Leaked benchmarks from AMD's internal validation labs suggest RDNA 5 delivers a generational leap in ray tracing - long the architecture's weakest point. Samples reportedly achieve 2.3x the ray-tracing throughput of RDNA 3 at iso-power.
A Dedicated Ray Engine
Unlike RDNA 3's shared shader-array traversal, RDNA 5 introduces a dedicated ray-traversal engine per shader engine, with a rebuilt BVH cache cutting memory round-trips by an estimated 40%.
"AMD finally stopped trying to bolt ray tracing onto compute units. The dedicated engine is the reset they needed," said analyst Chips and Cheese.
Where It Lands
The chip is expected to debut in the Radeon RX 9000 series late this year on TSMC's N3P node. If leaks hold, AMD could narrow NVIDIA's ray-tracing lead to within 15% in flagship titles - down from over 60% today.
Power efficiency gains also matter for the laptop market, where AMD has struggled to match NVIDIA's RTX 50-series battery life.
Related Articles
NVIDIA Blackwell Ultra: The $40,000 GPU Powering the AI Revolution
NVIDIA's latest data center GPU delivers 4x the AI performance of its predecessor, but the eye-watering price tag raises questions about who can afford to participate in the AI race.
Apple M5 Pro Benchmarks Leak — And They're Extraordinary
Leaked Geekbench scores suggest Apple's next-generation chip delivers desktop-class performance in a laptop form factor, with the neural engine getting a 3x boost for on-device AI.
The Race to Build 2nm Chips: TSMC vs Samsung vs Intel
The three semiconductor giants are in a fierce competition to manufacture the world's first 2-nanometer processors. The winner could reshape the global chip industry for a decade.
Comments (0)
Log in or sign up to leave a comment.
No comments yet. Be the first to share your thoughts.