Home / AMD / AMD Radeon 8065S Graphics: Performance and Specs

AMD Radeon 8065S Graphics

AMD Radeon 8065S Graphics
AMD Radeon 8065S Graphics graphics card review

AMD Radeon 8065S Graphics: The High-End iGPU for Ryzen AI Max PRO for Local AI

AMD Radeon 8065S Graphics - the high-end integrated graphics for the Ryzen AI Max PRO 400 platform. However, in terms of the GPU itself, this is not a significant leap compared to the Radeon 8060S. The graphics unit remains similar: 40 compute units (CUs) based on the RDNA 3.5 architecture, but the frequency has increased to 3000 MHz. The main difference lies not in the additional 100 MHz, but in the platform surrounding it.

The Radeon 8065S appears in the Ryzen AI Max+ PRO 495 - an updated version of AMD's large APU for local AI, compact workstations, and professional systems. Here, AMD shifts the focus even more from gaming to AI: up to 192 GB of unified memory, up to 160 GB of total memory can be used as GPU memory, the NPU delivers up to 55 TOPS, and the overall AI ceiling of the platform reaches 131 TOPS.

This is no longer just a gaming story. The Radeon 8065S is significant as part of a platform where AMD is attempting to expand its territory in local AI: providing a compact device with a large shared memory pool, a powerful CPU, substantial RDNA graphics, and an NPU without a separate discrete graphics card.

Why the Radeon 8065S Matters

The Radeon 8065S does not have its own GDDR6 memory like a discrete graphics card. It operates with the shared memory of the entire platform. For a typical gaming GPU, this might seem like a limitation, but for local AI, the situation is more complex: often, not only speed matters, but also the volume of available memory.

If a model or context does not fit into the available video memory, high GPU speed no longer helps: the task either fails to start properly or requires significant compromises. The Ryzen AI Max PRO 400 aims to fill this gap. The platform provides the client device with a large local memory pool for LLM, image generation, long contexts, and multiple AI tasks simultaneously.

AMD claims the ability to locally run models with over 300 billion parameters with 4-bit quantization. This does not mean that such a computer replaces a server with professional accelerators. But for a laptop, mini-PC, or compact workstation, this is a significant proposition: large models can be discussed theoretically, but now can actually be run locally with the appropriate software stack.

8065S vs. 8060S: The Key Difference Is in the Platform

The Radeon 8065S is a close relative of the Radeon 8060S. The number of CUs remains the same, the architecture is the same, and the increase in frequency is minor. Therefore, the transition from 8060S to 8065S does not promise a significant increase in FPS on its own.

Parameter Radeon 8060S Radeon 8065S
Architecture RDNA 3.5 RDNA 3.5
Graphic Units 40 CU 40 CU
GPU Frequency up to 2900 MHz up to 3000 MHz
Platform Ryzen AI Max 300 Ryzen AI Max PRO 400
Maximum Unified Memory up to 128 GB up to 192 GB
Memory Available to GPU up to 112 GB up to 160 GB

Comparing solely by frequency overlooks the main difference - the platform's memory limit. For gaming, the difference between 8060S and 8065S will likely be moderate. For local AI, the increase in unified memory from 128 to 192 GB is more important than the slight increase in GPU frequency.

Expected Gaming Performance

Currently, there is limited independent data on the Radeon 8065S, so it is logical to gauge gaming performance through the Radeon 8060S and the small difference in frequency. In configuration, these are very close GPUs: the 8065S has the same wide 40-CU block but with slightly higher frequency. Therefore, the gaming gain relative to the 8060S will likely be small: 100 MHz extra does not elevate this graphics unit to a new class.

The practical scenario remains as follows:

  • 1080p - the primary mode, often with medium or high settings;
  • 1440p - possible in less demanding games or with FSR;
  • 4K - more suitable for older and lighter projects;
  • Ray Tracing - supported but not a strong suit of this iGPU.

