| Cortex-A77, 20% improved IPC over A76
| Mali-G77, High-end High-Definition Graphics Delivery
| Arm Machine Learning Processor, 5 Trillion Operations per Watt

Arm unveiled a large lineup of new mobile IP products at Computex 2019, which opened on the 28th. Over the past year and more, Arm has provided new solutions spanning from network edge to cloud.

The product lineup introduced this time comprises the Cortex-A77, Mali-G77, and Arm machine learning processor. Among these are Project Trillium, Arm Neoverse, two new automotive processors with enhanced safety features, and the Pelion IoT Platform for IoT devices with security capabilities.

Arm's new mobile IP product lineup is designed not only to provide better computational speed and machine learning performance from always-on notebooks that consume power continuously to smartphones requiring advanced security features, but also to provide additional benefits to developers.

The CPU handles not only general computing tasks but also machine learning computing across more devices beyond current limitations, making its importance higher than ever. Additionally, it is important to realize immersive wireless AR/VR applications and HD gaming that can be enjoyed on the move.

The Cortex-A77 CPU achieves 20% improved IPC performance compared to Cortex-A76 devices, and based on this, provides consumers with advanced machine learning technology as well as smooth AR and VR experiences.

The A77 improved overall machine learning performance by 35 times over the A76 through optimization of hardware and software, achieving the performance and efficiency required in the smartphone market.

Arm plans to continuously work toward implementing computing performance in smartphones that is comparable to that of notebooks.

The Mali-G77 GPU delivers approximately 40% improved performance compared to devices equipped with the existing Mali-G76 through a new Valhall architecture.

The G77 improved energy efficiency and ISO-based performance density by over 30% compared to the G76 through core microarchitecture improvements including the engine, texture pipe, and load store cache.

Machine learning performance was also increased by 60%, strengthening inference and neural network (NN) performance for on-device AI applications.

These improvements provide developers with the optimal environment to implement more realistic and immersive games required in today's mobile applications.

Arm's Project Trillium is a machine learning computing platform comprising an Arm machine learning processor and an open-source-based Arm NN software framework, and is currently deployed in over 250 million Android devices.

As demand for various application areas utilizing machine learning technology increases, developers are exploring ways to use hardware platform environments leveraging dedicated NPUs (Neural Processor Units).

Since the announcement of Project Trillium last year, Arm has continuously enhanced its machine learning processor. As a result, energy efficiency has been increased more than twofold compared to the existing standard, enabling processing of up to 5 trillion operations per watt, and memory compression ratio has also been improved by up to 3 times compared to the existing standard.

Based on this, Arm plans to announce a multiprocessor capable of processing up to 32 trillion operations per second by connecting up to 8 cores in the future.