AMD Radeon R8 M535DX

The AMD Radeon R8 M535DX is an integrated graphics processor (IGP) from AMD's Gem System Hybrid generation (Rx M500 series), built on the GCN 3.0 architecture. Fabricated on a 28 nm process at TSMC, it packs 1,550 million transistors into a 125 mm² die, yielding a transistor density of 12.4M / mm². The part is marked as end-of-life and was released in April 2017. It operates with a base clock of 780 MHz and a boost clock of 891 MHz, and its memory subsystem is entirely system-shared.

Power and Cooling

The fact pack lists no TDP, no power connectors, and no suggested PSU for this part. Its slot width is IGP, which means it is integrated into the motherboard or system board rather than occupying a discrete expansion slot. The absence of power connector data is consistent with an IGP design, as such parts draw power from the system's existing rails. The 28 nm process and the 1,550 million transistor count on a 125 mm² die indicate a relatively small, power-conscious design, but the lack of a TDP figure means the actual thermal envelope is not specified. The bus interface is also IGP, reinforcing that this is not a card that requires external power or a dedicated cooling solution. Since no cooler dimensions or wattage are provided, the practical cooling requirements are system-dependent, but the IGP form factor suggests reliance on the host system's cooling. The 12.4M / mm² transistor density is a metric of the manufacturing efficiency, but without a TDP, it is not possible to estimate heat output. The data shows no PSU recommendation, so builders must assume the system's existing power delivery is sufficient. This is typical for integrated graphics, where the GPU shares the system's power budget.

Ray Tracing and Feature Set

The fact pack explicitly lists no ray tracing cores and no tensor cores. This means the M535DX lacks dedicated hardware for ray-traced lighting and AI-based tensor operations. API support includes DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170. The DirectX 12 (12_0) support indicates compatibility with modern DirectX 12 titles, but without RT cores, any ray-traced effects would have to be computed via the shading units, which would be extremely slow given the modest FP32 throughput. OpenGL 4.6 and Vulkan 1.2.170 provide cross-platform API coverage, allowing the part to run a wide range of software that supports these APIs. The absence of tensor cores also rules out any hardware-accelerated AI-based upscaling or similar features. The feature set is therefore purely based on the GCN 3.0 architecture's compute and rasterization capabilities. The FP16 rate is listed as 570.2 GFLOPS (1:1) with FP32, meaning there is no half-rate or double-rate FP16 execution, which limits compute-heavy workloads that could benefit from FP16. The shading units (320), TMUs (20), and ROPs (8) define the raw processing elements, but without RT or tensor cores, the feature set is focused on traditional graphics rendering.

Memory Subsystem

The memory size, type, and bus width are all listed as "System Shared". This is a critical specification, as it means the GPU does not have its own dedicated VRAM. Instead, it uses a portion of the system's main memory. The bandwidth is listed as "System Dependent", meaning the effective memory bandwidth is not a fixed number but rather depends on the host system's memory configuration, such as the number of memory channels and the RAM speed. This has profound implications for high-resolution gaming or memory-intensive workloads. Because the GPU shares memory with the CPU, the available bandwidth is limited by the system's memory controller and the speed of the system RAM. At high resolutions, the demand for texture data and frame buffers increases, and a shared memory subsystem can become a severe bottleneck. The lack of a dedicated bus width figure means the GPU's memory interface is not a fixed path; it is whatever the system provides. This makes the M535DX unsuitable for high-resolution or high-detail settings, as the memory latency and bandwidth are not optimized for graphics workloads. The system-dependent bandwidth also means that two systems with the same GPU could exhibit different memory performance if they have different RAM configurations. For a user planning to run modern games at high resolutions, the shared memory subsystem is a major limitation.

How It Compares

The fact pack provides an empty nearestRivals array, meaning there are no direct rival comparisons available in the benchmark database. The percentile ranking versus all GPUs is 50, which places this part exactly at the median of the tracked GPU population. However, this percentile is based on the database's overall distribution, and with no rival scores listed, it is not possible to make specific head-to-head comparisons. The absence of rival data means that any positional claims would be speculative. The only concrete positional indicator is the 50th percentile, which suggests that it sits in the middle of the performance spectrum when considering all GPUs ever tracked. Yet, because the average benchmark score is 0, this percentile is not backed by any measured performance data. Without rival names or delta percentages, the analysis must rely on the absolute specifications. The part's end-of-life status and April 2017 release date place it in an older generation, but the lack of rivals means no historical context is provided. The data shows that this is a standalone entry with no comparative anchors.

Benchmark Performance

The benchmarks array is empty, and the average benchmark score is 0. This indicates that no performance samples have been recorded for this part in the database. Consequently, the analysis must rely on theoretical peak rates derived from the hardware specifications. The pixel rate is 7.128 GPixel/s, which is the maximum number of pixels the ROPs can output per second. The texture rate is 17.82 GTexel/s, representing the texture fill rate. The FP32 performance is 570.2 GFLOPS, and FP16 is also 570.2 GFLOPS (1:1). These numbers are quite modest by modern standards. A pixel rate of 7.128 GPixel/s and a texture rate of 17.82 GTexel/s indicate a very low fill capability, which will bottleneck any resolution above basic desktop output. The FP32 throughput of 570.2 GFLOPS is suitable for very light compute tasks but will struggle with modern game physics or shader complexity. The 1:1 FP16 ratio means there is no advantage for FP16 workloads. The 50th percentile is a database position, but with zero benchmark scores, it is a placeholder rather than a measured result. The data shows no evidence of real-world performance, so any claims about gaming or compute capability are purely theoretical.

Who Should Consider It

Given the system-shared memory, the low pixel and texture rates, and the modest FP32 throughput, the AMD Radeon R8 M535DX is suited for basic desktop tasks such as office productivity, web browsing, and video playback. The shared memory subsystem means that any application requiring high memory bandwidth, such as modern games at high resolutions, will be severely limited. The end-of-life status and April 2017 release date indicate that this is a legacy part, likely found in older laptops or compact systems. Users who need to run games should consider that the 7.128 GPixel/s pixel rate and 17.82 GTexel/s texture rate are far below what is needed for smooth gameplay at any reasonable resolution. The 570.2 GFLOPS FP32 performance is adequate for 2D acceleration and basic 3D rendering but not for modern titles. The absence of ray tracing and tensor cores further limits its appeal for any advanced graphics features. In summary, this is a part for basic computing needs, not for gaming or content creation. The 50th percentile ranking is misleading because it is not backed by any benchmark scores, so the actual performance is likely lower than the median of all GPUs. High-resolution workloads are not advisable, and the system-dependent bandwidth means that even light tasks could be hampered if the system memory is slow. For anyone considering this part, the data suggests it is only suitable for legacy or low-demand scenarios.