The ASIC market is growing. This has long been a consensus in the industry. However, what is surprising is that the speed of ASIC growth is incredibly fast. Morgan Stanley predicts that the AI ASIC market size will grow from $12 billion in 2024 to $30 billion in 2027, with a compound annual growth rate … Read more
Latest developments in Broadcom’s AI-related ASIC business 3nm Process AI ASIC Chip Mass Production Initiated: Google TPUv6 (the world’s strongest AI XPU) has begun mass production, with projected revenue exceeding$15 billion over its lifecycle. Next Generation TPUv7 Design Completed, set for mass production in 2026; collaboration with Google onTPUv8 (2nm process). Meta’s 3nm MTIA Chip: … Read more
Content Summary1. Computational Demands of Large Models: The need for massive parameter storage, dense matrix operations, and frequent memory access drives chip architecture towards high performance and low latency;2. Challenges in Chip Design: The imbalance between chip computing power and bandwidth, with the “memory wall” becoming the biggest bottleneck to development;3. Innovative Chip Architecture Design: … Read more
UnderstandingCommon AI Chips Introduction In today’s rapidly advancing technology, chips serve as the core “brain” of technological products, and their importance is self-evident. Among them,<span><span>GPU, NPU, TPU, FPGA, and ASIC</span></span> play key roles in various fields. Let’s delve into their technical characteristics, advantages and disadvantages, and application scenarios. 01GPU Technical Characteristics GPUs were originally designed … Read more
Before diving into AI chips, let’s start with a small quiz to see how much you know about AI chips! AI Chip Quiz 1. The core difference between AI chips and traditional CPUs is: A) Higher clock speed B) Stronger parallel computing capability C) Larger storage capacity D) Lower power consumption 2. Which of the … Read more
Source: DeepHub IMBA This article has a total of 4500 words, and it is recommended to read in 5 minutes. This article will delve into five main types of computing accelerators. In today’s rapidly evolving computing technology, traditional general-purpose processors (CPUs) are gradually being supplemented or replaced by dedicated hardware accelerators, especially in specific computing … Read more
01 Introduction The raindrop sensor, also known as the rainwater sensor, is used to detect whether it is raining and the amount of rainfall. It consists of a large rainwater detection board and a control board. When raindrops fall on the detection board, the control board converts this into a digital signal (D) and an … Read more
Since the emergence of ChatGPT at the end of 2022, followed by the hundred-model battle in 2023, and the recent releases of GPT-4.5 by OpenAI, Grok3 by xAI, Claude 3.7 Sonnet by Anthropic, and Llama4 by Meta, the iteration speed of large models has been accelerating. In China, there has been a surge in open-source … Read more
In 2017, Google updated their TPU series. Google referred to this generation of TPU as a “domain-specific supercomputer for training neural networks.” It is evident that compared to TPU v1, which focused on inference scenarios, TPU v2 shifted its design focus towards training-related scenarios. Looking back at history, around 2017, groundbreaking work in deep learning … Read more
👇 Follow our official account and selectStar, receive the latest insights daily Introduction Google has officially launched the seventh generation TPU, Ironwood, designed specifically for inference, with performance soaring 3600 times compared to the first generation, rivaling NVIDIA’s B200. Additionally, Google has introduced several upgraded models, including Veo 2, and even announced the “Google version” … Read more