ARM Mali-G72 MP12 vs Qualcomm Adreno 642 vs ARM Mali-G78 MP22
ARM Mali-G72 MP12
► remove from comparisonThe ARM Mali-G72 MP12 is an integrated high-end graphics card for ARM based SoCs (mostly Android based). It was introduced in October 2017 in the Huawei Mate 10 and uses 12 clusters (hence the MP12 name). In our benchmarks it performs between a Adreno 530 and 540 and is therefore suitable for demanding mobile games.
The G72 is based on the second generation of the Bifrost architecture and offers improvements in the machine learning efficiency and a bigger tile buffer for 16x anti-aliasing.
The GPU supports all modern graphics APIs like OpenGL ES 3.2, Vulkan 1.0, OpenCL 2.0, DirectX 12 FL11_1 and Renderscript.
Qualcomm Adreno 642
► remove from comparisonThe Qualcomm Adreno 642 is a smartphone and tablet GPU that is integrated within the Qualcomm Snapdragon 780G SoC. The chip will be available from mid 2021 and will be used mainly in upper mid-range Android devices.
According to Qualcomm, the Adreno 642 GPU offers a 50% improved performance over the Adreno 620, its predecessor, which is integrated in the Snapdragon 768G SoC. This is also thanks to the fast LPDDR4X-4200 memory support of the SoC.
The Adreno 642 supports OpenCL 2.0 FP, OpenGL ES 3.2 and Vulkan 1.1.
The Snapdragon 780G is manufactured in the modern 5nm LPPE process with EUV at Samsung that should provide a good power efficiency.
ARM Mali-G78 MP22
► remove from comparisonThe ARM Mali-G78MP22 is an integrated high-end graphics card for ARM based SoCs (mostly Android based). It was introduced late 2020 in the HiSilicon Kirin 9000E (e.g. Huawei MatePad Pro 12.6). It integrates 22 of the 24 possible cores and is based on the second generation of the Valhal architecture. According to ARM it offers two new features: asynchronous top level and fragment dependency tracking.
The graphics card is one of the fastest in Android devices of 2020 and therefore able to run all games fluently.
The GPU supports all modern graphics APIs like OpenGL ES 3.2, Vulkan 1.1 and OpenCL 2.0.
|
|