According to a report by Electronic Enthusiasts (Author: Huang Jingjing), recentlyWu Xiongang, Chairman of CoreLab Technologyshared his views on the development of next-generation computing architectures in the AI era during a speech, stating that the exponential growth of computing power will inevitably give rise to new computing architectures.Looking back over a decade ago, when developers chose computing platforms, 95% focused onthe Intel-dominated x86 architecture,while less than 5% chose ARM.Today, the computing platform for smart terminal phones and PCs has completed a critical transition fromx86 toARM. Now entering the AI era, a new architectural revolution is quietly brewing.As a professional who has been involved in the PC CPU field since the early 1990s and has deep expertise in processor architecture and computing platforms for over 30 years, Wu Xiongangbelieves thatevery exponential increase in computing power will inevitably give rise to a new generation of mainstream computing devices, while also driving the birth of entirely new computing architectures.He cited an example,not widely known, that when Apple was looking for chip partners for its first-generation smartphone, it initially targeted not ARM but Intel. However, at that time, Intel rejected this collaboration opportunity.The reasons were twofold: on one hand, Apple had never ventured into the mobile phone sector, which posed uncertainties; on the other hand, Intel did not recognize the demand for a custom low-power, high-performance chip proposed by Apple. In the absence of alternatives, Apple turned to ARM. At that time, ARM could only provide embedded processors based on the MP3 player iPod, which were then developed into the first-generation smartphone chip AP (Application Processor) for Apple.With the advent of the mobile computing era, the data structures, usage scenarios, and data generation methods have fundamentally changed compared to the PC era, leading to a fundamental shift in computing demands.For example,the amount of data generated by mobile networks is tens to hundreds of times that of the PC internet era. The data generated by a high-end autonomous vehicle in one hour is equivalent to a person’s mobile phone usage data over a year. In the AI computing era, such enormous demands for computing power will inevitably drive the emergence of new computing architectures.Wu Xiongangstated that open source isthe key path to align with this trend. We always hope to enable innovators to turn their dream architectures into reality through open-source frameworks.However, if each company innovates based on its own architecture, it is easy to trigger the “fragmentation” problem, leading to incompatibility among different manufacturers’ RISC-V processors, GPUs, and other products.How to solve the compatibility issue? ARM’s customizable CSS platform provides us with an excellent reference. According to ARM’s annual report, it is predicted that this year the CSS platform will generate one billion dollars in revenue, and ARM is leveraging this “public version” platform to help more companies complete platform migration. For standard Android products, the CSS platform performs excellently, but when creating high-performance innovative products in the AI field,there are more possibilities.CoreLab is committed to building a compatible and unified development systemOpenCSS, connecting various RISC-V processor IPs and platforms, including self-developed ones, while being open to all manufacturers, integratingNPUs, GPUs, and other resources. This open platform not only possesses flexibility but can also further reduce costs and shorten development cycles through industry collaboration, ultimately empowering developers to freely explore in the high-performance AI computing field. Wu Xiongangmentioned that CoreLab’s processor team has developed multiple generations of processors for ARM, with shipments exceeding one billion units, occupying an important position in both domestic and international CPU fields. Now, we hope to share these technological accumulations with industry partners to jointly improve the platform ecosystem.It is worth looking forward to that next week, CoreLab will jointly release a top-performing computing platform in the RISC-V community with TensTorrent in Shanghai,marking a new breakthrough in the field of high-performance computing. Wu Xiongangsaid,we share a common vision with TensTorrent,which isto use an open platform as a foundation, allowing more innovators to utilize RISC-V as a shared ecological platform for software, from CPUs, GPUs, NPUs, to custom chips, creating an open space for innovation in AI computing.
Statement：This article is original from Electronic Enthusiasts and must be cited as above. For group communication, please add WeChat elecfans999,for submission, exposure, or interview requests, please email [email protected].
Read more hot articles
-
In the era of high-mix low-volume, electronic structure verification is turning its attention to 3D printing materials
-
AI is not a bubble; it is an order! NVIDIA’s Q3 performance exceeded expectations, with data center revenue exceeding $51.2 billion
-
Originating from Nobel Prize-winning physicists, breakthroughs in nanorobots!
-
Breaking news! The Netherlands returns control of Nexperia Semiconductor!
-
Head-to-head with GPT-5! This Chinese open-source AI model ranks first globally; what are its killer features?
Click to follow and star us
Originating from engineers, serving engineers, Huaqiu is committed to providing one-stop services for the electronic industry chain, making hardware innovation simpler.