Arm unveils first new architecture in a decade – Armv9


Arm’s drive to adopt the x86 CPUs in HPC, AI, and specialty computing workloads was reflected today with the introduction of Armv9, the first new Arm architecture in ten years. The launch will follow the introduction of the 7 nm EPYC Milanese CPUs from AMD on March 16 and before the new 3rd generation Intel Xeon CPUs, which will be launched on April 6.

Armv9 includes Scalable Vector Extension (SVE), a technology Arm developed in collaboration with Fujitsu and built into Fugaku, the world’s fastest supercomputer. Building on this work, Arm developed SVE2 for Armv9 to enable machine learning (ML) and digital signal processing (DSP) for a wider range of applications.

Vectors, one-dimensional data fields, came to HPC in 1976 with the Cray-1 supercomputer. The more vectors a computer can process in parallel and the longer these vectors are, the more powerful the computer becomes. Arm said that while Armv8-A’s vectors are 128 bits long, Armv9’s SVE2 upgrade allows chip designers to choose a vector length in multiples of 128, up to 2048 bits. “That’s an enormous amount of parallel calculations. While SVE was originally developed for the HPC space, in Armv9 SVE2 extends SVE support for a range of specialized augmented and virtual reality (DSP) workloads, from genomics to computer vision. “Arm said.

“SVE2 improves the processing capabilities of 5G systems, virtual and augmented reality, and ML workloads that run locally on CPUs such as image processing and smart home applications,” said Arm in his announcement. “Over the next few years, Arm will continue to expand the AI ​​capabilities of its technology and significantly improve matrix multiplication within the CPU, in addition to ongoing AI innovations in its Mali GPUs and Ethos NPUs.”

Arm Ltd., a United Kingdom-based Softbank semiconductor and software design company, is the subject of a $ 40 billion acquisition effort announced by Nvidia last September. It’s a potentially powerful combination that is currently under development through an expanded regulatory review in the US, UK and China. If approved, the acquisition would give Nvidia a CPU that can be tightly integrated with its industry-leading GPU chips for HPC and AI workloads, just like AMD has its EPYC CPUs and MI100 GPUs, and Intel may have from next year marrying his “Ponte Vecchio”, GPU becomes his advanced Xeon chips.

In its announcement, Arm claimed to have increased CPU performance many times over annually for the past five years and to increase Armv9 performance by more than 30 percent in the next two generations of mobile and infrastructure CPUs. This attributed the company to its Total Compute design methodology, which Arm said will be used in its IP portfolio of automotive, customer, infrastructure and IoT solutions. Arm said there are also plans to increase frequency, bandwidth, cache size and reduce memory latency.

“Meeting the demand for more complex AI-based workloads requires more secure and specialized processing that will be key to opening up new markets and opportunities,” said Richard Grisenthwaite, SVP, Chief Architect and Fellow of Arm. “Armv9 enables developers to build and program the trusted computing platforms of tomorrow by closing critical gaps between hardware and software. At the same time, standardization enables our partners to reconcile faster time to market and cost control and create their own unique product solutions. ”

According to Arm, more than 100 billion Arm-based devices have been shipped in the past five years. According to Arm, 100 percent of the data shared around the world will soon be processed on Arm.

“This penetration puts Arm’s responsibility in providing greater security and performance along with other new features in Armv9,” the company said. “The new features in Armv9 will accelerate the transition from general-purpose to more specialized computing for any application as AI, Internet of Things (IoT) and 5G gain momentum around the world.”

In terms of security, Arms says, the Armv9 roadmap introduces the Arm Confidential Compute Architecture (CCA), which protects pieces of code and data from being accessed or modified while in use, even from privileged software, by doing calculations in a hardware-based secure System to be carried out environment.

“Arm CCA will introduce the concept of dynamically created realms that can be used by all applications in a region that is separate from both the secure and non-secure world,” said Arm. “For example, in business applications, Realms can protect commercially sensitive data and code from the rest of the system while it is in use, at rest, and on the move.”

“The increasing complexity of use cases from edge to cloud cannot be addressed with a single solution,” said Henry Sanders, corporate vice president and chief technology officer of Azure Edge and Platforms at Microsoft. “As a result, heterogeneous computing is becoming more and more ubiquitous and requires greater synergies between hardware and software developers. A good example of this synergy between hardware and software are the confidential computing functions of ArmV9, which were developed in close cooperation with Microsoft. Arm is in a unique position to accelerate heterogeneous computing at the heart of an ecosystem and drive open innovation for an architecture that powers billions of devices. “

Source link


Comments are closed.