AMD says RDNA 3 architecture will have leading energy performance

by ddos ·

The need for immersive, realistic gaming experiences has pushed graphics technologies to push the boundaries of technology, pushing the boundaries to support ray tracing, variable rate shading, and various graphics technologies. These improvements require improved performance through continuous advances in chip design and architecture, which in turn pushes up the power consumption of graphics cards. At present, the power consumption of desktop discrete graphics cards has reached 400W or more, requiring larger cooling solutions, coupled with the recent soaring energy prices, making power consumption and heat dissipation a topic of concern for many players.

AMD published a new blog post in which Sam Naffziger, AMD senior vice president, enterprise researcher, and product technology architect, describes the performance-per-watt improvements that have been achieved over the past three generations of Radeon graphics cards. It is said that when the RDNA architecture and the Radeon RX 5000 series GPU based on the 7nm process were launched in 2019, the performance per watt increased by an average of 50% compared with the GCN architecture. The RDNA 2 architecture and Radeon RX 6000 series GPUs released in 2020 have improved performance per watt by 65% ​​when also based on the 7nm process. The RDNA 3 architecture will use a 5nm process and a small chip design, and the Radeon RX 7000 series will be its first consumer-grade GPU in an MCM multi-chip package, with a performance-per-watt improvement of more than 50% compared to the RDNA 2 architecture.

In June of this year, Sam Naffziger said in an interview that the total power consumption of the next generation of GPUs will increase, and the efficiency improvement is to maximize performance on this basis. Sam Naffziger has been with AMD for 16 years, overseeing multiple product areas, and focusing on driving performance-per-watt improvements to improve AMD’s competitiveness in CPUs and GPUs. He is also one of the main drivers behind AMD’s chiplet architecture, and a similar line of thinking has been successful with the Ryzen and EPYC series of CPUs.

Sam Naffziger said AMD, as the only company developing high-performance CPUs and GPUs today, is uniquely positioned to leverage the engineering team’s learnings and leverage the best IP in its portfolio. At present, AMD integrates some design methods of the CPU micro-architecture into the GPU design, such as simplifying the physical composition of the chip and making higher frequencies possible. Infinity Cache is implemented using the CPU’s L3 cache design idea.
To further refine and improve the AMD RDNA 2 graphics architecture and deliver greater efficiency and performance gains, the team implemented several other key changes, including:

  • Optimized Switching – Improving the fundamental design of the architecture to ensure every gate switched and every clock toggled directly contributed to performance, thus removing any wasted activity and excess routing to optimize the graphics pipeline.
  • High-Frequency Design – Tuning the design for high clock speeds, pushing AMD RDNA™ 2 frequencies beyond AMD RDNA™ by up to 30 percent, which enables the GPU to run at a lower voltage to ensure the new architecture could maintain the same clock speeds at lower power.
  • Smarter Power Management – Implementing intelligent power management within the GPU, which identifies the best opportunities to exploit higher frequencies and does so only when it directly improves performance, then reduces back down, thus eliminating excess energy use.

Sam Naffziger looks forward to the upcoming RDNA 3 architecture that will drive more efficient gaming, reiterating that under the blessing of the 5nm process and chiplet design, the RDNA 3 architecture will improve performance per watt by more than 50% over the RDNA 2 architecture, providing gamers with best-in-class gaming performance and a quiet and energy-efficient design.

The RDNA 3 architecture improves upon the RDNA 2 architecture’s adaptive power management technology to set workload-specific operating points, ensuring that each component of the GPU uses only the power needed for optimal performance. The new architecture also introduces a new generation of Infinity Cache, which is expected to provide a higher-density, lower-power cache, which can reduce the power consumption of video memory and help the RDNA 3 architecture achieve leading energy performance.

Tags: AMDRDNA 2 architecture