AMD Radeon RX Vega 3
2 laptops with AMD Radeon RX Vega 3 in comparion
2 laptops with AMD Radeon RX Vega 3 in comparion
The AMD Radeon Vega 3 is an integrated GPU (iGPU) that was presented at the CES in Las Vegas on January 8, 2018. It is the smallest of the integrated Vega-based graphics solutions that are installed in the accelerated processing units (APUs) of the Raven Ridge series. The iGPU is used in entry-level APUs such as the Ryzen 3 2200U and replaces the low-end iGPU Radeon R5 (Bristol Ridge). Not much is known about the integrated Vega GPUs. As with the Vega 10 and Vega 8 models presented in October 2017, AMD only provides information about the chip configuration and the maximum clock frequency. The name tells us that the AMD Radeon Vega 3 has 3 compute units (CUs) with a total of 192 shader units and 12 texture mapping units (TMUs). According to AMD, the graphics unit can reach a maximum clock frequency of up to 1000 MHz, but the base clock is not specified. As there is no dedicated memory, the GPU must access the system memory. Dual-channel DDR4 memory with up to 2400 MHz is supported.
Architecture
For years, AMD has only used the term Graphics Core Next (GCN) as the underlying graphics architecture for graphics cards. Regardless of the version of the architecture, which was constantly developed further, all GPUs were always "GCN-based". With the introduction of the Polaris graphics cards in 2016, there are signs of a change here. Although the new GPUs are still GCN-based, AMD now uses the chip name almost exclusively. In the case of the 5th generation of GCN architecture, this name is "Vega". Vega introduces a number of improvements that help to increase performance and reduce power consumption. A far-reaching modification of the shader units allows the use of Double Rate FP16 Math (Rapid Packed Math) technology for 8-bit, 16-bit and 32-bit numbers. This technology significantly improves the performance of floating point calculations with half precision (FP16). The new "Primitive Shaders" stage in the geometry pipeline will accelerate triangle culling, i.e. the removal of superfluous triangles from the pipeline at an early stage. If the primitive shaders are adopted by developers in the program code, the throughput of the pipeline can be doubled. Vega GPUs can now also perform tile-based rendering with the Draw Stream Binning Rasterizer. This technology, which can increase performance and energy efficiency in some instances by removing invisible pixels, is already known from mobile SoCs and was already used in NVIDIA's Maxwell architecture.
Features
The AMD Radeon Vega 3 supports the APIs Direct3D 12 Feature Level 12_1, OpenCL 2.2, OpenGL 4.6 and Vulkan 1.0. Further features are not known at the time of the product launch.
Performance
The AMD Radeon Vega 3 is by far the weakest model of the integrated Radeon Vega GPUs. Although it should deliver more performance than the Radeon R5 (Bristol Ridge) with 4 compute units, it is still only roughly on a par with the Radeon R7 (Bristol Ridge) with 6 compute units. Only less demanding games such as Sims 4 and Dirt Rally can therefore be played at a low resolution.
DirectX version | 12 |
Chip frequency | 1000 MHz |
Dedicated memory | No |
Pipelines | 192 |
Texture Mapping Units | 12 |
Production process | 14 nm |