The Radeon 8065S should not be marketed as a direct replacement for mobile RTX cards. This is integrated graphics with shared memory, and its performance will depend on power limits, cooling, and the specific device. In gaming, the Radeon 8065S remains an unusually strong iGPU, while in AI, the main advantage comes not from frequency but from access to a large amount of shared memory.

AI: The Main Scenario for Ryzen AI Max PRO 400

The Ryzen AI Max PRO 400 is significantly more oriented towards local AI than serving as a typical gaming platform. The CPU, GPU, NPU, and unified memory operate as parts of a single APU platform rather than as separate components with different memory pools.

The Radeon 8065S may be interesting for tasks such as:

  • Running local LLMs, especially if models do not fit into 8-12 GB VRAM;
  • Inference and testing of AI pipelines;
  • Working with long contexts;
  • Image generation and diffusion models;
  • Local AI agents and automated scenarios;
  • Tools like PyTorch/ROCm, ONNX, Ollama, llama.cpp, Amuse, and others - provided specific configuration support.

Training large models from scratch remains a task for server accelerators. But launching, testing, local development, small fine-tuning, and working with models that do not fit into a typical mobile graphics card is precisely where the Ryzen AI Max+ PRO 495 and the Radeon 8065S appear particularly interesting.

ROCm, PyTorch, and Limitations

The strong point of this platform is not its compatibility with CUDA but the large unified memory pool and the scenarios supported by the AMD stack. However, it is essential not to overstate this. AMD is still catching up to NVIDIA in the AI ecosystem, and compatibility must be checked against specific OS, ROCm version, PyTorch, model, and tool.

In one scenario, the Radeon 8065S can be used effectively as an accelerator, while in another, the software may not utilize it as a GPU or may require workarounds. Therefore, it is better to describe the 8065S not as a universal AI accelerator for any software but as part of a new AMD platform for local AI, where the significant advantage is the very large amount of available memory.

CUDA-dependent software remains a non-target scenario for the Radeon. If the program requires CUDA, an NVIDIA graphics card is needed. For the Radeon 8065S, the point is different: to run local models through tools supported by the AMD stack and handle tasks where memory is more critical than pure compatibility with CUDA.

Where the Radeon 8065S Fits in the Lineup

The Radeon 8065S currently appears as the high-end model in the Radeon 8000S family. Below it are the Radeon 8060S, Radeon 8050S, and Radeon 8040S. The difference between the 8065S and the 8060S is minor regarding the graphics unit but significant concerning the platform context.

The Radeon 8060S was the high-end iGPU for the Ryzen AI Max 300. The Radeon 8065S has become the updated flagship for the Ryzen AI Max PRO 400. For gaming, these are almost in the same class, while for AI, the 8065S shines primarily due to the Ryzen AI Max+ PRO 495 and its extended memory limit.

The Main Downsides - Price and Niche

The Radeon 8065S cannot be purchased separately. It is part of the expensive professional platform Ryzen AI Max+ PRO 495. Therefore, its value depends not on the line “8065S Graphics” itself but on the price of the entire device and whether local AI scenarios are needed by the user.

If a simple gaming laptop is needed, a model with a discrete RTX 4060 or RTX 4070 may be a more logical choice: separate video memory, DLSS, a familiar gaming ecosystem, and CUDA for compatible software. But if a compact workstation with enormous unified memory, a strong CPU, integrated graphics, and the ability to run large models locally is required, the Radeon 8065S becomes much more interesting.

This is not a mass-market solution for typical gaming laptops. It is a high-end iGPU from AMD's niche platform for local AI, workstations, and compact systems where large memory is just as important as raw graphical power.

Conclusion

AMD Radeon 8065S Graphics is not a revolution compared to the Radeon 8060S in pure GPU terms. It is a newer and expanded version of the same idea: 40-CU RDNA 3.5 graphics within a large APU platform, where the main bet is placed on local AI and unified memory.

