Computex Chronicles Part 3: ARM unveils new architectures and AI libraries
The third keynote in the Computex CEO Address Series was delivered by Arm’s (NASDAQ:ARM) Rene Haas, along with Chris Bergey, senior vice president and general manager of the company’s client business. Their keynote provided some historical context about The role Arm has played in the computing industry over the past several decades, we discussed some new AI software tools, and then expanded on some important architectural announcements for the CPU and GPU IP and related subsystems they debuted last week.
Haas begins with the little-known story of Arm’s origins in providing the first CPU for Apple’s Newton mobile device (AAPL). Although it proved to be one of Apple’s few product failures, the anecdote highlighted the fact that building low-power computing engines that ran on battery power was in Arm’s DNA from the beginning. He went To discuss just how important the focus on low power is today, if not more so, than it was when the company first started – both for battery-powered computing devices and for the high-power server CPUs used in servers.
In fact, as he explained, the three largest cloud computing providers — Amazon (AMZN) AWS, Google (GOOG, GOOGL) GCP and Microsoft (MSFT) Azure — along with Nvidia (NVDA), are using or recently announced custom chips that take advantage of… CPU IP designs for Arm’s Neoverse server. Given the concerns raised about the massive power requirements that will be required by data centers running GenAI workloads, it is a timely and appropriate point.
Haas also highlighted the massive installed base of Arm-based computing devices that have been sold over the past several decades. The number and scope of these devices and their applications has enabled support across an impressive range of operating systems and has helped enable a large library of software development tools. To add to this toolkit, Haas detailed the company’s new KleidiAI software libraries, which are designed to provide a set of middleware-like capabilities that translate between popular AI frameworks such as the Pytorch, Arm-native, and Silicon platforms.
KleidiAI’s idea is to make the process of building AI-powered applications on Arm-based systems as easy as possible and take advantage of some of the unique accelerated instruction sets used in Arm designs. Now, this could be useful for a single product category like cell phones – where arm-based devices are absolutely dominant. But with Arm’s reach beyond smartphones and now into PCs via Qualcomm’s (QCOM) Snapdragon . In addition, the company has a version called KleidiCV for computer vision applications, which is relevant to industrial, medical, and many other vertical industries.
After discussing AI, Chris Bergey took the stage to talk about new client IP products and expanding their Compute Subsystem (CSS) concept into the world of client devices. Arm’s latest ‘big’ CPU is the Cortex-X925 – based on the company’s v9.2 architecture. According to Arm, the X925 features a comprehensive redesign that enables up to a 35% increase in IPC (instructions per clock) performance – the biggest jump the company has ever announced. The company also announced a new “mini” core called the Cortex-A725 that focuses on power efficiency, and a Cortex-A520 core optimized for lower-level applications.
On the GPU side, the company’s latest Immortalis-G925 delivers a 37% increase in graphics performance and up to 34% for GPU-enabled AI applications. As with the CPU family, there is also an upgraded Mali-G725 and Mali-G625 version.
In addition to the performance improvements, what’s interesting about these new IP CPU and GPU designs is that they are specifically designed to increase the range of applications and packages that can be bundled together. For example, it’s possible to create more powerful combinations of large CPU cores and different GPU elements for things like next-generation Arm-based PCs, AR headsets, wearables, and more.
To make the process of designing these chips easier for its partners, Arm has also tailored its CSS offerings to customers. First introduced in the world of Arm-based server designs, the idea behind CSS is to help partners circuit and connect elements to help create the final design of the chip. Think of it as the difference between getting the basic raw materials to build a project and having those materials, glue, screws, and a complete blueprint of how to put everything together. By bringing CSS to customer devices, Arm believes it can help companies spend up to a year of development time. In the case of customer designs, CSS is specifically designed and optimized for the latest 3nm process technology, allowing companies to move to these cutting-edge chip production facilities as quickly as possible.
One element noticeably missing from Arm’s computing story is an IP for an NPU that can be used in traditional computing devices (it provides the low-power Ethos NPU for IoT applications). While I have no doubt we’ll see that eventually, Arm has made it clear that for many Android-based smartphone applications and Windows-based PC applications, CPUs are still the most commonly used computing element (about 70% of Time depends on the company). Intel (INTC) has been making similar arguments for a while, and it appears that a combination of different computing elements — CPU, GPU, and NPU — will be crucial to supporting all AI applications for some time to come.
Finally, when it comes to Arm, it’s important to remember — even though we’re still not well understood — that the work they do and the announcements they make tend to have a broader, longer-lasting impact than almost any of the chip companies’ individual product announcements. Grand. That’s because Arm’s business model is to create chip designs licensed by many of these companies (as well as hardware makers like Apple and Samsung ( OTCPK:SSNLF )), who in turn leverage those designs in their own chips and then build them into products. For a long time, this twice-removed status made Arm a difficult company for many people — and the stock market — to truly appreciate and understand. It seems that the world is finally starting to realize the value of what they are doing and their impact on computing.
Disclaimer: Some of the author’s clients are vendors in the technology industry.
disclosure: no one.
Original source: author
Editor’s note: The summary points for this article were selected by Seeking Alpha editors.