AMD Radeon 680M

The AMD Radeon 680M, an integrated graphics processor built on the RDNA 2.0 architecture, occupies a unique position in the benchmark hierarchy. With an average benchmark score of 16,407, it lands in the 58th percentile of all GPUs, placing it squarely in the mid-range of performance among both dedicated and integrated parts. The data indicates a tightly contested field at this performance tier, with the 680M essentially trading blows with several professional workstation GPUs, making it a remarkably capable integrated solution.

Power and Cooling

The Radeon 680M is defined by its extraordinarily modest power envelope, with a Thermal Design Power (TDP) of just 50 W. This figure is the cornerstone of its design, as it allows the GPU to function within the thermal constraints of a mobile or compact system without the need for active cooling solutions dedicated solely to the graphics processor. Because of this low power draw, the data shows that the chip requires no external power connectors, listing "None" as its power connector requirement, and the slot width is classified as "IGP," confirming its integrated nature.

Consequently, the FACT PACK data provides no Suggested PSU recommendation, as the power supply requirements are entirely dependent on the host system's CPU and other components rather than the GPU itself. The 50 W TDP is a critical advantage for system integrators and users looking to build ultra-portable or low-noise machines, as it eliminates the complex power delivery and thermal management challenges associated with discrete graphics cards. The 680M confirms efficiency, delivering performance that rivals discrete options while drawing a fraction of the power, making it an ideal candidate for thin-and-light laptops or mini-PCs where heat dissipation is a primary constraint.

Who Should Consider It

Benchmark results indicate that the Radeon 680M is a versatile performer, but its suitability hinges on the user's resolution and graphical expectations. For gamers targeting 1080p resolution, the 680M provides a viable path to playable frame rates in less demanding esports titles and older AAA games, where the 3.379 TFLOPS of FP32 compute power can be effectively utilized. Its performance, however, is not aligned with high-refresh-rate or ultra-settings gaming at that resolution; users should expect to moderate graphical settings to achieve smooth gameplay.

At 1440p or higher resolutions, the integrated nature of the memory subsystem becomes a significant limiting factor, and the 680M is best suited for productivity tasks, media consumption, and light creative work rather than intensive gaming. The data suggests that users who primarily engage in office applications, video playback, and photo editing will find the GPU more than adequate, as these workloads benefit from the 768 shading units and 48 texture mapping units without stressing the memory bandwidth. In contrast, users seeking a high-end gaming experience or those working with complex 3D rendering should look towards discrete solutions, as the 680M's performance profile, while impressive for an IGP, remains below the threshold for demanding visual workloads.

Memory Subsystem

The memory subsystem of the Radeon 680M is unconventional, as it relies entirely on "System Shared" memory. This means the GPU does not possess its own dedicated VRAM; instead, it accesses a portion of the host system's main memory. The memory type, bus width, and bandwidth are all listed as "System Shared" or "System Dependent," indicating that the GPU's memory performance is intrinsically linked to the speed and configuration of the host system's RAM.

This architectural choice has profound implications for high-resolution gaming. While the 768 shading units provide ample compute throughput, the lack of a dedicated, high-bandwidth memory bus means that data transfer between the GPU and system RAM becomes a bottleneck, particularly at higher resolutions where larger textures and frame buffers are required. The performance is directly correlated with the system's memory bandwidth; a system with fast, dual-channel DDR5 memory will yield significantly better GPU performance than one with slower, single-channel memory. Therefore, the "System Dependent" bandwidth figure is not a weakness but a variable, and users must consider their system's memory configuration as a critical component of the 680M's overall performance. This shared architecture is efficient for cost and power, but it places a premium on having a well-balanced system.

How It Compares

The Radeon 680M finds itself in a fascinating competitive position, with its average benchmark score of 16,407 placing it directly alongside several professional-grade discrete GPUs. Against the AMD Radeon PRO W7500, the 680M is statistically tied, with a negligible delta of 0.1% in favor of the 680M. This indicates that the integrated solution matches the performance of a dedicated workstation card, a surprising result that highlights the efficiency of the RDNA 2.0 architecture.

