Talk about those top chip designers

In the semiconductor industry, Gordon Moore must be one of the most important chip engineers in history, and his proposal of Moore's Law, the number of transistors that can be accommodated on an integrated circuit doubles approximately every 18 to 24 months, has always been regarded as the classic law of semiconductor development.

Over the years, the semiconductor industry has also been evolving according to Moore's predictions, but Moore's Law has developed rapidly, requiring not only founders such as Gordon Moore, but also countless excellent chip design engineers in the continuous pursuit of technological progress.

01

Legendary chip designer Jim Keller

In the chip industry, there is one who can be said to be known and known to everyone - Jim Keller. Its legendary experience and brilliant performance have become the object of thirst for talents in major companies. His education is only a bachelor's degree, but for the past 20 years, he has single-handedly stirred up the storm in Silicon Valley.

Jim Keller began his career at DEC. In the 6th and 70s of the last century, DEC was one of the most successful computer manufacturers in the world, at that time DEC used the Alpha21264 microprocessor designed by RISC to be the world's fastest processor, and the fledgling Jim Keller was deeply involved in the development of the Alpha series.

Keller has always been a man of character. When he was a newcomer to chip development and training at DEC, someone came in and asked Jim Keller a question about hierarchical design, Jim Keller thought that half of what this person said made sense, and the other half was very stupid, and then Jim Keller started arguing with this person, and after an hour, no one convinced anyone, until the person left, and the next person told Jim Keller, that person was Gordon Bell, DEC's chief technology officer.

Beginning in 1998, Jim Keller joined AMD to help design the Athlon K7 processor and soon became the main architect of the K8, two generations of processors that beat Intel's 64-bit Itanium and gave AMD the first foothold in the lucrative server chip space.

When everyone thought Jim Keller would be at AMD, he jumped ship without warning to Sibyte, a chip startup, responsible for the development of MIPS chips. A year later, Sibyte was acquired by chip giant Broadcom, and Jim Keller became Broadcom's chief architect. But Keller didn't stay at Broadcom all the time.

In 2004, Jim Keller jumped to a startup, PA-Semi, which was acquired by Apple 4 years later, and Jim Keller naturally became Apple's chip designer. In four years at Apple, Jim Keller led the team to develop the pioneering A4 of Apple's A-series processors, as well as the second-generation A5 (corresponding to iPhone4 and iPhone4S), opening Apple's brilliant core manufacturing road, which also achieved Jobs' self-developed chip strategy.

After helping Apple start its glory in the field of chips, Keller has returned to the arms of his old club AMD. He became AMD's vice president and chief core architect, leading the development of a new microarchitecture, codenamed "Zen", when AMD encountered a major crisis, he took a pragmatic approach, aligned with the design of competitors, abandoned the architecture started with Bulldozer, and became a key figure in AMD's "turnaround again".

After saving AMD twice, Keller quit and joined Tesla. During his time at Tesla, Keller developed self-driving engineering for the company's electric vehicles, oversaw Autopilot and led the low-voltage hardware division. He developed the FSD self-driving chip at Tesla, which is more than 20 times more powerful than the replacement Nvidia solution.

At that time, Tesla's self-developed AI chip had not yet been released, but Musk had already publicly boasted several times: "The special AI chip that Jim Keller is developing will be the best AI hardware." ”

Later, Keller was poached by Intel to serve as senior vice president of silicon engineering at Intel, a team of 10,000 people. Who would have thought that Intel's journey didn't seem to go well. So in June 2020, Keller announced his resignation, handed over as a consultant within 1 year, and started a new life in 2021.

It is understood that Keller's work at Intel aims to simplify the development process of a large number of silicon wafer products and build a strategic platform to pave the way for the development of future products. Before announcing his resignation, Keller proposed innovative approaches such as 3D stacking chips.

Geniuses always have different personalities, and Jim Keller always likes to be constantly challenged, after all, in his own words, he wants to try as many different fields as possible. Jim Keller once said in an interview: "Tesla and Intel are another whirlwind, so you can say I jumped in and out and had a good time." ”

Jim Keller's story continues, and now six months after leaving Intel, Jim Keller is the CTO of Tenstorrent, an AI chip startup in Toronto, Canada, and the next goal is to develop AI chips.

02

Chief chip architect at major companies

Mike Clark, Chief Architect at AMD Zen

As the key to AMD's turnaround again - Zen architecture, it brings a new design and process to AMD, IPC performance increased by 52%, exceeding the original 40% increase. It can be said that the credit of the Zen architecture is unparalleled, so the world is also arguing about who is the father of Zen.

Many people refer to Jim Keller as the "father of Zen", but when asked head-on in the media, Jim said: "I'm a crazy uncle at best."

In fact, Mike Clark, the chief architect of AMD Zen, is strictly the true father of Zen architecture. Mike joined AMD for almost 30 years after graduating from college, and he is now the holder of AMD's highest technical honor Enterprise Academician.

In a media interview about Zen, Mike said he agreed with Jim Keller that there are too many people behind the Zen architecture. But if you have to judge a so-called father of architecture, he may be the most suitable, because since the first day of the Zen architecture project in 2012, he has spent the past 10 years with it.

Mike's story seems to be one of a decade of sharpening a sword. In 1993, Mike graduated from the University of Illinois at Urbana-Champaign, where he received many offers, but he chose to join AMD. Mike believes that AMD is for a company that can really get him into CPU design.

