs and one decoration"

  • So far, Huang Renxun, known for wearing a leather jacket at the press conference, on the one hand humiliated and burdened Microsoft by pleading soft and pleading for orders, on the other hand, Intel reached a patent cross-licensing agreement.

The balance began to tilt, Intel rolled AMD; AMD was forced to fight on two fronts; Invid's high-performance graphics product

Yahoo Finance (Deep Blue AMD; Light Blue Intel; Purple Nvidia)

In 2005, Intel's Tick Tok was born. Gordon Moore, 76, set the timeline for Intel's processor microarchitecture and chip fabrication updates (following the Moore's law that transistors on a chip double every 24 months, that is, the performance of the semiconductor industry doubles every two years).

AMD also had OPTERON in 2003.The serverIt can surpass Intel in manufacturing, but with the pendulum plan, AMD will technically stall, unleashing a decade of grief that has been suppressed by Intel: product lines, prices and brand image will have to continue to sink.

In 2010, because of the name of the Bulldozer architecture and the similarity between AMD and ADM, an agricultural manufacturer, AMD, which has been nicknamed "farm (private) enterprise" (industry), has declined to the point where it is necessary to grind the corner of the display card manually to solve the heat dissipation bug. One hundred thousand HD6970 graphics cards that have been delayed since their launch date have been nicknamed "purely hand-made" by players. This God operation that shocked the jaw of global players has laid the title of AMD's "agricultural enterprise".

In 2011-2015 years, AMD's continuous losses were on the verge of bankruptcy.

In 2006, AMD made a strategic decision to buy ATI, Nvidia's arch-rival in GPU, and initiated a new solution for CPU integration. However, the purchase of ATI meant that former comrades AMD and Nvidia became rivals, and AMD began the bitter days of fighting Intel on the CPU line and Nvida (and ATI indigestion) on the GPU line.

AMD was not the only one who was caught in the attack. In 2006-2012, Nvidia suffered a series of "N-shocks" from AMD, Intel stopping patents to cross-share its sleeves into 3D graphics accelerators, and a huge human and financial effort to develop a CUDA (Universal Parallel Computing Platform) that burned money without short-term returns, with its share price falling from $37 to $6.

In 2012, with the breakthroughs in artificial intelligence and deep neural network technology, Nvida, which provides the computing infrastructure, finally survived the chance of turning over. Tesla, a large-scale parallel computing chip based on CUDA architecture, was born for the AI era. Jen-hsun Huang's insistence on CUDA began to generate returns.

04 AMD Miracle: The Jedi counterattack; Intel's housekeeping power, Moore's Law, failed; Invida stayed up to the tuyere of deep learning

AMD, who saw the street coming, ushered in the new CEO Lisa Su in 2014.

AMD's last run was a seven-fold increase from 2015 to early 2017, followed by the launch in 2017 of a cheaper Razor processor and EPYC series that directly rivaled Intel on high-performance processors, snatching back some CPU market share from Intel's mouth.

In July, AMD unveiled the Threadripper family of high-end desktop CPUs in an attempt to change its reputation as a low-end product and directly hit Intel's monopoly of the high-end CPU processor market. Of course, the biggest hit was this year's announcement of AMD's 7-nanometer integrated memory, announcing the start of providing customers with Radeon Instinct samples of the 7-nanometer Radeon Vega architecture.

Contrary to AMD's Jedi rally, Intel's "pendulum deterrence" failed. After 2014, Moore's Law, the rationale for the pendulum program, encounters the physical limit ceiling of component transistors. Insiders, such as CEO Jen-hsun Huang, even claimed that "Moore's law is dead". Although Intel's dead duck's mouth refuses to admit that Moore's law is invalid, let alone 7 nanometers, Intel's 10 nanometer chip has been jumping tickets for three years, repeatedly delaying the launch date.

AMD, which has reached 7 nanometers, has completely reversed repression and even has a one-to-two-year manufacturing advantage over Intel.

In addition to catching up with Intel in the server market and making breakthroughs, the potential and competitiveness of AMD in the GPU market have also been highlighted. Despite slow progress in GPU products in 2017, the release of 7Nm GPU means that AMD will move into the forefront of AI computing. The Rdeon Vega GPU chips, which focus on artificial intelligence and in-depth learning, provide diversity for AMD products. The first Radeon Instinct accelerator has provided prototypes to the first batch of customers and will be shipped in the second half of the year.

