AMD Compute Cores
||With HSA, both the CPU and GPU cores are compute cores
- hUMA advances how processors inside an APU access memory
- hQ revolutionizes how processors inside an APU interact with each other to handle computational tasks
At AMD, we recognize that our customers often think of processors (CPUs) and graphics cards (GPUs) in terms of the number of cores that each product has. This approach can help customers compare apples to apples.
However, AMD's accelerated processing units (APUs) combine CPUs and GPUs on a single chip, so each APU has both CPU cores and GPU cores on it.1 The question is how to talk about the number of cores on any given APU.
To answer this question, AMD uses the term "Compute Core." We have established a definition of the term "Compute Core" so that we are taking a consistent and transparent approach to describing the number of cores in our accelerated processors.
A Compute Core is any core capable of running at least one process in its own context and virtual memory space, independently from other cores.
Based on the above definition, a Compute Core can be either a CPU core or GPU core.
To illustrate, we can describe a given APU, such as the AMD A10-7850K APU, as having 12 Compute Cores, consisting of 4 CPU cores and 8 GPU cores.
Beginning in 2014 with the first generation of heterogeneous processors based on the architecture of the APU codenamed “Kaveri”, AMD will begin designating the number of Compute Cores in the following manner.
AMD A10-7850K APU with AMD Radeon™ R7 graphics
- 12 Compute Cores (4 CPU + 8 GPU)
For a deeper dive into Compute Core characteristics, check out our Compute Cores white paper.
- AMD defines a “Radeon Core” as one Shader/Shader Array. The term “GPU Core” is an evolution of the term “Radeon Core”. “GPU Core” is defined as having 4 SIMDS each comprising of 64 Shaders/Shader Arrays. For example, 512 “Radeon Cores” equals 8 “GPU Cores“ (8 GPU Cores x 4 SIMDs x 16 Shader Arrays = 512 Radeon Cores).