Two Cards, One Battleground
AMD’s Radeon RX 9070 and NVIDIA’s RTX 5070 are priced close enough to make buyers genuinely hesitate. Both cards target the upper-mid-range tier, both promise solid 4K performance, and both launched into a market where ray tracing has stopped being a novelty and started being a baseline expectation for PC gaming in 2025. The question is not whether these cards can handle ray tracing at 4K – they can. The question is how much of a gap separates them when the workload gets heavy.
The RX 9070 carries AMD’s RDNA 4 architecture, which brings meaningful improvements to ray tracing hardware compared to RDNA 3. The RTX 5070 runs on Blackwell, NVIDIA’s latest generation, with the full weight of the company’s RT core evolution behind it. On paper, NVIDIA holds the ray tracing advantage it has maintained for several generations. In practice, the gap at 4K is more complicated than a spec sheet comparison suggests.
Architecture and Ray Tracing Hardware
RDNA 4 represents AMD’s most serious effort at competitive ray tracing acceleration. The RX 9070 includes second-generation ray accelerators per compute unit, and AMD restructured how those accelerators handle BVH traversal – the mathematical process of figuring out which rays intersect which geometry. The practical result is that the RX 9070 handles ray tracing workloads roughly twice as efficiently per compute unit compared to the RX 7000 series. That is a real architectural leap, not a marginal revision.
The RTX 5070’s Blackwell architecture builds on Ada Lovelace’s already strong RT core foundation. NVIDIA’s dedicated RT cores operate independently from shader execution, which means the GPU can run shading work and ray intersection calculations in parallel rather than sequentially. This design has been NVIDIA’s structural advantage in ray tracing for years. Blackwell refines the throughput and scheduling of those cores rather than redesigning them from scratch. The result is consistent, high-efficiency ray tracing that does not choke shader pipelines the way AMD’s earlier designs occasionally did.
What matters at 4K specifically is how each card manages memory bandwidth under heavy ray tracing loads. Ray tracing is bandwidth-hungry because BVH structures require frequent texture and geometry lookups. The RTX 5070 ships with GDDR7 and a 192-bit bus. The RX 9070 also uses GDDR6 with a 256-bit bus, which gives it a bandwidth edge in raw numbers. Whether that bandwidth advantage translates into better frame rates under ray tracing pressure depends heavily on how efficiently each architecture uses what it has.
Upscaling and Frame Generation
Ray tracing at 4K without upscaling is brutal on both cards. This is where DLSS 4 and FSR 4 enter the conversation directly, because neither the RX 9070 nor the RTX 5070 is intended to be run at native 4K with full ray tracing enabled in demanding titles. DLSS 4 on the RTX 5070 uses transformer-based machine learning models that produce sharper, more stable upscaled images than FSR 4’s spatial approach in most tested scenarios. NVIDIA also has Multi Frame Generation available on Blackwell, which can produce multiple AI-generated frames per rendered frame – a feature the RX 9070 simply does not have access to.
FSR 4 is a genuine step up from FSR 3.1, and AMD’s frame generation through Fluid Motion Frames 2 works across a broader range of titles than NVIDIA’s implementation. But when comparing quality at 4K with ray tracing, DLSS 4 produces cleaner output with less ghosting on fast motion. That gap narrows when FSR 4 is running at Quality mode, but it does not close entirely. For buyers who plan to run demanding ray tracing titles at 4K, the upscaling quality difference is a real factor – not an abstract one.
4K Ray Tracing Performance: Title by Title
In Cyberpunk 2077 with Ray Tracing: Overdrive enabled and DLSS/FSR set to Quality mode, the RTX 5070 delivers notably higher average frame rates than the RX 9070. Path tracing in Cyberpunk is one of the heaviest ray tracing workloads available on PC, and the RTX 5070’s RT cores handle it without the frame time spikes that show up on the RX 9070 during complex scenes – particularly interiors with multiple light sources and wet reflective surfaces. The RTX 5070 wins this workload clearly.
Alan Wake 2 tells a similar story. The game’s full ray tracing mode hammers VRAM and ray tracing hardware simultaneously, and while the RX 9070’s 16GB VRAM gives it headroom the RTX 5070’s 12GB occasionally lacks, the RTX 5070’s faster ray tracing throughput keeps it ahead in average frame rates at 4K. The RX 9070 does show less VRAM pressure in edge cases – loading into large outdoor environments with the full RT pipeline active – but frame rates still favor the RTX 5070 in most sequences.
Where the RX 9070 closes the gap meaningfully is in titles that use hybrid rendering rather than full path tracing. Hogwarts Legacy, Black Myth: Wukong, and Indiana Jones and the Great Circle all use selective ray tracing – reflections, shadows, or ambient occlusion rather than the full lighting pipeline. In these titles, the gap between the two cards at 4K shrinks considerably. The RX 9070’s improved ray accelerators handle hybrid workloads well, and paired with FSR 4 at Quality mode, it delivers playable frame rates that sit within striking distance of the RTX 5070. These are also the kinds of games most people actually play, which makes this comparison more relevant to everyday use than synthetic path tracing benchmarks.
Rasterization performance at 4K, which is the baseline for any ray tracing comparison since most games still blend rasterized and ray-traced elements, is genuinely competitive between the two cards. The RX 9070 sometimes pulls ahead in titles optimized for AMD hardware, and the RTX 5070 leads in NVIDIA-optimized titles – a predictable pattern that has defined GPU comparisons for years. The CPU platform you pair with either card also affects 4K ray tracing outcomes, since ray tracing can expose CPU bottlenecks in scene traversal that a pure rasterization workload would not.
The RTX 5070 carries a higher MSRP than the RX 9070, and in markets where actual retail pricing reflects MSRP – which has not always been the case in recent GPU generations – the RX 9070 is the better value card. If your library skews toward full path tracing titles or you want the best possible 4K ray tracing ceiling without moving to the RTX 5070 Ti or RTX 5080, the RTX 5070’s RT performance advantage justifies the premium for a specific kind of buyer. But for the majority of 4K gaming use cases, including most of the highest-profile releases of the past two years, the RX 9070 stays competitive enough that the price difference is hard to rationalize on ray tracing performance alone.
