Introduction

The ASUS ROG Astral GeForce RTX 5080 OC is the company’s most premium custom design graphics card powered by NVIDIA’s new enthusiast-segment GPU. ASUS designed the ROG Astral brand to be positioned above its ROG Strix series, offering a combination of premium aesthetics and powerful air cooling. This is the only air-cooled graphics card in its segment to feature four large fans ventilating a heavy heatsink. The ROG Astral is designed to visually pair well with not just the company’s ROG Strix line of premium motherboards, but also enthusiast-segment ROG Maximus and ROG Crosshair boards. The RGB LED lighting isn’t as flashy as the ones you find on ROG Strix series products—it’s subtle and neatly executed—but what draws your attention are the textured metal alloy surfaces. It’s no surprise that the ROG Astral RTX 5080 OC also comes with the company’s highest factory overclock for the GPU.

The GeForce RTX 5080 is designed to meet much of the same use-case as the flagship RTX 5090 we reviewed last week. You get to play any of today’s AAA games at 4K Ultra HD with maxed-out settings and ray tracing enabled. The RTX 5080 is targeted squarely at gamers, while the RTX 5090 addressed other segments such as AI development and acceleration. The RTX 5080 comes in at a starting price that’s exactly half that of the RTX 5090, and if you go through its specs sheet, many of the numbers you see are either nearly or exactly half the ones on the RTX 5090 specs list, too.

The RTX 5080 debuts NVIDIA’s second-largest gaming GPU based on the Blackwell graphics architecture, the GB203. This has very similar die-area and transistor counts to the previous-generation AD103 powering the RTX 4080, because NVIDIA is building the GeForce Blackwell generation on the exact same 5 nm NVIDIA 4N foundry node at TSMC. Whatever performance-per-Watt gains you see are purely by merit of Blackwell and its redesigned power-management architecture.

The GeForce RTX 5080 maxes out the GB203 silicon, enabling all 84 streaming multiprocessors (SM) present on it, besides the chip’s full 256-bit GDDR7 memory interface, and two each of the latest generation NVDEC and NVENC video accelerators. All 64 MB of the chip’s on-die L2 cache is enabled for the RTX 5080. With all SM enabled, the RTX 5080 enjoys 10,752 CUDA cores—a numerically higher figure than the 10,240 of the RTX 4080 SUPER. It also gets all 336 Tensor cores, 84 RT cores, 336 TMUs, and 112 ROPs. The GPU runs at speeds of over 2500 MHz, which ASUS further overclocked, while the memory is a lavish 30 Gbps GDDR7, working out to 960 GB/s of bandwidth, a 34% increase over the RTX 4080. The GPU will need this bandwidth to drive two of the architecture’s highlights—Neural Rendering and DLSS 4 Multi Frame Generation.

The Blackwell architecture debuts a new concept in consumer 3D graphics called Neural Rendering, which lets the GPU use a generative AI to create certain neural objects, and combine them with classic raster 3D graphics, just like NVIDIA discovered how to combine ray traced objects with raster 3D with the very first RTX. Generative AI hence plays a more participatory role in the rendering stack, and isn’t just used to reconstruct details in DLSS super resolution. To make this happen, NVIDIA innovated a way for 3D graphics applications to directly access the GPU’s Tensor Cores, and created a hardware scheduler called the AI management processor (AMP), letting the GPU accelerate generative AI models and render 3D graphics in tandem.

DLSS 4 introduces many new changes. It replaces the convoluted neural network (CNN) based AI models used in its subcomponents with new transformers based models, which are more accurate, and hence increase image quality in every performance setting. Transformer models replace CNNs for super resolution, ray reconstruction, and frame generation. DLSS 4 also introduces the Blackwell-exclusive Multi Frame Generation (MFG), which uses AI to conjure up not just every single frame succeeding a conventionally rendered one, but up to four succeeding frames, effectively making one conventionally rendered pixel succeed with up to sixteen AI generated ones. DLSS 4 MFG can hence increase smoothness and apparent frame rates by over 4 times, unlocking new use cases such as 8K 60 Hz gaming.

The ASUS ROG Astral RTX 5080 OC features a new cooling technology from the company called Quad Fan Force. It involves three large fans along the front of the card, and a single large fan along its backplate. The PCB underneath is barely two-thirds the length of the card, so a portion of the airflow from the middle fan and all of it from the third one goes through the heatsink and out a large vent along the backplate. ASUS decided to add a fourth “pull” fan here, increasing air pressure through the directly ventilated portion of the heatsink. This increased airflow allows ASUS to run the fans at conservative speeds, lowering noise. ASUS has given the ROG Astral RTX 5080 OC its highest factory overclock, with the GPU boosting up to 2760 MHz compared to 2617 MHz reference. ASUS is pricing the card at $1500, a shocking 50% premium over the $1000 baseline price for the RTX 5080.