The Asia-Pacific region is witnessing a rapid evolution in its digital landscape, driven by significant advancements in GPU technology. This surge in technological capability is transforming how industries approach data processing and analysis, crucial for the sophisticated demands of modern AI applications.
As we explore the dynamic role of GPUs in this transformative era, we delve into how these technologies are not only supporting but propelling the integration of intelligent systems across various sectors.
Adapting AI strategies in APAC
Reflecting on this trend, the Asia Pacific market is becoming increasingly recognised as a vibrant hub of AI innovation, expected to grow to a market value of US$78 billion by 2027. While the technology is still emerging, the demands and applications in APAC are largely similar to its global counterparts, with a focus on innovation and pushing the boundaries of what’s possible.
While the specific needs of individual APAC countries may differ based on their respective AI readiness and economic developments, initial trends around its use cases have surfaced. The technology is currently extensively deployed in key industries such as healthcare, science, research, industrial, and security fields.
As global economies, including those in APAC, shift towards an intelligence era, massive computational resources and specialised hardware become key priorities in the roadmap towards AI readiness. This is alongside other key factors such as a robust talent ecosystem and an ethical data governance framework.
At AMD, we believe AI is the single most transformational technology of the last 50 years. A comprehensive portfolio of energy-efficient CPUs, GPUs, adaptive SoCs, and other accelerators is essential to serve the inference and training needs of demanding AI workloads from cloud to edge.
In the APAC region, adapting AI strategies requires robust support for developers and seamless integration of diverse programming models. Collaborative efforts across all AI markets, from endpoints to edge to cloud, are crucial for delivering effective AI inference and training solutions through an open ecosystem.
Collaborative AI projects
In line with the need for robust AI strategies in the APAC region, collaborations between companies are driving advancements. One example is the partnership between KT (formerly Korea Telecom) and Seoul-based AI start-up Moreh to develop an AI platform architecture. This platform, which utilises high-performance hardware and Moreh’s software technology, drives KT’s AI cloud service called Hyperscale AI Computing. Additionally, it supports the creation of a large language model (LLM) to meet commercial application needs in the Korean market.
The chosen hardware was selected for its compatibility with industry-standard programming frameworks and libraries, enabling developers to write hardware-agnostic code. This compatibility allows developers to leverage their existing codebase and expertise, facilitating a smooth transition and deployment on the new platform.
Through this collaboration, KT Cloud achieved notable performance improvements, with processing speeds averaging 1.4x faster than competitor-based servers and a 70% reduction in the effective price of its GPU cloud service. The language model also saw significant gains, achieving 1.9x higher throughput and 117% faster results while requiring only 25% of the network switches and cables compared to competing systems.
APAC trends in GPU usage for AI
Right now, generative AI stands out as the main use case for GPUs, transforming various aspects of work, life, and entertainment. Training generative AI models, including LLMs and generative adversarial networks, continues to grow in size and complexity, involving up to billions of parameters and demanding substantial computing power.
The surge in generative AI and the rapid pace of AI application innovation is evident across multiple industries. These tools are becoming invaluable for creating realistic virtual environments in gaming, aiding scientists in treating and curing diseases, and helping humanity better prepare for climate change. Consequently, the need for compute performance is growing exponentially for both training and inference tasks.
In response to these demands, high-performance GPUs play a crucial role in delivering the necessary compute capabilities. For instance, GPUs enable leadership performance for data centres, from single-server solutions to the world’s largest exascale-class supercomputers. They are well-suited to power demanding AI and high-performance computing (HPC) workloads, offering substantial compute performance, large memory density, high bandwidth memory, and support for specialised data formats.
Several companies in the Asia-Pacific region and beyond have adopted these high-performance GPUs for their AI workloads. For example, Microsoft Azure, Lamini, and MosaicML have implemented these GPUs, while numerous original equipment manufacturers and other cloud service providers, including Dell Technologies, Supermicro, Lenovo, and Oracle Cloud, are also planning to introduce GPU-based solutions this year.
Future trajectory
The importance of GPUs will continue to grow with AI’s increasing pervasiveness, especially as the technology becomes a core driver for both business and economic growth. This trend is expected to persist, as GPUs remain the dominant computing platform for accelerating machine learning workloads, supporting many of the largest AI models in recent years.
Cloud GPUs are also gaining traction as an attractive alternative to physical on-premises hardware. These cloud-based solutions make deep learning projects more accessible to a wider range of organisations and industries, often matching the performance levels of self-built GPU machines at a fraction of the cost of purchasing and maintaining on-premises GPU infrastructure.
As the APAC region continues to harness the power of GPUs for AI advancement, these technologies are set to fundamentally shape future economic and technological landscapes. The ongoing enhancements in GPU capabilities are crucial for meeting the increasing demands of AI applications, ensuring that the region remains at the forefront of innovation.
