Once upon a time, under Moore’s Law, the performance of chips was soaring, leading the innovation of thousands of industries.
But as the development of chips gets closer to the legendary singularity, the pace of Moore’s Law has quietly slowed down. When the tide is ebbing, some people shout that Moore’s Law is dead, and some pessimistically think that the cold winter has come, but there are still unyielding innovators, with perseverance and persistence, walking on this one-way street that belongs to the warriors.
Turning the field of view back to China, since Huawei was sanctioned in 2019, the original leaders have been silent, but the challengers are not admitting defeat. After the twilight of the gods, the seeds of new forces have germinated.
In August of this year, U.S. President Biden officially signed the “Chip and Science Act” to support the development of the U.S. semiconductor manufacturing industry while restricting the rising Chinese semiconductor industry. Under the action of external forces, the domestic chip industry’s “stuck neck” is getting tighter and tighter, but there are still unyielding people who are willing to persevere on this most difficult road.
Semiconductor manufacturers who dare to climb the peak, their hymn is a hymn of courage.
As one of the top industrial media in the technology industry, Leifeng.com launched the 2022 “Flying to the Earth and Going to the Sea” annual technology list selection plan in October 2022.
The 2022 “Flying into the Earth and Going to the Sea” annual technology list, formerly known as Leifeng.com’s annual “Best AI Nuggets Year-end List”, was founded in 2016, and has gone through six selections, with a total of more than 2,000 registered technology companies . In 2022, the list will be fully upgraded and renamed the 2022 “Flying into the Earth and Going to the Sea” Annual Science and Technology List, hoping to find out those products and companies with the most vitality from a more professional and rich dimension.
For this reason, in the past few months, Leifeng.com has considered artificial intelligence, urban IoT AIoT, industrial intelligent manufacturing, automotive technology, enterprise services, commercial robots, Future interaction, consumer electronics, chips, medical AI and digitalization, cloud and data infrastructure, 12 vertical fields, looking for pioneers and evangelists of the new business order. A total of 49 companies won awards in this selection.
Among them, in the field of “chips”, four companies, Intel, Cambrian, Qualcomm, and Synopsys, stood out and won the Computing Performance Benchmark Award, Cloud AI Chip Landing Benchmark Award, End-to-side AI Chip Landing Benchmark Award, and AI+EDA Landing Benchmark Award prize.
Intel: Computing Performance Benchmark Award
In recent years, the growth rate of chip performance has faced bottlenecks, but as a legend of desktop processors and high-performance servers, Intel has never stopped trying to challenge the limits of computing performance.
Since the new CEO Pat Kissinger took office, Intel has launched a grand IDM2.0 plan.
In terms of traditional projects, Intel launched the Core 13 generation processor, which is popular among users and is known as “the strongest desktop processor on the earth”, making the CPU main frequency enter the 6GHz era for the first time.
In terms of chip manufacturing, the key project of the Intel IDM2.0 plan, Intel not only completed the plan to bring the Intel4 process into mass production, but also made a plan to fully enter the Amy era in 2024, launching the Intel20A and Intel20A based on the new transistor architecture RibbonFET. Intel18A process chip.
Intel is confident that its future process technology and chip technology will continue to make breakthroughs in the future, and has made a prediction that Moore’s Law will continue for ten years.
In addition to seeking new breakthroughs in advanced manufacturing processes and allowing Moore’s Law to continue to challenge the limits, Intel has been constantly moving and exploring new battlefields. It has set up a 1 billion fund to promote the design and development of advanced process chips, and has become a RISC- A senior member of the V Foundation, join the RISC-V architecture track. Then to the acquisition, the acquisition of Gaota Semiconductor was completed on February 15th for 5.4 billion, and the x86 instruction set authorization was opened in IFS foundry services.
And in 2022, the new independent GPU product line Ruixuan series was launched, breaking the long-standing pattern of “red and green competition” in the independent graphics card market, and provided large-scale optimization support for DX9 games for the graphics card in the large-scale update in December , showing Intel’s long-term determination to promote the graphics card business.
On January 11 this year, Intel officially released the fourth-generation Xeon scalable processor (code-named “Sapphire Rapids”), and at the same time launched the Intel Xeon CPU Max series (code-named “Sapphire Rapids HBM”) and the Intel Data Center GPU Max series ( Codenamed “Ponte Vecchio”), in addition to the continuous increase in the number of cores, the new built-in accelerators, involving artificial intelligence, scientific computing, security, network, data analysis, storage and other fields, have an average performance increase of 1.53 times compared with the previous generation. And released the first flagship data center GPU, using 3D package Chiplet technology, integrating 47 small chips on a single product, integrating more than 100 billion transistors. The Max series of GPUs offers up to 128 Xe cores and ray tracing units, and up to 128 GB of high-bandwidth memory.
The Cambrian: Cloud AI Chip Landing Benchmark Award
On March 15, 2016, Cambrian was established. Just one year later, Cambrian announced that it had completed the A-round financing of US$100 million and entered the ranks of AI chip unicorn companies.
More people know the Cambrian because the Huawei Kirin 970, the industry’s first mobile SoC integrated with an AI chip, uses the Cambrian AI processor.
