Businesses seeking to innovate with AI or real-time data must harness vast processing power to achieve their goals.
But that means data centres need more power, thus renewing focus on the sustainability of the IT industry.
The last decade has seen significant advances in data centre design, and the efficiency of servers and their processors, or CPUs.
Data centre designers can pack in more processing power. And businesses can use this to centralise applications and data storage, bringing efficiencies, improved security, and resilience.
But AI brings its own challenges.
Research by IDC, found that AI uses up to ten times the power of general-purpose computing. On current estimates, AI data centre energy use was 23 Terawatt hours in 2022, with power use almost doubling every two years. IDC expects it to reach 146.2 Terawatt hours by 2027.1 Restrictions on power could even hinder the development of new data centres in the future, unless the industry acts.
AI – always on
When it comes to energy use, AI is not just increasing demand. It is also changing the nature of data centre energy consumption, as AI workloads have different characteristics to other business applications.
Enterprise software often has peaks and troughs of usage. But AI is often described as “always on”, especially for inference models.
It might be possible to run AI training in batches – perhaps to take advantage of cheaper energy – but AI users will need their queries answered around the clock.
The onus is on data centre operators to be as efficient as possible. Selecting the latest-generation processors helps organisations to reduce their energy consumption overall. And it can also promote efficiency, by matching the right hardware to the workloads.
As industry analysts IDC note, “to maximise the business value of AI, IT leaders need to pivot toward a fit-for-purpose CPU selection and deployment mindset to optimise data centre and cloud computing capabilities that match legacy and modern workloads to the best CPUs for the job at hand”.2
AMD, for example, offers 5th generation EPYC processors to accelerate data centre, cloud, and AI workloads.3
Systems based on AMD EPYC processors can take up less space, deliver more performance per watt, and be more energy efficient.4 AMD EPYC 5th Generation processors have up to 192 cores on a single processor – the efficiency-optimised ‘Zen 5c’ core powers the CPUs with the highest core count of any x86-architecture processors, delivering the highest core density for virtualised and cloud workloads.5
All this reduces power use, the need for cooling, and potentially, the demand for real estate.
AMD’s goal is to achieve a 30x increase in energy efficiency for high performance computing by 2025, known as “30×25”.
This is based on EPYC CPUs and Instinct Accelerators. This is one way the industry is staying ahead of the demand created by AI, both in the data centre and the cloud.6
Investing in the latest data centre technology now allows enterprises to capture the benefits of big data and AI, but also stay ahead of the curve when it comes to energy consumption, and the move towards Net Zero.
As Justin Murrill, AMD’s director of Corporate Responsibility, says: “We are in the midst of a major inflection point in the computing industry, and it is more important than ever to put our values into action and align our efforts with some of the biggest challenges facing our world, including addressing climate change and energy demand.”7
- IDC blog post, Sustainable AI to AI for Sustainability, July 19, 2024, https://blogs.idc.com/2024/07/19/sustainable-ai-to-ai-for-sustainability/ ↩︎
- IDC Spotlight Paper, Sponsored by AMD, CPU Choices Impact AI Ready Infrastructure Cost, Performance, and Sustainability.pdf, #US52029024, April 2024 https://www.amd.com/en/solutions/data-center/insights/cpu-choices-impact-ai-cost-performance-sustainability-idc-report.html ↩︎
- 5th Gen AMD EPYC™ Processors https://www.amd.com/en/products/processors/server/epyc/9005-series.html ↩︎
- See 4th Gen AMD EPYC™ Processors – Architectures Optimized for Enterprise Workloads: https://www.amd.com/en/solutions/data-center/insights/4th-gen-amd-epyc-architectures-optimized-for-enterprise-workloads.html ↩︎
- See 5th Gen AMD EPYC™ Processors – Architectures Optimized for Enterprise Workloads: https://www.amd.com/content/dam/amd/en/documents/epyc-business-docs/white-papers/5th-gen-amd-epyc-processor-architecture-white-paper.pdf ↩︎
- AMD Blog, How AMD is Advancing the 30×25 Energy Efficiency Goal in High-Performance Computing and AI
https://community.amd.com/t5/corporate/how-amd-is-advancing-the-30×25-energy-efficiency-goal-in-high/ba-p/650223 ↩︎ - AMD press release, AMD 2023-24 Corporate Responsibility Report: Advancing Sustainability, Collaboration and Inclusion
https://ir.amd.com/news-events/press-releases/detail/1213/amd-2023-24-corporate-responsibility-report-advancing ↩︎