Photonics, Power and RF Packaging Forum 2022 – powered by Yole Group and CITC
- 9:00 AM to 5:00 PM
Photonics and Lighting
IT DOES COMPUTE: THE WORLD IS DIGITAL
RESPECTING THE PHYSICAL LAW: THE EXPONENTIAL COMPUTING POWER OF SEMICONDUCTORS Gordon Moore, co-founder of Fairchild Semiconductor and Intel, published a paper in 1965 in which he predicted that the number of components per integrated circuit, especially for a processor, would continue to double every year. The intervening decades have seen chip complexity increase as Moore predicted, enabling incremental technological leaps. Now it’s half a century later, and a first CPU era has already been supplanted by the GPU era. The market is now seeing itself flooded by a new generation of processing and computing chips. Smarter and ultra-specific, they fuel Artificial Intelligence and neural networks. They are also driving a drastically-changed computing industry that is moving from edge computing toward balanced edge and cloud computing.
With the emergence of semiconductor devices, the processing speed increased spectacularly, and the components became smaller by orders of magnitude.
ASICs, FPGAs, DSP and microcontrollers remain strong fundamentals across the entire industry and are key to understanding the global market. In parallel to the standard computing approach, Artificial Intelligence and Machine Learning have become the centerpiece of most computing chips, like your new smartphone’s Artificial Intelligence accelerator-driven Application Processor Units (APU).
Meanwhile, variations in the architecture such as MPU (Micro Processor Unit), TPU (Tensor Processor Unit), VPU (Vision Processing Unit), and HPU (Holographic Processor Unit) are also being developed alongside new strategies like chiplets.
New approaches to computing have also emerged, like neuromorphic computing mimicking neuro-biological architectures present in the nervous system. There is also in-memory computing, a technique of running computer calculations entirely within computer memory to allow for faster computation.
This multimodal approach enables new ways of addressing computing bottlenecks and triggers smarter and more efficient ways of handling, analyzing and extracting value from data.
From end markets to device functionalities and the related manufacturing supply chains, we at Yole Group have the expertise to forecast market and technology trends, as well as to analyze the key players’ strategic changes and their financial performance. We decipher this very complex market for you, tracking the computing forecast in terms of units, dollars and wafers, providing you a with uniquely comprehensive vision of the industry all year long.
We are also deeply involved in the understanding of the latest technical innovations. We provide real and accurate data, via our reverse-engineering activities of processors and analysis of such devices’ floor plan.
A fast-growing trend within the processor market is the use of chiplets as an alternative to the ever-growing silicon area requirements, especially at the highest tiers of performance. We track the use of chiplets in the main processor segments and forecast their penetration. Yole Group’s analysts trace the market value of processors, not just back to their respective designers but also to the various foundry partners. We analyze the level of wafer production at each process node.
We also develop a market model built from the consolidation of top-down and bottom-up market data, rationalized through a technology model that keeps track and forecasts key technological metrics like die size, process node, and transistor count.
Our aim is to deliver a relevant and accurate understanding of the shared evolution between hardware computing and its associated software.
Through our investigations, we explain and analyze how that can be integrated into silicon. We combine the multiple function integrations.
Beyond the market trends, these technical shifts – front-end, back-end and packaging integration, and software development – need to be deeply analyzed to forecast the evolution of the computing industry.
The landscape has undergone a dramatic transformation in the years since the big leaders of the computing world were IBM, Intel, AMD, TI and Motorola. Only two of them – Intel and AMD – remain in the race, and the competition is only increasing. Due to the specialization of computing units, along with a fabless strategy initiated over 20 years ago, the value of this fragmented market is distributed across the entire chain. This happens from the design stage – using ARM or RISC – to manufacturing, alongside players like TSMC, Intel and Samsung for high-end processors and APU, or through a fabless approach like Qualcomm, Nvidia or AMD and Xilinx.
OEMs like Apple, Xiaomi and Google also tend to develop their own systems-on-a-chip to achieve ultra-high optimization of hardware when combined with software – which requires huge investments. On the industrial side, the datacenter market drives the increasing demand for computational chips to address the work-from-home and cloud gaming trends, and, more globally, the shifting edge-cloud computing balance.
The XPU market is expected to reach more than $180 billion in five years, making it one of the biggest markets for semiconductor devices.
Computing chips are the centerpiece of the global semiconductor wave that has been unfurling over our daily lives for the last 60 years. Yole Group’s computing team provide you the most up-to-date data and analysis about the computing field by closely monitoring the six markets:
And we do this across the most demanding end-systems, such as:
We give you deep insights into the technologies that are inherently driving the future of computing and the supply chain.
We can also help identify the right partners for you, all so that you draw the most benefits from the computing market.
The increase of XPUs’ computing power leads to both evolution – remember Moore’s law? – but also to the development of new integration schemes: chip partitioning, new packaging strategies, new interfaces featuring on-board memories. This evolution explains the major investment by the semiconductor industry’s leaders, like Intel, TSMC, Samsung, NVIDIA, and AMD.