Time flies, and now the unicorn has been hidden in the mountains and forests, but the footsteps of the Cambrian have never stopped. On March 21, 2022, Cambrian officially released the new training accelerator card MLU370-X8. MLU370-X8 is equipped with dual-chip four-core Grain Siyuan 370, integrated Cambrian MLU-Link multi-core interconnection technology, mainly for training tasks, in the industry’s widely used training tasks such as YOLOv3 and Transformer, the parallel performance of the 8-card computing system An average of 155% of a 350W RTX GPU.
The adaptation work of the MLU370-X8 accelerator card and domestic mainstream server partners has been completed, and small-scale shipments have been realized to customers.
Zhang Qiang, deputy general manager of Inspur Information’s artificial intelligence-related product line, said: “Inspur and Cambrian are currently cooperating smoothly on the Siyuan 370 series products, and they are gradually landing in the fields of the Internet, finance, and manufacturing together; the performance of the MLU370-X8 is excellent. We It is expected that the two parties can continue to strengthen cooperation and bring excellent artificial intelligence computing power to more industries and customers.”
Cambrian uses products to confirm its original intention and determination to customers: to provide excellent AI chip products for the explosion of artificial intelligence technology, so that machines can better understand and serve human beings.
Qualcomm: End-to-side AI chip landing benchmark award
In the era of mobile Internet, Qualcomm took off because of smartphones. After AI became a popular track, Qualcomm quickly popularized AI on the terminal side through the huge product shipments of smartphones.
As early as 2007, Qualcomm had discovered the potential of artificial intelligence and made efforts on the AI track. After AI became popular, Qualcomm has made outstanding contributions to the popularization of end-side AI through the resources and reputation accumulated in the field of mobile devices over the years, using smartphones as the entrance. So far, Qualcomm AI product shipments have exceeded 2 billion.
In addition, Qualcomm has also entered the smart cockpit track, which has changed the ecology and gameplay of this industry, causing many car owners to shout out the account name “No Qualcomm, no flagship”.
Qualcomm has been deploying smart cars for more than 20 years. As early as 2002, Qualcomm and General Motors jointly launched the OnStar car networking solution based on its wireless communication technology. Between 2014 and 2021, four generations of cockpit platforms were successively released.
The fourth-generation Snapdragon cockpit platform released by Qualcomm in 2021 has become a benchmark product in this field, leading the industry skyline.
The fourth-generation Snapdragon cockpit platform is the center of high-performance computing, computer vision, AI and multi-sensor processing. Through flexible software configuration, it can meet the computing, performance and functional safety requirements of corresponding partitions or domains, and supports the cockpit to regional The evolution of system electronic and electrical computing architecture. The platform adopts 5nm process technology, enhances the performance of graphics, multimedia, computer vision and AI, and uses Qualcomm’s advanced AI engine to support the top immersive driving experience, including personalized settings for drivers and passengers, natural voice Control, language understanding, driver monitoring, and adaptive human-machine interface, etc., can also support the system to continuously learn and adapt to the preferences of drivers and passengers, bringing optimized situational awareness and adaptive cockpit systems, and can The preferences of consumers are constantly evolving to meet the needs of the next generation of cockpit experience. With enhanced functions, the fourth-generation Snapdragon cockpit platform has become one of the most comprehensive top-level solutions in the automotive industry, bringing excellent in-vehicle user experience as well as safety, comfort and reliability, and setting a new standard for cockpit solutions in the automotive industry. A new benchmark. Jidu Auto’s first model, JIDU ROBO-01, adopts the fourth-generation Snapdragon cockpit platform, and its mass-produced version will be officially launched in 2023.
Synopsys: AI+EDA Landing Benchmark Award
As a leader in EDA products, Synopsys is committed to innovation to change the world. In many fields from chips to software, Synopsys has always led the technology trend, and works closely with global technology companies to jointly develop electronic products and software applications that people rely on. Synopsys is the world’s number one chip automation design solution provider, the world’s number one chip interface IP supplier, and a global leader in information security and software quality.
As the core technology driver of industries such as semiconductors, artificial intelligence, automotive electronics, and software security, Synopsys’ technologies have been profoundly influencing the world’s five emerging technological innovation applications: smart cars, Internet of Things, artificial intelligence, cloud computing, and information security .
Synopsys has launched DSO.ai (Design Space Optimization AI), the industry’s first autonomous artificial intelligence application for chip design. As an artificial intelligence and reasoning engine, DSO.ai can search for optimization goals in the huge solution space of chip design, while automating other decisions, enabling SoC teams to operate at an expert level and significantly amplifying overall production efficiency, thereby completely Change the efficiency of the overall chip design.
With the continuous evolution of the semiconductor manufacturing process, the number of transistors integrated in a single chip is as high as tens of billions, and the design challenges are becoming more and more serious. At the same time, the software and algorithms of terminal applications are accelerating iterations, and the chip design cycle urgently needs to be shortened. The combination of EDA tools and AI technology can not only design chips with better PPA (performance, power consumption, area), but also significantly shorten the chip design cycle. While achieving the vision of providing better, faster, and cheaper chips, it will also greatly reduce the threshold for chip design, allowing more people and companies to design the chips they need. The new era of AI independent chip design is coming!
This article is transferred from: https://www.leiphone.com/category/chips/bxJCyJH7TsW1yzyW.html
This site is only for collection, and the copyright belongs to the original author.