For gaming, the Radeon 8065S remains very strong integrated graphics, but it does not replace discrete graphics cards. For AI, it is more interesting: not because AMD has caught up to NVIDIA in terms of software ecosystem, but because the Ryzen AI Max+ PRO 495 provides what is rarely found in mobile systems - up to 192 GB of unified memory.

The significance of the Radeon 8065S is not due to it being noticeably faster than the 8060S in games but rather because it solidifies a new meaning for the Ryzen AI Max: it is no longer just a powerful APU, but a client AI platform where memory for local models becomes the main argument.

Basic

Label Name
AMD
Platform
Integrated
Launch Date
May 2026
Model Name
AMD Radeon 8065S Graphics
Generation
Radeon 8000S
Boost Clock
3000 MHz
Bus Interface
Integrated
RT Cores
40
Compute Units
40
Tensor Cores
?
Tensor Cores are specialized processing units designed specifically for deep learning, providing higher training and inference performance compared to FP32 training. They enable rapid computations in areas such as computer vision, natural language processing, speech recognition, text-to-speech conversion, and personalized recommendations. The two most notable applications of Tensor Cores are DLSS (Deep Learning Super Sampling) and AI Denoiser for noise reduction.
No
TMUs
?
Texture Mapping Units (TMUs) serve as components of the GPU, which are capable of rotating, scaling, and distorting binary images, and then placing them as textures onto any plane of a given 3D model. This process is called texture mapping.
160
Foundry
TSMC
Process Size
4 nm
Architecture
RDNA 3.5

Memory Specifications

Memory Size
System Shared
Memory Type
System Shared LPDDR5x
Memory Bus
?
The memory bus width refers to the number of bits of data that the video memory can transfer within a single clock cycle. The larger the bus width, the greater the amount of data that can be transmitted instantaneously, making it one of the crucial parameters of video memory. The memory bandwidth is calculated as: Memory Bandwidth = Memory Frequency x Memory Bus Width / 8. Therefore, when the memory frequencies are similar, the memory bus width will determine the size of the memory bandwidth.
256-bit
Memory Clock
LPDDR5x-8533
Bandwidth
?
Memory bandwidth refers to the data transfer rate between the graphics chip and the video memory. It is measured in bytes per second, and the formula to calculate it is: memory bandwidth = working frequency × memory bus width / 8 bits.
273 GB/s

Theoretical Performance

Pixel Rate
?
Pixel fill rate refers to the number of pixels a graphics processing unit (GPU) can render per second, measured in MPixels/s (million pixels per second) or GPixels/s (billion pixels per second). It is the most commonly used metric to evaluate the pixel processing performance of a graphics card.
192 GPixel/s
Texture Rate
?
Texture fill rate refers to the number of texture map elements (texels) that a GPU can map to pixels in a single second.
480 GTexel/s
FP16 (half)
?
An important metric for measuring GPU performance is floating-point computing capability. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable. Single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks, while double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy.
30.72 TFLOPS
FP64 (double)
?
An important metric for measuring GPU performance is floating-point computing capability. Double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy, while single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable.
480 GFLOPS
FP32 (float)
?
An important metric for measuring GPU performance is floating-point computing capability. Single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks, while double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable.
15.36 TFLOPS

Miscellaneous

Shading Units
?
The most fundamental processing unit is the Streaming Processor (SP), where specific instructions and tasks are executed. GPUs perform parallel computing, which means multiple SPs work simultaneously to process tasks.
2560
OpenCL Version
2.1
OpenGL
4.6
DirectX
12
CUDA
No
Power Connectors
None
Shader Model
6.8
ROPs
?
The Raster Operations Pipeline (ROPs) is primarily responsible for handling lighting and reflection calculations in games, as well as managing effects like anti-aliasing (AA), high resolution, smoke, and fire. The more demanding the anti-aliasing and lighting effects in a game, the higher the performance requirements for the ROPs; otherwise, it may result in a sharp drop in frame rate.
64

Benchmarks

FP32 (float)
Score
15.36 TFLOPS

Compared to Other GPU

FP32 (float) / TFLOPS