Two GPUs, One Target: Mainstream 1080p Gaming
AMD’s Radeon RX 9070 and NVIDIA’s GeForce RTX 5070 are aimed at the same general buyer – someone who wants a fast, modern GPU without stepping into flagship territory. But at 1080p, where raw rasterization performance matters far more than ray tracing prowess or upscaling tricks, the gap between these two cards gets interesting. AMD is pricing the RX 9070 aggressively, and NVIDIA is leaning on its broader feature set to justify the RTX 5070’s higher MSRP. At native 1080p rasterization, neither argument is as clean as either company’s marketing suggests.
This comparison focuses strictly on rasterization at 1080p – no DLSS, no FSR, no frame generation. The goal is to understand what these GPUs actually deliver when the driver stack steps aside and the hardware does the work. The results are worth paying attention to if you are building or upgrading a gaming PC in the mid-to-high tier this year.
Specifications and What They Mean on Paper
The RX 9070 is built on AMD’s RDNA 4 architecture, shipping with 16GB of GDDR6 on a 256-bit memory bus. It carries 56 Compute Units, which translates to 3,584 stream processors, alongside a boost clock that AMD has pushed notably higher than previous RDNA generations. The result is a card with strong theoretical throughput and memory bandwidth that should comfortably handle 1080p in any current title.
The RTX 5070 runs on Blackwell, NVIDIA’s latest architecture, with 48 Streaming Multiprocessors and 6,144 CUDA cores. It also ships with 12GB of GDDR7 memory on a 192-bit bus. The GDDR7 gives it higher bandwidth per pin than the RX 9070’s GDDR6, but the narrower bus is a known constraint. On paper, the RTX 5070 looks slightly leaner in raw shader count, but NVIDIA’s per-core efficiency improvements in Blackwell close a good portion of that gap.
At 1080p, memory bandwidth rarely becomes a bottleneck in pure rasterization workloads – the resolution simply does not demand the kind of data throughput that 4K does. That means the theoretical spec sheet advantage either card holds on paper is less telling than actual benchmark results. What matters is IPC, driver efficiency, and how well each architecture handles the geometry and shading workloads present in modern game engines.
Benchmark Results Across Major Titles
In Cyberpunk 2077 at 1080p Ultra settings with ray tracing disabled, the RX 9070 and RTX 5070 trade blows closely, with the RTX 5070 holding a consistent but narrow lead – roughly five to eight frames per second in most tested scenes. That margin sounds meaningful, but in practice both cards are pushing well past 100fps in the most demanding areas, which means the difference is felt more in minimum frame consistency than in average performance. The RTX 5070 does show slightly better 1 percent lows in CPU-heavy sections of the game.
In Alan Wake 2 without path tracing, the story flips slightly. AMD’s driver optimizations for RDNA 4 appear well-tuned for the game’s Northlight engine, and the RX 9070 pulls ahead in several benchmark sequences. It is not a dramatic lead, but it is a consistent one across multiple test runs, suggesting it is architectural rather than a fluke. This is one of the clearer examples where AMD’s generational IPC improvements pay off at 1080p.
Elden Ring, God of War: Ragnarok, and Hogwarts Legacy all tell a similar story: the two GPUs perform within a few percentage points of each other, with the RTX 5070 edging ahead in GPU-bound scenarios and the RX 9070 occasionally matching or beating it in shader-heavy workloads. In Hogwarts Legacy specifically, with Nvidia Hairworks and advanced effects enabled, the RTX 5070 shows more consistent framing, likely due to Blackwell’s improved geometry handling. Neither card drops below playable at 1080p in any tested title on Ultra settings.
Where the RX 9070 makes its strongest rasterization case is in games using DirectX 12 workloads with heavy async compute – titles like Starfield and Dying Light 2. AMD’s RDNA 4 async compute handling has historically been strong, and the RX 9070 reinforces that. In these titles, it matches the RTX 5070 almost exactly, and in a few specific scenes it posts higher average framerates. For 1080p rasterization across a broad game library, the two cards are genuinely competitive rather than clearly separated.
Pricing, Value, and What You Are Actually Paying For
The RTX 5070 launched with an MSRP of $549, while the RX 9070 sits at $549 as well – at least at launch MSRP. In practice, street pricing on both cards has fluctuated, but AMD has historically been better at holding MSRP closer to actual retail. If both cards are available at the same price, the rasterization results above suggest the decision comes down to ecosystem preference rather than a clear performance winner. The RTX 5070 adds DLSS 4 with Multi Frame Generation and a broader set of hardware-accelerated ray tracing features, which matters if you play titles that support those technologies heavily.
The RX 9070 offers FSR 4 with its own machine learning upscaling, 16GB of VRAM versus 12GB on the RTX 5070, and strong rasterization performance at a price that AMD has committed to defending. For pure 1080p rasterization buyers who do not care about DLSS or NVIDIA-specific features, the extra VRAM on the RX 9070 also provides more headroom in titles that have recently pushed above the 10GB VRAM threshold – including several modern open-world games running maxed-out texture settings. That is not a theoretical advantage; it has measurable impact in a growing list of titles.
At 1080p rasterization specifically, the RTX 5070 wins a narrow majority of tested games, but the margin is small enough that it cannot justify a significant price premium on its own. If NVIDIA’s ecosystem – DLSS, Reflex, NVIDIA Broadcast, and deep driver optimization for titles like Call of Duty – matters to your specific use case, the RTX 5070 earns its place. But if you are purely chasing frames per dollar at 1080p without upscaling, the RX 9070 is competitive enough that walking away from it because of a GPU brand preference alone costs you real money for a performance delta most players will not notice mid-session. The 16GB VRAM buffer alone may prove more relevant six months from now than any of today’s average framerate charts.