According to AMD, the accelerator is "the fastest training accelerator in the world," optimized for large-scale parallel applications to meet the needs of large data set machine intelligence and HPC-level system workload. The figures also show that revenue from the computing and image sectors has been a key driver of AMD's sustained high growth and improved gross margins over the past few quarters.

In 2012, with the help of Google's Artificial Intelligence Team, Nvida built the largest artificial neural network of its time, and after that, in-depth learning teams began using Nvida graphics cards in large numbers. In 2013, invidia reached a partnership with IBM in building enterprise data centers. In 2017, NVIDIA released Pegasus for L5 completely unmanned development platform. The price of NVIDIA has more than 9 times since 2014.

Big Three stock price 2012-present (Yahoo Finance) (dark blue line AMD; light blue line Intel; purple line NVIDIA)

05 NVIDIA grew into a GPU overlord, and Intel and AMD even managed to deal with the NVIDIA alliance.

Over the past two-and-a-half years, Nvidia has remained on top of the pack in high-end GPUs, high-performance computing, HPC, and data centers, and has been hailed as one of the three major breakthroughs in advanced learning (massive data and training sets require high-powered GPUs).

In 2014, Nvidia's revenue was $4.13 billion, compared with $9.71 billion in fiscal year 2018, a 2.35-fold increase corresponding to a 3.4% increase in Nvida's GPU (CPU with built-in GPUs), suggesting that Nvida was eating the fattest piece of meat in the market.

At present, the two main businesses of NVIDIA are games and data centers. Although the volume of data center business is only one third of the game business, it is in the tuyere of deep learning and cloud computing, and has great potential for future development. According to Tianfeng Securities, Nvidia now accounts for 96% of the global server GPU market, while AMD happily accounts for the remaining 4%.

The data center business is now Nvidia's second-largest business, maintaining a year-on-year growth of 100 percent in each quarter since the second quarter of 17 years. Nvidia's revenue for the first nine months of this year has been $1.326 billion, up 148 percent from the same period last year.

Intel and AMD will not turn a blind eye to Nvida's bullishness. "The enemy of the enemy is a friend." In November 2017, Intel even announced an alliance with AMD. In order to suppress the "common enemy" Yingweida, Intel and AMD tear up for so many years to be able to work together to integrate AMD semi-custom graphics chip into an Intel CPU, the melon crowd is also surprised.

Soon after the announcement of the alliance, Intel poached Raja Koduri, one of AMD's core technology leaders and graphics director. Just joined the alliance to insert a knife allies, this reversal hit the melon-eating masses caught off guard, can only lament that the world of the three giants is really a surge of sneaky clouds and undercurrents.

06 the past of Tian Ji racing, the future of AI era.

In 2014, there was a question about Intel, AMD, and Nvidia that said, "AMD graphics cards can't sell NVIDIA, CPUs can't sell Intel, why can they survive to this day?"

One humble answer is this: "Because AMD graphics cards do better than Intel, CPUs do better than Invid."

This cunning and perfect summary of the survival of the Big Three at that time, the answer was only 89 praises, really heroic. Now, 4 years later, the pattern of Tian Ji racing is changing.

AMD's market share hit a four-year high of 14.9% in the first quarter of 2018, according to the latest GPU market share data from JPR consulting, which specializes in the GPU market (including CPUs with built-in GPUs). Unfortunately, the number fell to 13% in the second quarter as the digital money market turned cold. For now, AMD seems to be shifting its focus to semi-customized and gaming consoles (mainly for Microsoft, Sony and Nintendo) in 2019, rather than consumer graphics, in spite of all the advances in server, GPU, and cloud computing this year.

"as the only company with GPU and x86 silicon chip technology," AMD, which is in the GPU CPU market for 10,000 years, is likely to play a synergistic role with GPU CPU, moving its product line back to the high end of the market, according to Tianfeng Securities. If AMD does, then Tian Ji horse racing three-nation pattern may be rewritten again.

Intel, which had not caught the bus to the data center and had stagnated in its own technological processes, did not wait to die, and adopted a strategy of acquiring the stones of other hills for its own use. Intel Capital, an Intel unit, has been one of the world's most aggressive AI investors in the past five years, according to Tianfeng Securities. Intel now has layouts in key segments of the AI era: FPGA chip giant Altera, deep learning company Nervana, driverless industry leader Mobileye, and machine vision chip maker Movidius have all been included.

We will wait and see the new season of the Three Kingdoms story.



China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments