Author page description
October 21
- High-Speed Packaging Technology and Reliability of Optical Chips
➀ Optical chips used in optical modules are made of various crystal materials and are very delicate.
➁ The packaging technology for different types of chips varies.
➂ The continuous advancement of AI's computing power drives the industry towards high-speed optical modules and raises market expectations.
➃ High-speed optical modules have characteristics such as high speed, large bandwidth, high power consumption, and increased attention to reliability due to internal temperature rise.
➄ Some processes affect both bandwidth and reliability.
➅ Packaging technology is a crucial part of the entire industry chain.
- Explosive HBM Demand Expected to Push DDR5 Pricing Up by 20%!
➀ The demand for HBM is rapidly growing due to the surge in AI server needs, with supply for 2024 and 2025 already sold out by Micron and SK Hynix.
➁ TrendForce analysts predict a 5% to 10% increase in HBM memory prices next year, with DDR5 prices expected to rise by 15% to 20%.
➂ HBM is much more expensive than standard DRAM, about five times the cost of DDR5, but its performance and capacity advantages justify the higher cost.
- AI Network Background: Why, What & How of RDMA
➀ The need for RDMA in high-performance computing, big data storage, and machine learning domains;
➁ Detailed introduction of RDMA network architecture, key components, and protocols;
➂ Explanation of the RDMA working mechanism, including software configuration and hardware design;
➃ Overview of RDMA technology protocols such as InfiniBand, iWARP, and RoCE;
➄ Future technological innovations and challenges for RDMA, including scalable RDMA controllers, efficient QP management, congestion control, network topology optimization, and security.
- Why Intel Launched OCSP? The Background and Goals of Intel's Open Common Server Platform
➀ The background of Intel OCSP: The server market is highly competitive with significant pressure on small and medium-sized brands, and profits of large manufacturers are thin.
➁ The初衷 and goals of Intel OCSP: To help small manufacturers reduce development pressure and deepen the Intel server ecosystem.
➂ The scope and objectives of Intel OCSP: To define open standards for server components like cases, power supplies, motherboards, etc., to build a healthy server component ecosystem in China.
October 20
- The Midfield Battle of Domestic CPUs
➀ The domestic CPU industry is entering a new stage with the expanding market demand for domestic processors, marked by three major routes: x86 architecture, ARM architecture, and LoongArch architecture;
➁ The core of evaluating domestic processors today should be how to find a balance between the two routes and let some routes develop first, with x86 and ARM showing significant potential while LoongArch carries the initial vision of independent development;
➂ The three routes complement each other, with x86 and ARM leading the market and LoongArch focusing on ecosystem development and performance improvement.
- Should Domestic Chip Start-ups Continue Financing?
➀ Chip start-ups in China have been facing a series of deep reflections over the past six years.
➁ The opportunities in the chip industry are cyclical and have an element of chance.
➂ The Chinese chip market may not be able to accommodate many domestic chip companies, and the industry is heading towards polarization.
➃ The two possible outcomes for domestic chip start-ups are listing or acquisition.
➄ The era of 'survival of the fittest' has passed in the Chinese chip industry, and only the 'winner takes all'.
- Intel and AMD Alliance to Counter Arm, but Cannot Halt Cloud Firms' Self-developed Arm
➀ Intel and AMD have formed the 'x86 Advisory Group' to counter Arm.
➁ Cloud firms like AWS are developing their own Arm processors, lowering costs and increasing flexibility.
➂ Intel and AMD have responded with high-core-count processors, but cloud firms are continuing to develop their own Arm processors.
- Collaborating with HarmonyOS Starlink, Creating an Intelligent IoT | Xinhai Technology Debuts at the 6th Core Chip Eco-System Conference
➀ The conference gathered hundreds of industry elites to discuss technological innovation opportunities and promote the strong development of China's semiconductor industry.
➁ Liu Longguang, General Manager of Xinhai Technology's AIoT product line, delivered a speech on 'Collaborating with HarmonyOS Starlink, Xinhai Technology Creates a New Era of Intelligence', showcasing the company's achievements in the HarmonyOS ecosystem.
➂ Xinhai Technology has developed a full-signal chain technology platform, providing an AIoT overall solution to create a differentiated competitive advantage and actively integrate into the HarmonyOS ecosystem.
- China's Electric Vehicle Market Captures 66% of Global Market, Yet Relies on Imports for 90% of Chips!
➀ China's electric vehicle market captures 66% of the global market with a monthly sales record of 1.1 million vehicles in September 2024;
➁ China's market share in the global electric vehicle market reached 720,000 vehicles from January to September 2024, with a year-on-year growth rate of 35%;
➂ China's electric vehicle industry faces the issue of over-reliance on imported chips, with more than 90% of chips being imported, and the dependency rate of computing and control chips reaching 99%;
➃ The Chinese government plans to increase the proportion of domestically purchased auto chips to 20%~25% by 2025.
- Semiconductor Manufacturing at a New Inflection Point?
➀ The semi-damascene technique, when combined with patternable metals like Ru, promises to achieve RC, area, cost, and power efficiency for interconnect scaling paths.
➁ The introduction of CU DUALdamascene in BEOL in 1997 marked a turning point in semiconductor history, shifting from subtractive Al patterning to wet processes like copper electroplating and CMP.
➂ Imec proposed a new metallization concept called 'semi-damascene' in 2020, which starts with direct patterning of the first local interconnect metal layer and then uses a single inlaid technique for the vias.
- Why Layout Engineers Must Learn FinFET? 'Advanced FinFET Process Simulation Layout Design' Online Live Class Registration!
➀ The advancement in semiconductor manufacturing processes, such as the transition from 28nm to 12nm FinFET, is driven by the need for higher integration, processing power, and lower power consumption to meet the growing demand in AI, high-performance computing, automotive electronics, 5G communication, and the Internet of Things.
➁ FinFET technology, with its innovative 3D structure, addresses the size scaling and performance bottlenecks of planar transistors, becoming a key technology to extend Moore's Law.
➂ The course offered by E课网 aims to teach advanced FinFET layout design skills in just two weeks, providing real-world project experience and preparing students for higher salaries and career development.
- 多位院士领衔，第二届集成芯片和芯粒大会开放注册！
➀ The 2nd Integrated Chip and Chiplet Conference will be held in Beijing from November 8 to 10, 2024, with the theme 'Integrated Chip: Stepping into the Era of Large Chips';
➁ The conference is organized by the Institute of Computing Technology of the Chinese Academy of Sciences and Fudan University, focusing on the cutting-edge dynamics and future development trends of integrated chips and chiplets;
➂ The conference will feature theme speeches, expert roundtable forums, black technology releases, technical forums, and open-source community competitions, discussing key issues in the field of integrated chips and chiplets.
- Salary Exposure for American Chip Engineers
➀ The average salary for U.S. engineers and other IEEE members increased by about 5% from 2022 to 2023, according to the IEEE-USA 2024 Salary and Benefits Survey.
➁ The median salary for circuit and equipment engineers is the highest, at $196,614, followed by communication ($190,000) and computer/software technology ($181,000).
➂ The median salary for professors is $190,000, associate professors is $118,000, and assistant professors is $104,500.
- Nvidia and Google Embrace RISC-V
➀ Nvidia and Google are discussing the use of RISC-V architecture at the RISC-V Summit;
➁ Nvidia has been using RISC-V CPU architecture for nine years and will have a 20-minute keynote speech at the summit;
➂ Google employees, including Cliff Young, the designer of TPU, will discuss the advantages of RISC-V in AI chips;
➃ RISC-V is gaining influence as companies seek cheaper and faster CPU design methods;
➄ Apple is also using RISC-V microcontrollers in its M series CPUs;
➅ RISC-V is trying to capture a market tired of proprietary options like x86 and ARM;
➆ Omdia predicts that by 2030, the shipment of RISC-V processors may reach 17 billion, with about 46% for automotive use;
➇ RISC faces challenges such as security issues and political concerns;
➈ Alibaba has released a new chip for embedded devices with many new security extensions;
➉ Some companies have discussed selling RISC-V chips, but manufacturing issues will not be widely discussed at the summit.
