Radeon RX 9060 XT 16GB Review – Performance, Benchmarks & Value

AMD continues its rollout of the new RDNA 4 graphics architecture with the Radeon RX 9060 XT. Targeting the heart of the market, it aims at gamers who play at 1080p and 1440p and are mindful of their budget.

However, this is not its only mission. It also aims to challenge Nvidia with highly competitive suggested retail prices, offering 8 GB of VRAM at a lower cost than the GeForce RTX 5060 Ti 8 GB. The latter is announced at suggested retail prices of $379 (€399) and $429 (€449) for the 8 GB and 16 GB versions, respectively, compared to $299 and $349 for the Radeon RX 9060 XT 8 GB and 16 GB versions. During its presentation, AMD claimed an average performance improvement of 46% over its predecessor, the Radeon RX 7600 XT 16 GB, at 1440p resolution.

For this review, we tested the Radeon RX 9060 XT 16 GB Reaper from PowerColor. It was compared against the Radeon RX 7600 8 GB and RX 7600 XT 16 GB, the GeForce RTX 4060 Ti 16 GB, the RTX 4060 8 GB, and the RTX 5060 Ti 16 GB at 1080p and 1440p with maximum settings, with and without Ray Tracing.

RDNA 4 Graphics Architecture

AMD specifies that RDNA 4 is the result of a long multigenerational journey with RDNA, focusing on several objectives, including meeting today’s trends while preparing for the future to immerse itself in the titles of the coming years. The focus with these Radeon RX 9070 series cards is to target the market of cards with suggested retail prices below $600.

RDNA 4 Architecture

RDNA 4 promises to improve performance in many key areas. To achieve this, compute units (CUs, aka Compute Units) benefit from optimizations, and 3rd generation Ray Tracing accelerators accompany 2nd generation AI accelerators. The GPU also features an improved multimedia engine, and boost frequencies approach 3 GHz.

Compute Units

The RDNA 4 architecture promises to improve general workloads (rasterization) and compute. Even though we are gradually seeing the use of new rendering techniques, these fundamental workloads remain at the heart of most real-time rendering experiences.

The work done on the RDNA 4 compute unit improves the memory subsystem and scalar units while enabling dynamic register allocation, all with much higher frequencies compared to RDNA 3. Collectively, this boosts performance per CU, allowing the Radeon RX 9070 Series to offer performance comparable to the Radeon RX 7900 despite a lower total number of CUs.

3rd Generation Ray Tracing Accelerators

This is probably the most evolved part with this generation to make up for an obvious lag compared to Nvidia solutions. These 3rd generation RT accelerators promise to offer twice the Ray Tracing capabilities of RDNA 3. In Ray Tracing, one of the important elements associated with heavy workloads is the construction of the BVH data structure, short for Bounding Volume Hierarchy.

Its role is to improve rendering performance by reducing the number of collision tests needed between rays and scene objects. When a graphics card executes a Ray Tracing algorithm, it must determine which objects are hit by a light ray. Checking each object individually would be extremely computationally expensive and therefore slow, especially in a scene containing millions of triangles. The BVH then organizes these objects in a hierarchy of bounding boxes (Bounding Volumes) that allow large portions of the scene to be quickly eliminated without having to test each triangle.

When a ray is launched into the scene, it passes through this BVH tree hierarchically. If a ray does not hit a bounding box, then all the objects inside that box are ignored, which avoids unnecessary calculations. On the other hand, if a ray passes through a box, the algorithm goes down the hierarchy to test the smaller boxes until it reaches concrete objects.

AMD graphics cards with Ray Accelerators have dedicated units for BVH acceleration. These units allow for efficient construction and traversal of the hierarchy to improve real-time rendering.

Returning to RDNA 4, AMD introduces a particular approach to BVH called “Oriented Bounding Boxes.” It works to reduce the size and complexity of BVH data. These changes improve Ray Tracing processing while reducing VRAM requirements.

A second ray intersection engine is introduced at the heart of the RT accelerators, which doubles performance for Ray/Box and Ray/Triangle tests. At the same time, a dedicated ray transformation block is implemented to increase performance when rays are traversed in the lower levels of the BVH tree.

