Introduction

Today we are reviewing the ASUS TUF Gaming GeForce RTX 5090, the company's second premium custom-design graphics card based on the flagship new generation GPU by NVIDIA. For the RTX 5090, ASUS offers the ROG Astral series as its most premium custom design, followed by the TUF Gaming we're looking at today. There are no ROG Strix or Prime series models based on the RTX 5090. The TUF Gaming series has over the years risen from being a value-segment brand to something premium. These cards are endowed by fairly powerful cooling solutions with heavy heatsinks and airy cooler shrouds designed to expose most of the heatsink for airflow, which is why ASUS refers to the cooling solution as the Ventilated Exoskeleton.



The GeForce RTX 5090 needs little introduction at this point—it's the fastest gaming GPU money can buy, and is the flagship of the RTX 50-series Blackwell generation. The card is designed for 4K Ultra HD gaming with maxed out settings, including ray tracing; with DLSS 4 Multi Frame Generation unlocking new use-cases, such as 4K high refresh-rate (144 Hz or even 240 Hz), and 8K. The GPU is endowed with nearly double the number crunching machinery as the next-fastest one from the lineup, the RTX 5080, and double its memory size, with nearly double the memory bandwidth. These prove crucial not just for high-resolution gaming, but also assist with what NVIDIA is trying to accomplish with Blackwell—bring AI closer to gaming.

The new Blackwell graphics architecture introduces a potentially revolutionary new technology called Neural Rendering. You already know about the incredible power of generative AI in conjuring photorealistic images and video, and NVIDIA figured if a locally running generative AI model could create objects for the game that are combined with conventional raster 3D graphics, much like ray traced objects are. To this effect, NVIDIA worked with Microsoft to standardize the technology, giving applications direct access to the Tensor cores. The company introduced a new hardware scheduler for all the AI acceleration resources, called the AI Management Processor (AMP).

The new Blackwell generation CUDA core offers generational IPC uplifts, and concurrent FP32 and INT32 capability on all cores in an SM. The shader execution reordering engine comes with support for neural shaders. The 4th generation RT cores come with even more dedicated hardware, including preparation for Mega Geometry—a concept that increases geometric complexity of ray traced objects. The 5th generation Tensor cores come with support for the FP4 data format for even more throughput by tracing in precision. The display and media engines receive significant upgrades, including support for hardware flip-metering and 4:2:2 video formats. The former also plays a crucial role in enabling Multi Frame Generation.

Introduced with DLSS 4, Multi Frame Generation is the logical next step to Frame Generation introduced with the RTX 40-series, it lets the GPU generate up to three frames following a conventionally rendered one, entirely using AI. The DLSS 4 feature set itself sees the replacement of older convoluted neural networks (CNN) based AI models with newer transformer-based models that are more accurate, and improve image quality for upscaling, frame generation, and ray reconstruction. While Multi Frame Generation is exclusive to the RTX 50-series, the rest of the DLSS 4 feature set is available even for the RTX 40-series and RTX 30-series.

The GeForce RTX 5090 is based on the GB202, the largest GPU in the family that attains its size because NVIDIA hasn't switched to a new process node to manufacture these chips—they're based on the same NVIDIA 4N process node as the RTX 40-series Ada generation. All energy efficiency upgrades you see are purely a function of the architecture. The RTX 5090 features as many as 21,760 CUDA cores across 170 SM, along with 680 Tensor cores, 170 RT cores, 680 TMUs, and 176 ROPs. The memory subsystem sees a massive upgrade over the RTX 4090, you now get 32 GB of memory across a 512-bit wide GDDR7 memory bus, and with a speed of 28 Gbps, you have a mammoth 1,792 GB/s of memory bandwidth on tap.

The ASUS TUF Gaming RTX 5090 features the most premium version of the TUF Gaming Ventilated Exoskeleton design, with two-tone metal surfaces on the frame, a trio of the company's latest Axial-Tech fans, and a heavy cooling solution that uses a vapor chamber plate to pull heat from the GPU. There are as many as 12 heat pipes that transfer heat from the GPU and memory across the aluminium fin-stack. The TUF Gaming comes at reference clock speeds, but there's also a TUF OC model that's clocked higher. ASUS is pricing the TUF Gaming RTX 5090 at $2,450, a $450 premium over the $2000 MSRP, although we've seen this card sell for nearly $4,000.

