AORUS GeForce RTX 5060 Elite Review

To NVIDIA, the future of graphics cards undoubtedly goes through the Artificial Intelligence. With the arrival of Blackwell, its new graphics architecture for the RTX 50 series, the company seeks to unify all the advances obtained in business AI, data science and supercomputing, to apply them in the world of Gaming and real time renderingThis includes significant improvements in energy efficiency, support for updated codecs, new graphics generation techniques, and optimized neural network processing.

The pillar of this generation remains DLSS (Deep Learning Super Sampling), which in its 4th version not only increases the fluidity of games, but also reduces the load on the hardware thanks to multi-frame generation and ray reconstruction by AI. In addition, the new neural shaders (Neural Shaders), a feature that seeks to gradually replace part of the traditional graphics pipeline logic with accelerated inference using neural networks. All this is accompanied by a new version of DLSS Ray Reconstruction, which optimizes the quality of ray tracing without the usual computational weight.

Let's start with the basics: the GeForce RTX 5060 specs.

La GeForce RTX 5060 It is based on the graphics chip GB206-250-A1, manufactured on the custom node TSMC4N. It has 3,840 CUDA cores, distributed in 30 SMs (Streaming Multiprocessors), a figure that represents a generational leap compared to its predecessor, the RTX 4060. This graphics card is launched with 8GB of GDDR7 memory a 28 Gbps, connected through an interface 128 bits. Overall, it offers a 448 GB/s bandwidth, which represents a considerable improvement in data transfer compared to previous generations with GDDR6X or traditional GDDR6.

One of the most striking aspects is the debut of the GDDR7 memory in the mid-range. Thanks to the use of signaling PAM3, greater energy efficiency and better performance are achieved signal-to-noise ratioThis allows for higher channel density and, therefore, higher performance without increasing power consumption. In practice, this translates into faster loading times, improved texture response in open-source games, and more stable performance in VRAM-intensive titles.

In terms of specialized computing units, the RTX 5060 maintains the 4th generation RT cores and 5th generation Tensor Cores. This allows you to fully exploit artificial intelligence technologies applied to rendering, such as DLSS-4, Ray Reconstruction, DLAA, frame generation, and new techniques such as neural shaders, which combine machine learning algorithm with traditional graphic programming.

The 5th generation Tensor Cores introduce support for the format FP4, which allows data to be represented with minimal memory usage and virtually no loss of precision compared to FP16. This improvement has a direct impact on the efficiency of AI models like DLSS or real-time neural network inferences. In a game, this can mean the difference between achieving 90 FPS or 60 FPS when Frame Generation is enabled.

The RT cores now double intersection testing capacity (ray-triangle intersection tests), reducing the bottlenecks that usually affected real-time ray tracing. Also included are improvements such as the micromapping opacity engine, which reduces the calculations required for complex effects such as transparencies or particles, and technologies such as mega geometry, sweeping spheres y grouped triangle acceleration, useful for representing detailed geometry such as hair, leaves, fabric, and complex reflections.

The graphics engine idTech8 and advanced engines such as Unreal Engine 5 o Unity HDRP They will be able to take advantage of these capabilities in conjunction with techniques such as Lumen, Nanite, or Global Illumination by ray tracing, offering more realistic environments with a mid-range GPU.

In compatibility, the Blackwell architecture offers full support for DirectX 12 Ultimate, Vulkan y OptiX 9.0. In addition, compatibility with Shader Execution Reordering (SER) 2.0, which dynamically reorganizes shader execution to optimize parallel loading, a feature especially useful in engines with dynamic lighting or physics-based animation.

The RTX 5060 also incorporates the new coprocessor AI Management Processor (AMP), based on architecture RISC-VThis component acts as a context scheduler, reducing the load on the CPU in Windows. By redistributing tasks such as memory management and frame synchronization, it achieves lower latency and better multitasking performance, ideal for those who work in the office. streaming or run multiple applications in the background.

The system DLSS-4 It's the heart of the Blackwell generation. This AI technology combines multiple components: multi-frame generation, ray reconstruction, super-resolution, and deep learning antialiasing. Together, they allow for improved graphics quality and performance at the same time, something that seemed contradictory just a few years ago. NVIDIA claims that DLSS 4 is up to 40% faster than DLSS 3, using 30% less video memory thanks to more compact AI models.

La Multi-frame Generation now works entirely through AI, eliminating dependence on the optical flow accelerator hardware. This allows for greater compatibility with various graphics engines and reduces thermal load. In addition, the Flip Metering allows you to adjust the presentation of frames more precisely, improving visual fluidity.

DLSS Super Resolution (SR) Reconstructs a high-resolution image using motion data, depth vectors, and previous frames. This results in sharper, more stable images with less ghosting on moving objects, ideal for competitive or open-world titles.

DLSS Ray Reconstruction (RR) replaces traditional noise removers with a trained neural network, producing more natural results, with smooth shadows and more accurate reflections in complex environments. Even on cards like the RTX 5060, this translates to a tangible visual improvement without the need for high-end hardware.

DLAA (Deep Learning Anti-Aliasing) is also part of the package, offering smoother edges and superior visual stability to traditional techniques such as TAA or FXAA, without the need to scale resolution.

With neural shadersNVIDIA is looking to redefine the graphics pipeline. These are AI-trained code blocks that replace traditional mathematical functions with more efficient inferences, from indirect lighting to texture compression, subsurface scattering, and facial animation.

Other notable technologies include: RTX Neural Texture Compression, which allows textures to be compressed and decompressed using neural networks with greater efficiency than traditional methods. Neural Radiance Cache (NRC) stores complex lighting patterns to generate high-fidelity global illumination and reflections. RTX Skin allows to represent human skin with light scattering effect, and Neural Faces reconstructs more realistic human faces using AI layers on 3D meshes.

As for video outputs, the RTX 5060 includes HDMI 2.1 y DisplayPort 2.1b, supporting resolutions up to 8K at 165 Hz or 4K at 480 Hz, allowing you to get the most out of high-end gaming monitors, especially at high resolutions with technologies like G-SYNC.

Finally, the energy consumption of the RTX 5060 is 115W TGP, which represents an excellent balance between efficiency and performance, allowing for compact configurations, simple cooling, and mid-range power supplies without compromising the gaming experience.

For more extensive coverage on what's new in the Blackwell architecture and DLSS 4, you can check out our article on the Editor's Day from NVIDIA during CES 2025.