- The Future Development of GPUs
➀ Nvidia's Ian Buck discusses the shift from focusing on single chips to integrated systems;
➁ The transformation of data centers into AI factories due to the rise of AI;
➂ The increasing power and cooling requirements of GPUs and Nvidia's involvement in the Coolerchips project.
- ReRAM: The Preferred Alternative to Flash Memory
➀ ReRAM is becoming the preferred alternative to flash memory due to its cost, complexity, power, and performance advantages.
➁ ReRAM is finding its place in power management ICs, IoT, MCU, and edge AI applications.
➂ ReRAM offers inherent security for secure applications and is poised to become a cornerstone for neuromorphic computing.
- Inside the Failure of Intel's Itanium: Revealed by a Former Engineer
➀ The article reveals the internal details of the failure of Intel's Itanium processor;
➁ It discusses how Intel's insistence on releasing x64 versions of Itanium CPUs could have led to victory in the x86-64 domain over AMD;
➂ The article highlights the importance of backward compatibility in the PC market and how Itanium's pure 64-bit architecture hindered its market performance;
➃ It describes the role of former Intel Chief x86 Architect, Robert Colwell, in the development of the Pentium 4 CPU and the challenges he faced with Intel's management.
- Learning Notes: Solidigm SSD Product Knowledge Points
➀ Solidigm, previously known as Intel, is a leading brand in the NAND solid-state drive industry, acquired by Intel and SK Hynix in 2020.
➁ The product naming rules of Solidigm are explained, using the example of a data center-oriented SSD.
➂ Advanced features of Solidigm's enterprise SSDs are highlighted.
➃ The product roadmap for PCIe NVMe includes U.2 form factor, increasing capacities, and the introduction of PCIe 5.0 products.
➄ The roadmap for SATA SSDs focuses on system drives with lower capacities and the continuation of SATA and M.2 form factors.
- The Military Prospects of Brain-Computer Interface
➀ Brain-computer interfaces offer multiple military applications, such as health monitoring, neural regulation, sensory expansion, telepathy, lie detection, human modification, and experience downloading;
➁ The concept of 'Centaur warfare,' which refers to human-machine collaboration, is gaining traction in the current US military strategy, with brain-computer interfaces being a key technology;
➂ 'Neuro warfare,' a new combat concept that aims to gain a military advantage through the dominance of the brain and nervous system, is gaining momentum in the US military.
October 19
- Negative Development Process Characteristics and Simulation Difficulty Analysis
➀ The characteristics of negative development process are discussed, including high spatial imaging contrast and suitability for making grooves and contact holes.
➁ The effect of KrF wavelength compared to ArF wavelength and the change in imaging performance with line/space ratios are analyzed.
➂ The tolerance of NTD process in making low CD patterns and its application in specific structures are highlighted.
- Using Large Area Analysis to Improve Semiconductor Yield
➀ Design Rule Check (DRC) is a technology used in the chip design process to ensure high yield manufacturing of devices.
➁ Advanced semiconductor technology nodes have increased the number and complexity of DRC rules, making traditional 2D DRC ineffective.
➂ Lam Research's SEMulator3D platform is used for 3D modeling and rule-based metrology of semiconductor devices.
➃ Large Area Analysis (LAA) is a powerful concept in semiconductor engineering research.
➄ LAA can help engineers develop the best semiconductor processes using a limited number of experimental wafer runs.
➅ SEMulator3D can identify 3D device failures such as shorts, opens, and material excess problems.
➆ The output of LAA includes rule-based metrology, fault identification, and defect maps.
➇ LAA can save wafer and mask costs and accelerate yield improvement curves.
- Countdown to Three Years: Where is the国产 Database 100% Replacement?
➀ The development of domestic databases has gone through several key periods, from breaking the monopoly to the "hundred-battle" era;
➁ The replacement rate of domestic databases in the finance industry is not high, with non-core systems around 40%;
➂ The main challenges for domestic databases include high migration costs, open-source database risks, and an imperfect ecosystem.
- A Ten-Year Journey of a RF Engineer — Who's Life Is Not Confused
➀ The author reflects on his first year in the RF industry, feeling lost and aimless, and his struggle to learn and adapt.
➁ After the second year, he decides to change his job and starts working in a射频 company, improving his skills and finding direction.