2nd Generation AI Accelerators

The 2nd generation AI accelerators of RDNA 4 promise to process advanced AI models more efficiently through a combination of different changes, improvements, additions, and optimizations. For example, AMD is adding more math pipelines for AI calculations, increasing the accelerator’s capabilities to support new data types such as FP8 and inference optimization techniques such as Structured Sparsity. This technique helps GPUs calculate faster by intelligently removing certain unnecessary values in neural networks while maintaining acceptable precision performance.

It reduces the number of calculations, knowing that a portion of the multiplications are removed, allowing operations to execute faster while relieving memory resources. Storing lighter matrices saves memory bandwidth and, more generally, guarantees better inference performance. For example, in tasks like image recognition or language processing, this allows for more requests to be processed per second.

Multimedia Engine

Regarding the media engine, improvements focus on the quality of encoding during recording and streaming through all major codecs such as H.264, HEVC, and AV1. Thus RDNA 4 supports:

• Decoding: VP9 – 4K210 / 8K48 / (4:2:0 8/10b), H.264 – 4K330 / (4:2:0 8b), H.265 – 4K210 / 8K48 / (4:2:0 8/10b), AV1 – 4K240 / 8K60 / (4:2:0 8/10/12b).
• Encoding: H.264 – 4K180 / (4 :2 :0 8b), H.265 – 4K180 / 8K48 / (4:2:0 8/10b), AV1 – 4K240 / 8K60 / (4:2:0 8/10b).

FSR 4

FidelityFX Super Resolution 4 (FSR 4) is a machine learning (ML)-based upscaling technology designed to transform low-resolution images into high-definition output. Unlike previous versions, which mainly relied on classic algorithms, FSR 4 uses an advanced AI model to improve image quality and optimize performance.

The process begins with low-resolution inputs and critical scene data, such as depth, colors, and motion vectors. This information allows the algorithm to better understand the positioning of objects, lighting, and movements in the scene. With this data, the FSR 4 upscaling algorithm applies AI processing accelerated by the RDNA 4 AI accelerators. This approach allows FSR 4 to function in real time even with high refresh rates, which is essential for competitive gaming and immersive experiences.

High-definition upscaling is then promised, offering more precise details, better sharpness, and a reduction in visual artifacts. FSR 4 can be coupled with advanced image generation technologies that insert additional frames to smooth the experience while attempting to ensure high visual quality.

As you may have guessed, FSR 4 requires an RDNA 4 GPU with dedicated AI accelerators, which limits its accessibility to Radeon RX 9000 series cards and games that support FSR 3.1 or later. Despite this constraint, this advance brings AMD closer to solutions like NVIDIA’s DLSS, which also use AI to improve real-time rendering.

Radeon RX 9060 XT 16 GB Reaper

As we have seen, the RDNA 4 architecture brings several major advances, including 3rd generation Ray Tracing accelerators, 2nd generation AI accelerators, and an advanced multimedia engine that supports hardware reading and recording of the most popular codecs like H.264, HEVC, and AV1.

The Radeon RX 9060 XT 16 GB is built around an RDNA 4 GPU from TSMC’s N4P process. On a surface area of 199 mm2, we find 29.7 billion transistors. All this allows us to build 32 compute units (2048 stream processors), 32 Ray Tracing accelerators, 64 ROPs, and 64 AI accelerators, not to mention a 3rd generation Infinity Cache of 32 MB. The frequency is calibrated at 2530 MHz in game mode and can reach 3130 MHz in boost mode. Added to this are 16 or 8 GB of GDDR6 at 20 Gbps used with a 128-bit memory bus.

The card features a PCIe 5.0 x16 interface, a notable difference, especially with older PCIe 4.0 or PCIe 3.0 configurations compared to Nvidia’s GeForce RTX 5060 Ti, calibrated in PCIe 5 x8. AMD announces a TBP, also known as Total Board Power, of 160 Watts in the 16 GB version and 150 Watts in the 8 GB version.

