Total Compute: A Holistic Approach to SoC Design
To meet future market demands, digital experiences must become more interactive, intuitive, and immersive. But without significant changes to current SoC design methodologies, that is nearly impossible. The Arm Total Compute strategy ensures that compute platforms can seamlessly and securely handle tomorrow’s workloads as new devices, use cases, and advanced apps proliferate.
All the performance, efficiency, and AI capabilities required for the next generation of devices.
Defense-in-depth for built-in, multilayered protection across the entire ecosystem.
Improved accessibility and increased performance across software and tools.
Second-Generation Arm Total Compute Solutions
The second-generation Arm Total Compute Solutions (TCS22) feature the latest Arm v9 CPUs, the new flagship GPU Immortalis-G715, new Mali GPUs, Arm System IP and interconnect technologies, Cortex-M and Cortex-R CPU processors, physical IP, software, tools, and standards.
The Three Pillars of Total Compute
Maximize Compute Performance
As gains in chip performance become more challenging, a significant leap forward is needed to accommodate more advanced, complex, and demanding multicompute workloads.
The complexity of new devices, apps, and experiences requires increasingly greater compute power in a similar, if not reduced power envelope. Total Compute evaluates the entire system to identify and develop opportunities to increase performance with minimal impact to the energy and area required.
- Evaluate performance across the whole system rather than on individual components.
- Analyze how interconnecting data and compute are best deployed between IP blocks and compute domains.
- Optimize workloads on devices.
- Develop each IP block with a common architectural approach for SoC performance, efficiency, and data exchange.
- Ensure all components work together seamlessly for first-rate performance and efficiency.
Bolster Security Through a Multilayered Solution
As attacks become more sophisticated, security must be significantly more robust to protect users and organizations who use these new technologies.
To address future vulnerabilities, security must evolve from point products to multilayered solutions that cooperate and communicate across the system. It must also be a fundamental element, built into the core of every device and extended to hardware, firmware, software, operating systems, apps, and cloud services.
- Standardized Platform Security: Establish device platform trust for apps and cloud services through open collaboration with hardware and software developers on standards for common security frameworks and consistent implementations.
- Compute Security: Strengthen software resilience to attacks via Memory Tagging Extension (MTE), which makes detecting memory safety violations easier and more efficient.
- Isolation Architecture: Sandbox trusted applications, segregate secure code and user data.
- Interconnectivity Security: Secure multiple concurrent processes running on CPUs, GPUs and NPUs without restricting their ability to interconnect.
Improve Developer Access
Today, developers learn different tools for each technology and for the various SoC components. As pressure to design paradigm-shifting experiences increases, they must be supported with streamlined development processes.
Total Compute provides frameworks for programming, analyzing, and debugging across all IP – including CPUs, GPUs, and NPUs. This improves programmability across all compute domains for a seamless developer experience, reduced costs, and accelerated time-to-market.
- Create easy, efficient app development processes that remove the need for tools optimized for each architecture or various, multiple discrete IP blocks.
- Give developers access to high-compute performance and allow them to conduct full-system performance analyses for all workloads.
- Help developers build more complex, performant apps through technologies, software, and tools that enable pre-shipment testing, in-field capabilities, and faster debug.
- Offer software frameworks and compute libraries, such as Arm NN, to improve performance across compute domains.
- Ease deployment across platforms through more effective security features.