➂ He transitions to the whole machine industry and aims to become a RF leader, working hard and achieving his goal.
➃ After becoming a RF leader, he faces another period of confusion and seeks to deepen his understanding of product design and communication protocols.
➄ He continues learning and eventually finds his next career goal — becoming a product director or technical director.
➅ The author emphasizes the importance of hard work and learning, leaving the rest to destiny.
- Bull Market Speculation Directions
➀ The manufacturing industry and technology are crucial for China's future development;
➁ In the manufacturing sector, excess profits are earned through high-margin products;
➂ In the technology sector, high profits are used to support wafer fabs and photolithography machines;
➃ How to create a positive feedback loop? For every million-dollar luxury car sold by Huawei, it equals one less luxury car sold in developed countries like Europe and the US, and high-level engineer positions shift from Europe, the US, and Japan to China; Huawei (high-tech) and its high-margin manufacturing sector are the directions for this bull market.
- Intel: The Decline of an Empire
➀ Intel is facing accusations of product vulnerabilities and security breaches, with China's Cybersecurity Association suggesting a cybersecurity review of its products sold in China.
➁ Intel's revenue has been declining, with a significant portion coming from the Chinese market.
➂ The company has been struggling with the competition in the CPU market and the lack of growth in new markets like AI chips.
➃ Intel's manufacturing technology has fallen behind, leading to increased costs and decreased performance compared to competitors like TSMC.
➄ Intel's market value has halved, and the company is facing challenges in new business expansion.
- Kioxia's Struggles: The Rise and Fall of a Semiconductor Giant
➀ Kioxia, a major Japanese NAND Flash manufacturer, has postponed its IPO plan due to investor demands for lower valuation.
➁ Kioxia's parent company, Bain Capital, aimed to use the IPO to recoup investments and inject new capital into the company.
➂ The company's market value has dropped significantly from its peak of $30 billion to around $5 billion in just three years.
- Arm Plans to Bypass ARM Technology and Sell IP Directly to Chinese Customers!
➀ Arm is reportedly attempting to provide IP licenses directly to Chinese customers to bypass its joint venture in China, ARM Technology (China) Co., Ltd. (Arm China), leading to strained relations between the two companies.
➁ Arm established its wholly-owned subsidiary in China in 2002 and planned to sell 51% of its shares to Hillhouse Capital in 2016 to address regulatory requirements and develop products for the Chinese market.
➂ Arm and Hillhouse Capital formed a joint venture, ARM Technology, in 2017 to accelerate Arm's global strategy and support China's IC industry. ARM Technology has developed its own processor IP and other products, which could compete with Arm's IP.
- How to Become the Best VC in Silicon Valley? The Answer: Organize Events
➀ The Silicon Valley tech scene is buzzing with events, with A16z being a prominent name in the Tech Week.
➁ A16z organizes a variety of events, including discussions, workshops, and investor meetings, to connect tech professionals and investors.
➂ A16z's approach to media and content creation has helped it become a key player in the tech venture capital community.
➃ A16z's partners, including Marc Andreessen and Ben Horowitz, have become influential figures in the tech industry.
➄ A16z actively participates in promoting its investment portfolio and supporting emerging technologies.
- What is NPU? Everything You Need to Know
➀ NPU stands for Neural Processing Unit and acts as a hardware accelerator for AI.
➁ NPU complements CPU and GPU, handling tasks like edge AI.
➂ NPU is designed for high throughput and parallel workloads, such as neural networks and machine learning.
➃ NPU is increasingly used in consumer devices like laptops and PCs, as well as in cloud-based systems.
➄ The rise of NPU is driven by the importance of edge intelligence and the need for local data processing.
- Understanding the Strengths of NVIDIA's BlueField-3 DPU
➀ The logic relationship between network cards, smart network cards, and DPUs;
➁ The reasons for the emergence of smart network cards and DPUs;
➂ The advantages of DPU over smart network cards;
➃ Product information of NVIDIA's BlueField-3 DPU;
➄ On-chip system layout of BlueField-3 DPU;
➅ Typical application scenarios of BlueField-3 DPU;
➆ Example of BlueField-3 DPU in VMware private cloud;
➇ NVIDIA's DPU roadmap.