When he first joined AMD, Mike started working on K5, which was their first basic design on x86, and Mike also copied TLB, but he didn't know anything about x86 TLBs and had to understand and reverse engineer how x86 TLBs worked, which went through many twists and turns.

By K6, Mike was responsible for integrating the NexGen. Then to K7, he became the main microcode designer of K7; When he started the Greyhound (K9) core, Mike became the chief architect of K9, and now Mike is the overall Zen core architecture leader and is responsible for the entire Zen roadmap.

In the development of Ryzen, there were also many problems. In 2006, AMD announced the acquisition of GPU company ATI for $5.6 billion, and although this decision made AMD the first manufacturer to have both high-performance CPUs and GPUs at the time, it also put AMD in financial difficulties.

After the acquisition, ATI's GPU market share began to decline. Intel proposed the Tick-Tock strategy in 2006, that is, in the next development, maintain a 2-year process update. When Intel launched its first product, the Core2 series processor, based on this strategy, it matched or even surpassed the Athlon K8 with its new core architecture performance.

Although after the Athlon K8 was equaled, AMD successively launched the Phemon series of processors, from the K10 architecture to the bulldozer architecture, but most of AMD's products can only attract consumers with cost performance and more cores, without exception have been defeated by Intel.

At the time, AMD was in financial trouble, and CPU development funds were frozen and isolated from other businesses. It was a tough time, the chip market needed a product every year, and AMD had to do it. Mike admits that one of the biggest problems is getting the team together, with many people choosing to leave AMD, and Mike needs to spend a lot of time convincing everyone that AMD will succeed.

Zen was named by Mike, and he was there from day one in 2012, feeling the pain and surprise of Zen's birth. Someone came and went, but he was with Zen from start to finish.

Jeff Wilcox, Apple's M1 chief chip designer

In November 2020, Apple released the M1, the first chip built specifically for Macs; Later, Apple successively launched M1 Pro and M1 Max, introducing the system-on-chip (SoC) architecture into the Pro system for the first time, achieving new breakthroughs in memory bandwidth, power efficiency, and capacity.

Jeff Wilcox, head of the Mac system architecture team, contributed to this series. From October 2010 to November 2013, Wilcox served as a principal engineer at Intel, responsible for the power management architecture for the Atom processor chipset program. After that, he joined Apple and began an eight-year career at Apple.

As director of Mac system architecture, Wilcox oversaw Apple's transition to its own chips, including the company's M1 chip. At Apple, he has full responsibility for all the architecture, signal integrity, and power integrity of the Mac system. Starting with the M1 chip, he led the Mac team to achieve the transition to Apple's self-developed chip, and developed the SoC and system architecture of the T2 coprocessor.

However, in January 2022, Wilcox posted on social media: "After an amazing eight years, I decided to leave Apple and seek another opportunity. It was an incredible journey, and I'm incredibly proud of all that I've accomplished there, culminating in the transition from Apple Silicon to the M1, M1 Pro, and M1 Max SoCs and systems. I will miss all my Apple colleagues and friends terribly, but I look forward to the next journey. At present, Wilcox has worked at Intel as an Intel Academician and CTO of the Design Engineering Group, and will focus on client SoC architecture design.

Murthy Renduchintala, principal engineer at Intel

Whether in Intel or AMD, the chief architect must basically be responsible for the management design and optimization of the hardware architecture, and be able to get the title of "chief architect" and must be able to convince the public in the semiconductor circle.

However, not all chief architects can lead chip companies to achieve great success, and planning a chip roadmap is not an easy task. In 2020, Intel announced that its chief engineering officer, Murthy Renduchintala, would be leaving, which seemed to have something to do with the delay in Intel's plan roadmap.

Dr. Renduchintala left Qualcomm Vice President and Qualcomm CDMA Technology (QTC) Co-President to join Intel in 2015, initially responsible for Intel's new business unit, Technology, System Architecture and Customer Group (TSCG).

In 2016, he sent a memo to other senior Intel leaders with an emergency plan to address the "competitiveness gap," and he himself was subsequently promoted to the position of chief engineering officer.

Renduchintala was once considered one of the candidates to take over Brian Krzanich as Intel's CEO. Just after joining the company, his entry bonus is as high as $10 million and his salary is more than $25 million. By 2019, Renduchintala was already one of Intel's highest-paid executives, with a total annual compensation of about $26.88 million.

From design to engineering to manufacturing, he is responsible for almost all of Intel's hardware. Speaking about Renduchintala, Intel also said that Renduchintala's team brings together all of Intel's major technical, engineering, and manufacturing functions, including semiconductor process technology, manufacturing and operations, system and product architecture, IP development, design and system-on-chip engineering, software and security, and managing Intel Labs.

Joining Intel at the time, Renduchintala was tasked with helping Intel open up a larger market beyond the CPU, which was the PC era. Soon, though, Intel sold the business to Apple for $1 billion.

At that time, the basic indicators of Intel's 10nm process technology were even stronger than TSMC's 7nm process, but Intel believed that 10nm was not the best node. To make up for the delay in 10nm chips, Intel is determined to work 7nm to ensure on-time delivery.

However, Murthy Renduchintala made a serious mistake in leading the team to develop 10nm and 7nm processor architectures, resulting in Intel's 7nm process development schedule being delayed by about 6 months from the original plan, and the time to market delayed by about a year.

As a result, Murthy Renduchintala, Intel's chief engineering officer, also announced his departure shortly after announcing the extension.