Shared memory

As a Unified Memory Architecture (UMA ) is called computer architectures in which different components ( eg, CPU and graphics processor) to work together on a shared memory instead of having dedicated memory.

X86 computer systems

The more general term is at x86 Memory Mapped Graphics ( MMG). This was in the early days of personal computers represent the usual way of graphics. Another name for the principle is the shared memory.

Due to the omission of a dedicated memory, a more cost-effective production of the system is possible. Shared access to the main memory will result in performance degradation for the system: In the times in which the graphics processor ( GPU) accesses the memory, the CPU is not available and the CPU must wait until idle. In 3D mode, this required by the GPU data transfer rate increased by a multiple texture by complex calculations. In addition, the used by the GPU memory area can be (usually 128-256 MB) permanently not used by the CPU and decimated by the available memory. The graphics performance is also lower, since the access to main memory is slower than dedicated graphics memory.

SGI / MIPS systems

SGI put the mid-1990s workstations on the market, in which the CPU and graphics processor on a single high-performance memory worked. The memory 256 was organized bit wide, which 3.2 GB / s at 100 MHz revealed. The chipset used was cobalt, the main benefit was a flexible division between main memory and graphics memory and the elimination of the PCI bus as a bottleneck.

UMA for game consoles

The products developed by Microsoft Xbox and Xbox 360 game consoles also use UMA. So these do not have a dedicated graphics memory, but a common memory for CPU and GPU.

Demarcation

Is it at the access to the shared memory to the parallel use of content within a symmetric multiprocessor system ( SMP), one speaks of uniform memory access.

• Computer Architecture