Arm Ltd. is rolling out a series of new chip designs, including its highest-performance mobile processor to date, that will enable handset makers to design faster and more efficient smartphones.

The designs made their debut today at the annual Computex conference in Taiwan. The event also saw Nvidia Corp. introduce a new addition to its data center chip lineup.

Record-setting mobile CPU

The first addition to Arm’s portfolio of chip designs is the Cortex-X4. It’s a processor core, the part of a central processing unit that performs computations. A typical mobile CPU includes multiple cores, as well as supporting components such as a cache that stores the data being processed. 

According to Arm, the Cortex-X4 is the fastest chip of its kind that its engineers have designed to date. It promises to provide 15% higher performance than the company’s previous-generation mobile CPU core. Moreover, the Cortex-X4 consumes 40% less power, which means handset makers can build smartphones with a longer battery life.

The Cortex-X4’s performance and efficiency improvements are the result of several architectural optimizations. Arm has reportedly doubled the size of the onboard L2 cache, which a CPU core uses to store the data it processes. The Cortex-X4 also features an enhanced branch predictor, a component that predicts what calculations will have to be made in near the future and performs them ahead of time.

Another component that Arm’s engineers have improved is the CPU front end. Before a processor carries out an application’s computing instructions, it must translate those instructions into a form that its cores can understand. That task is carried out by the CPU front end.

Arm has equipped the Cortex-X4 features with a new front-end design. As a result, it can fetch computing instructions more efficiently and performs some memory-related tasks with less latency. 

Faster hybrid CPUs

Many mobile processors are based on a hybrid design that incorporates not one but two sets of cores. One set of cores is optimized for performance, while the other offers lower speeds and higher power efficiency. Demanding apps run on the CPU’s high-performance section.

Arm today debuted two new core designs, the Cortex-A720 and Cortex-A520, that chipmakers can use to build hybrid mobile CPUs. Both core designs offer a 20% increase in power-efficiency compared with the company’s previous-generation silicon. Additionally, they feature performance optimizations. 

Arm’s new Cortex-A720 design is meant to power the high-performance cores in hybrid CPUs. It’s 15% faster than the company’s previous-generation product. The speedup is the result of a lower-latency cache and a feature that allows the Cortex-A720 to expend less processing power on branch mispredictions, a type of common CPU error.

The Cortex-A520, the chip design that debuted in conjunction, will power hybrid CPUs’ power-efficient cores. The design uses less electricity than its predecessor because Arm has removed multiple components. The company also improved the power efficiency of the core’s pointer authentication algorithm, which is used to detect and block certain types of cyberattacks.

Arm is rolling out the core designs alongside a module called the DSU-120. It includes the various supporting components necessary to build a hybrid CPU that mixes performance-optimized and efficiency-optimized cores. According to Arm, the DSU-120 makes it possible to build CPUs with up to 14 cores. 

“The DSU-120 provides advanced features like new intelligent power-saving modes to improve efficiency across the CPU cluster,” Arm Director of Product Management Saurabh Pradhan detailed in a blog post. “These new efficiency levels increase days of use for whatever consumer device a particular cluster is targeting.”

New GPUs

At Computex today, Arm also refreshed its lineup of mobile graphics processing units. Chipmakers often combine Arm’s CPU cores and GPU designs in a single processor. Such processors, which are referred to as systems-on-chip, frequently include other components as well.

The Immortalis-G720 is Arm’s new flagship mobile GPU design. The company says that the chip provides 15% higher peak performance than its previous-generation entry into the product category. Furthemore, the Immortalis-G720 uses 15% less power.

According to Arm, the GPU is particularly adept at running mobile games. The reason is that it includes a new feature called DVS, or deferred vertex shading.

DVS is designed to reduce the amount of memory bandwidth required for geometry-related computations. Memory bandwidth is a metric that measures how much data a chip moves to and from RAM every second. Geometry-related computations, meanwhile, are calculations involved in video game rendering.

In complex games, geometry-related computations require a large amount of memory bandwidth. As a result, there’s often insufficient bandwidth left for the other calculations the GPU must perform, which slows performance. According to Arm, the Immortalis-G720’s new DVS feature makes geometry-related computations more efficient and thereby reduces their performance impact on other calculations.

“This architectural innovation also brings smoother gameplay and more lifelike, realistic gaming experiences commonly associated with PC and Console to mobile,” detailed Arm Director of Product Management Dan Wilson. “And it is not just gaming scenes and applications, DVS delivers 37% less bandwidth for a leading CAD application for architects.”

Chipmakers can use the Immortalis-G720 to build mobile GPUs with up to 10 cores. For devices with lower performance requirements, Arm is rolling out two other GPU designs called the Mali-G620 and Mali-G720. They support up chip configurations with up to five and nine cores, respectively.

Image: Arm