I'm looking to understand the architectural evolution of Radeon GPUs, how real-time ray tracing is implemented, and the performance advantages in gaming.

26 Tem 18:40

In the evolution from RDNA to RDNA3, the most noticeable changes are the organization of the Compute Units (CU) and the introduction of ray tracing hardware. RDNA1 used a relatively relaxed CU layout, offering good frequency performance but limited ray tracing support. RDNA2 introduced Ray Accelerator units within each CU, bringing real-time ray tracing instruction scheduling and acceleration closer to dedicated RT Cores. While its acceleration was slightly lower than NVIDIA's second-gen RT Cores, it had a power efficiency advantage. RDNA3 further integrated ray tracing units and AI acceleration (Tensor Cores) within the CUs, along with higher frequencies and a refined 7nm process, resulting in a 15% to 20% improvement in ray tracing frame rates compared to RDNA2 at the same power consumption. When I compared the RX 6700 XT (RDNA2) and RX 7900 XT (RDNA3) while playing *Cyberpunk 2077* and *The Medium*, the 6700 XT delivered around 10% to 12% lower frame rates than the RTX 3070 with ray tracing enabled at medium settings, but at a power draw of just 140W. Meanwhile, the 7900 XT matched the RTX 3080's frame rates under the same ray tracing settings while consuming around 210W—about 15% less than the RTX 3080 in the same tier. From a driver perspective, AMD's Radeon Software has become increasingly intelligent in resource scheduling, automatically reducing non-critical Compute tasks in ray tracing scenarios to maintain stable frame rates. Overall, Radeon has closed the gap in ray tracing performance while maintaining strong power efficiency, making it a worthwhile consideration for budget-conscious players who prioritize power consumption.