In its presentation, AMD compares the Radeon RX 9060 XT 16 GB to the GeForce RTX 5060 Ti 8 GB. This choice may seem strange since the amount of VRAM is not identical, which will play in Nvidia’s favor. However, there is logic to this comparison since the GeForce RTX 4060 Ti 16 GB has a lower suggested retail price than the GeForce RTX 5060 Ti 8 GB.

Summary of Specifications

PowerColor Radeon RX 9060 XT 16 GB Reaper

The Radeon RX 9060 XT is not available in an MBA version, which stands for Made by AMD. Therefore, you have to turn to partners. For this review, we tested the Radeon RX 9060 XT 16 GB Reaper from PowerColor.

The card comes in a small cardboard box without frills. The bundle is reduced to the bare minimum, with nothing else besides the card.

Its form factor is compact with a dual-slot chassis requiring a space of 20 cm to fit. Its single 8-pin PCIe connector is positioned on the side.

The black casing is made of plastic. It overhangs a massive heatsink occupying the entire length of the PCB.

Composed of three radiators with thin aluminum fins oriented vertically, it is armed with copper heat pipes to accelerate and optimize heat transfers.

The whole is accompanied by a backplate and two 9-blade fans attached to an external ring, a choice intended to improve static pressure and the orientation of the airflow so that it is as efficient as possible.

On the equipment side, there is no RGB, and the connectivity includes two DisplayPort video outputs and one HDMI.

Test Protocol

We used games, several resolutions, several benchmarks, and GPGPU applications. For each title, all rendering parameters are at maximum.

Test Platform

• Motherboard: Asus ROG Maximus Z690 HERO
• Processor: Intel Core i9-12900K
• RAM: Trident Z5 RGB 2 x 16 GB DDR5-6000 CL28
• Storage: Kioxia Exceria Pro 2 TB SSD
• Power Supply: NZXT C1500 Platinum
• Cooling: DeepCool LS720 WaterCooling

Video Games

• Metro Exodus
• Far Cry 6
• Shadow of the Tomb Raider
• Horizon Zero Dawn
• Watch Dogs Legion
• Cyberpunk 2077
• Star Wars Outlaws

Benchmarks

• 3DMark Fire Strike (Performance, Extreme)
• Blender (Monster, JunkShop, and Classroom scenes)
• LuxMark 3.1
• FAHBench
• IProcyon – AI Text Generation (PHI 3.5, Mistral-7B, Llama 3-1 and Llama-2 models)

We also tested power consumption, cooling performance, operating noise, and frequency stability.

Cooling Performance

The entire monitoring is ensured by GPU-Z. We stressed this solution for 10 minutes by maximizing its GPU load. The tests are done in open air to abstract the cooling performance of the case.

Temperatures

The two fans of the card are able to stop in the event of a low load. PowerColor mentions this mode on the condition that the GPU temperature does not exceed 60°C. At idle, 48°C is reached but not exceeded, so the card is inaudible.

After 10 minutes of intensive testing with a GPU load rate of 100%, it stabilizes at around 60°C, which is perfect. The ventilation is far from its maximum. The two fans are running at 1600 rpm, or about 50% of their maximum speed. The VRAM is warmer on its side with temperatures between 7 and 89°C!

Noise Levels

At idle, the result is perfect with an inaudible card. Under load with two fans at 1600 rpm, the increase in dBA is well controlled. The card is not noisy with a set under 40 dBA.

Frequency Stability

The control of temperatures allows to maintain, over time, optimal power. The GPU frequency is stable with an average value of 2806 MHz against 2502 MHz for the VRAM. The card maintained its full potential throughout our tests.

Power Consumption

This power consumption measurement corresponds to the minimum and maximum needs of the card to be able to make a global evaluation of its operating cost and its carbon footprint. Note that the TGP is the maximum power limit for the GPU boost frequency. It will be optimized according to this limit but also other parameters such as temperature. This power can be reached with very demanding applications.

We are looking at an energy demand between 8 and 161 Watts depending on the load level. The card consumes less than an RTX 5060 Ti and as much as the Radeon RX 7600.