Packaging

The Card

The ASUS TUF color theme for GeForce RTX 50 Series is very similar to what we saw on the GeForce 40 Series. Metal surfaces are combined with various shades of gray and black. Both the main cooler and backplate are made of metal.


Dimensions of the card are 35.0 x 15.0 cm, and it weighs 2317 g.


Installation requires four slots in your system. We measured the card's width to be 70 mm.


Display connectivity includes three standard DisplayPort 2.1b and two HDMI 2.1b.

Standard for all GeForce RTX 50-series Blackwell cards is a new display engine that supports three DisplayPort 2.1b outputs, each capable of UHBR20; and one HDMI 2.1a. Both interfaces support DSC (display stream compression). With DSC enabled, a single DisplayPort on this card can drive 4K 12-bit HDR at 480 Hz; or 8K 12-bit HDR at up to 165 Hz. The RTX 5090 features an updated media acceleration engine with support for 4:2:2 video formats, AV1 UHQ, and MV-HEVC. There are three independent NVENC units, and two NVDEC.


The card uses a single 16-pin connector, which allows a maximum power draw of 600 W.


ASUS has installed an RGB lighting zone near the corner of the card.


This BIOS switch lets you select between the default "Performance" BIOS and a "Quiet" BIOS that runs a more relaxed fan curve.

Teardown

The fan assembly can be removed easily, which makes it easy to replace a broken fan in a couple of years—no need to touch the thermal paste of the card.


ASUS has installed 12 heatpipes on their thermal solution, the main heatsink also provides cooling for the memory chips and VRM circuitry.


The backplate protects the card against damage during installation and handling. Note the lines pattern, it's not embossed but printed—great idea.

High-resolution PCB Pictures

These pictures are for the convenience of volt modders and people who would like to see all the finer details on the PCB. Feel free to link back to us and use these in your articles, videos or forum posts.


High-resolution versions are also available (front, back).

Circuit Board (PCB) Analysis

A 24-phase VRM powers the GPU, highlighted in the red rectangles above. These are controlled by a Monolithic Power Systems MP29816 controller.


All GPU power phases use Vishay SiC654A DrMOS, rated for 50 A.


Powering the GDDR7 memory chips is a 7-phase VRM driven by a uPI uP9512R controller.


Just like GPU, the memory is handled by Vishay SiC654A DrMOS chips.


The GDDR7 memory chips are made by Samsung, and bear the model number K4VAF325ZC-SC28, they are rated for 28 Gbps.


The NVIDIA GB202 GPU at the heart of the GeForce RTX 5090 is massive. It is fabricated using a 5 nanometer "NVIDIA 4N" process at TSMC Taiwan (same process as Ada). The die measures 750 mm², and comes with a transistor count of 92.2 billion.

Test System

Benchmark scores in other reviews are only comparable when this exact same configuration is used.

• All games and cards are tested with the drivers listed above—no performance results were recycled between test systems. Only this exact system with exactly the same configuration is used for all results in this review.
• All graphics cards are tested using the same game version.
• All games are set to their highest quality setting unless indicated otherwise.
• All games are running at their native resolution without upscaling (no DLSS or FSR), except for the results on the "DLSS 4" page where we specifically look into that feature's performance
• All games have ray tracing disabled, except for testing on the "Ray Tracing" page
• AA and AF are applied via in-game settings, not via the driver's control panel.
• Before starting measurements, we heat up the card for each test to ensure a steady state is tested. This ensures that the card won't boost to unrealistically high clocks for only a few seconds until it heats up, as that doesn't represent prolonged gameplay.
• For better real-life applicability, all game tests use custom in-game test scenes, not the integrated benchmarks
• All cards used for comparison are reference designs. When a reference design does not exist, we go the extra mile and buy the closest possible match, using reference clocks and default power limit.

Each game is tested at these screen resolutions:

• 1920x1080: Most popular monitor resolution.
• 2560x1440: Intermediary resolution between Full HD and 4K, with reasonable performance requirements.
• 3840x2160: 4K Ultra HD resolution, available on high-end monitors.