Introduction

ASUS TUF Gaming GeForce RTX 5080 OC is the company's third premium custom-design graphics card based on the GPU. The TUF Gaming brand of graphics cards from ASUS provide a balance of cooling performance and value, while not being as flashy as the company's ROG Strix or ROG Astral graphics card lines. The card comes with a powerful cooling solution ASUS calls the Vented Exoskeleton. A meaty aluminium fin-stack heatsink is ventilated by a trio of axial airflow fans. The cooler shroud is designed both for rigidity, and to expose most of this heatsink outside, so its airflow moves out unimpeded. The GeForce RTX 5080 is NVIDIA's new enthusiast-segment product positioned below the RTX 5090 flagship.



The GeForce RTX 5080 is designed for maxed out gaming at 4K Ultra HD, including with ray tracing. This is something even the flagship RTX 5090 does, but while that card has other use-cases such as AI compute acceleration and pro-visualization thanks to its 32 GB memory; the RTX 5080 is a more laser-focussed product for gamers and game-streamers playing at 4K. The RTX 5080 is powered by the new GeForce Blackwell graphics architecture, which introduces Neural Rendering, a potentially revolutionary concept in consumer 3D, where generative AI models are leveraged to create objects in the scene in real time, which offload the raster graphics machinery. You probably know the incredible capabilities of generative AI, and can imagine where NVIDIA is headed with this. NVIDIA worked with Microsoft to standardize this at the DirectX API level. Neural Rendering is exclusive to Blackwell as the architecture introduces a new hardware scheduler called the AI Management Processor (AMP), letting the GPU accelerate AI and render graphics in tandem.

The other big features being introduced with this generation are DLSS 4 and Multi Frame Generation. DLSS 4 replaces the convoluted neural networks (CNN) based AI models driving DLSS with a new Transformer-based model, which is more accurate, and introduces image quality improvements at every performance preset. DLSS 4 implements Transformer-based AI models for upscaling, ray reconstruction, and frame generation. Speaking of which, the new Multi Frame Generation feature allows the GPU to generate up to 3 frames following a conventionally rendered frame, completely using AI, effectively quadrupling frame rates at any given image quality. When used in combination with upscaling, this can increase frame rates by larger percentages, as every pixel is being used to generate up to 15 other pixels. While DLSS 4 and its Transformer models are being extended to even the RTX 40-series Ada and RTX 30-series Ampere, Multi Frame Generation is exclusive to Blackwell because it requires hardware flip-metering that's being introduced with the new display engine of Blackwell.

NVIDIA updated its display engine to support DisplayPort 2.1b with UHBR20, which can drive compatible 8K displays with a single cable. The new media acceleration capabilities of Blackwell introduce 4:2:2 color support for AV1, HEVC, and H.264, both for encoding and decoding, significantly improving bitrates without affecting quality. Blackwell also introduces AV1 B-frame support, that significantly lowers file size or stream bitrate.

The GeForce RTX 5080 introduces GB203, the second-largest silicon based on the GeForce Blackwell architecture. NVIDIA is sticking to the same exact process node it used to build the RTX 40-series Ada—NVIDIA 4N, and so all energy efficiency improvements you see are purely a function of the architecture. The RTX 5080 maxes out the GB203 silicon, enabling all 84 streaming multiprocessors (SM) present on the silicon, which work out to 10,752 CUDA cores, 336 Tensor cores, 84 RT cores, 336 TMUs, and all 112 ROPs present on the GB203 silicon. The card offers 16 GB of GDDR7 memory, and maxes out the 256-bit wide memory bus of the GB203 silicon. With a memory speed of 30 Gbps, the RTX 5080 enjoys 960 GB/s of memory bandwidth.

The ASUS TUF Gaming RTX 5080 OC comes with factory overclocked speeds for the RTX 5080, with the GPU boosting up to 2700 MHz, compared to 2617 MHz reference, while the memory is left untouched at 30 Gbps (GDDR7-effective). The cooler uses a nickel-plated copper baseplate to pull heat from the GPU and memory; from which 7 heat pipes spread it to two aluminium fin-stacks. These are ventilated by three 100 mm Axial-Tech fans with double ball bearings and an impeller designed for axial airflow. ASUS is pricing the card at an MSRP of $1375, sold out everywhere, scalpers want 1500 USD. In Europe the card is readily available at a price of €1475, which converts to $1375 without VAT.