Rasterization Performance

Now let’s look at performance. You will find results in Rasterization, Ray Tracing (Ultra mode), and Ray Tracing (Ultra mode) + FSR (Performance mode). In these results, we take the performance of the GeForce RTX 5060 Ti 16 GB as a reference. We focused here on 1080p and 1440p with maximum graphics options.

With our list of games mixing old and new titles, we are almost at the level of the GeForce RTX 5060 Ti 16 GB, whether at 1080p or 1440p. The respective difference is 3 and 4%.

The 3DMark Fire Strike benchmark returns equivalent results. The RTX 5060 Ti is in the lead with however larger gaps. The lead is 16.8% at 1080p and 14.3% at 1440p.

We are on a card capable of meeting the needs at 1080p without worrying too much about the graphics options. The most recent titles require a little attention if you want to maintain 60 fps at a minimum. At 1440p, the situation gets a little more complicated. If this resolution does not pose any problems with old games (>90 fps), the framerate can drop below 60 or even 45 fps. In this case, to guarantee optimal gameplay, you will have to play with the graphics options.

Ray Tracing Performance

On this terrain, Nvidia continues to dominate AMD. The RTX 5060 Ti 16 GB stands out by taking the lead. The Radeon RX 9060 XT 16 GB is 9% less performant at 1080p and 8% at 1440P.

The activation of Ray Tracing generates significant calculation needs so that with some titles like Star War Outlaws, the framerate can collapse below the fateful 30 fps.

In such a context, the game is not playable and the activation of an upscaling technology like FSR can transform the experience while maintaining maximum graphics options and the activation of Ray Tracing.

Performance with FSR and Frame Generation

The FSR 2 Feature Test of 3DMark allows you to get a first idea of the potential of this technology. At 1440p, the cap is impressive as it reaches a factor of 2.4. In other words, the number of frames per second goes from 46.4 to 113.

In practice under Cyberpunk 2077, the framerate climbs to 141 at 1080p and 70 fps at 1440p while we record 127 fps at 1080p under Star War Outlaws and 67 at 1440p. Finally, under The Last of US Part II, the numbers soar with 191 and 132 fps.

GPU Compute Performance

Although targeting gamers, the Radeon RX 9060 XT 16 GB can be useful in certain contexts such as video creation, photo editing, and 3D rendering. Many applications are able to exploit its GPU to benefit from hardware acceleration with time and productivity gains as a bonus.

This manufacturer dominates with its RTX 5060 Ti 16 GB and RTX 4060 Ti. The Radeon RX 9060 XT takes third place with a marked difference compared to its elder the RX 7600 XT (+11%)

Here is the detail of some benchmarks that we used.

Blender

LuxMark

FAHBench

AI Text Generation

Using the Procyon AI Test Generation benchmark, we evaluated performance in a context of calculations around AI, particularly with several local LLM models.

The RTX 5070 Ti thus dominates as well as the RTX 4060 Ti RTX 5070 and the RTX 5080. With a relative performance index of 42 (100 for the RTX 5060 Ti 16 GB), the Radeon RX 9060 XT is positioned behind the RTX 4060 8 GB.

Conclusion

Without making a splash, the Radeon RX 9060 XT 16 GB Reaper from PowerColor proves to be solid, balanced, and well-designed for those aiming for comfortable gaming at 1080p or even 1440p, provided that they remain measured on Ray Tracing. However, FSR and Frame Generation allow it to catch its breath where it matters.

Compact, quiet, and energy-efficient, it focuses on efficiency rather than showmanship. Its cooling is of good quality, its noise levels are contained, and its consumption is controlled. It remains behind the GeForce RTX 5060 Ti 16 GB in Rasterization and especially in Ray Tracing, but its price plays in its favor.

For those looking for an efficient, well-cooled, and discreet card, this Radeon RX 9060 XT 16 GB Reaper deserves attention, especially if its price aligns with the MSRP, then putting Nvidia under pressure. In France, AMD announces the Radeon RX 9060 XT 16 GB at $422.89 (€369.90) and the 8 GB edition at $365.72 (€319.90), while Nvidia positions its RTX 5060 Ti 16 GB at $513.31 (€449) compared to $456.15 (€399) for the 8 GB model.