Similarly, the comparison with the AMD Radeon Pro 5700 and the AMD Radeon Pro 5600M shows an identical pattern. The 680M leads both by a margin of 0.3%, a difference that is well within the margin of error for benchmarking. This effectively places the integrated 680M on par with these older, dedicated mobile and workstation GPUs, offering comparable performance without the need for a separate graphics card. The final rival, the NVIDIA T400, presents a slight reversal, with the 680M trailing by 0.5%. This negligible deficit confirms that the 680M is firmly entrenched in a performance class that was previously the exclusive domain of discrete graphics solutions.

Benchmark Performance

The benchmark data reveals a consistent performance profile for the Radeon 680M across different testing methodologies. In the demanding 3DMark Steel Nomad DX12 test, the GPU scores 378 points, a figure that reflects its capabilities in modern, low-level API gaming workloads. While this score is modest in absolute terms, it aligns with the GPU's position as an integrated part, indicating its ability to handle contemporary game engines at acceptable settings.

The compute-oriented benchmarks paint a more impressive picture. In Geekbench OpenCL, the 680M achieves a score of 22,994, demonstrating strong general-purpose compute performance. More notably, its Geekbench Vulkan score of 25,850 is significantly higher, suggesting that the architecture is particularly well-optimized for the Vulkan API, which is increasingly used in modern games and professional applications. The average benchmark score of 16,407, derived from these tests, shows the 680M performing within 0.1% to 0.3% ahead of the AMD Radeon PRO W7500, Pro 5700, and Pro 5600M, while being just 0.5% behind the NVIDIA T400. These minuscule deltas, ranging from -0.5% to +0.3%, indicate that the 680M delivers performance that is virtually indistinguishable from its nearest discrete rivals, despite being an integrated component with a 50 W TDP.

FAQ

Q: Is the AMD Radeon 680M a discrete graphics card?

A: No, the 680M is an integrated graphics processor (IGP), as indicated by its slot width classification, and it is designed to be part of a larger system rather than a standalone expansion card.

Q: What is the power consumption of the Radeon 680M?

A: The 680M has a TDP of 50 W, which is exceptionally low and requires no external power connectors, as it draws power directly from the motherboard.

Q: How much dedicated video memory does the 680M have?

A: The 680M has no dedicated video memory. It uses "System Shared" memory, meaning its VRAM size, type, and bandwidth are dependent on the host system's main memory configuration.

Q: How does the 680M compare to the NVIDIA T400?

A: Benchmark results show the 680M is slightly behind the NVIDIA T400, with an average score difference of -0.5%, indicating nearly identical performance levels.

Q: Does the 680M support hardware ray tracing?

A: Yes, the 680M includes 12 dedicated ray tracing cores, and its DirectX 12 Ultimate (12_2) API support confirms hardware-level ray tracing capabilities.

Q: What is the manufacturing process for the 680M?

A: The 680M is manufactured on a 6 nm process node at TSMC, which contributes to its high transistor density of 63.0 million transistors per square millimeter.

Ray Tracing and Feature Set

The Radeon 680M is equipped with 12 dedicated ray tracing cores, a feature that is notably uncommon in integrated graphics. This hardware support enables the GPU to handle real-time ray tracing effects in compatible games, a capability typically reserved for high-end discrete cards. While the overall performance is limited by its 50 W TDP and shared memory architecture, the presence of these cores means that ray tracing is not entirely off the table for users of this IGP, provided they are willing to accept lower resolutions and settings.

Complementing the ray tracing hardware, the 680M boasts a robust feature set through its API support. It supports DirectX 12 Ultimate (12_2), ensuring compatibility with the latest gaming features, as well as OpenGL 4.6 and Vulkan 1.4 for broad software compatibility. The GPU's architecture is based on RDNA 2.0, which is a proven and efficient design. The 768 shading units deliver a peak FP32 performance of 3.379 TFLOPS, while the 48 texture mapping units and 32 raster operations pipelines provide a texture fill rate of 105.6 GTexel/s and a pixel rate of 70.40 GPixel/s. These specifications, combined with the 12 RT cores, define a feature-rich integrated GPU that offers a taste of modern graphical technologies without the power and thermal overhead of a discrete solution.