AMD announced the 30×25 goal in 2021, its vision to deliver a 30x energy efficiency improvement for AMD EPYC CPUs and AMD Instinct accelerators by 2025 from a 2020 baseline. Now, AMD SVP, Corporate Fellow, and Product Technology Architect Sam Naffziger shared the latest update on this goal
AMD 30×25 Update
The company reveals that through the combination of architectural advanced and software optimizations, it achieved a ~28.3x energy efficiency improvement in 2024 using AMD Instinct MI300X accelerators paired with AMD EPYC 9575F host CPUS. This is when compared to the 2020 goal baseline.
Energy Efficient Design Starts at the Architecture Level
AMD takes a holistic approach to energy-efficient design, balancing advancements across the many complex architectural levers that make up chip design, incorporating tight integration of compute and memory with chiplet architectures, advanced packaging, software partitions, and new interconnects.
AMD Instinct MI300X accelerators pack 153 billion transistors and leverage 3.5D CoWoS packaging to minimize communication energy and data movement overhead.
It is built with eight 5nm compute die layered on top of four 6nm IO die tightly connected to 192GB of high-bandwidth memory (HBM3) capacity. This runs at 5.2 terabytes per second so these accelerators can ingest and process massive amounts of data.
Microsoft and Meta leverage MI300X accelerators to power key services including all live traffic on Meta’s Llama 405B models.
AMD increases the memory on chips and improves the locality of memory access via software partitions and optimizes how data is processed by enabling high bandwidth between chiplets to lower interconnect energy and total communication energy consumption. It does so to reduce the overall energy demand of a system that can multiply across clusters and data centers.
In addition to accelerators though, pairing them with the right CPU host impacts AI performance and energy efficiency. The AMD EPYC 9575F CPUs are tailor-made for GPU-powered AI solutions with the company’s testing claiming up to 8% faster processing than a competitive CPU thanks to higher boost clock frequency.
Continuous Improvement with Software Optimizations
The AMD ROCm open software stack also delivers major leaps in AI performance that allows the company to continue driving performance and energy efficiency optimizations for the accelerators.
The MI300X accelerators have doubled inferencing and training performance across a wide range of AI models through ROCm enhancements since the accelerators launched.
AMD has shared that it is continuously finetuning with partners like PyTorch and Hugging Face to help ensure applications on ROCm libraries are optimized.
ROCm has also expanded support of lower abstraction AI-specific math formats like FP8 to enable greater power efficiency for AI inference and training.
The latest ROCm 6.3 release continues to extend performance, efficiency, and scalability.
What’s Next for AMD?
AMD EPYC CPUs and Instinct accelerators are powering AI at scale to uncover insights through the world’s fastest supercomputers and enabling data centers to do more in a smaller footprint.
In the post, Sam Naffziger shared the company’s commitment to continue pushing the boundaries of performance and energy efficiency for AI and high-performance computing. Moreover, its open software approach enables it to harness collective innovation across the open ecosystem to drive performance and efficiency enhancements consistently and frequently.
In addition, he shared that the company is confident in its roadmap to exceed the 30×25 goal and is excited about the possibilities ahead.
“As AI continues to proliferate and demand for compute accelerates, energy efficiency becomes increasingly important beyond the silicon, as we broaden our focus to address energy consumption at the system, rack, and cluster level. We look forward to sharing more on our progress and what’s after 30×25 when we wrap up the goal next year.”
Ram found his love and appreciation for writing in 2015 having started in the gaming and esports sphere for GG Network. He would then transition to focus more on the world of tech which has also began his journey into learning more about this world. That said though, he still has the mentality of "as long as it works" for his personal gadgets.
