What is A Graphics Processing Unit?

A graphics processing unit (GPU) is a specialized electronic circuit made to rapidly manipulate and alter memory to accelerate the creation of images in a frame buffer designed for output to a display device. GPUs are found in embedded systems, cell phones, computers, workstations, and game consoles. Modern GPUs have become efficient at manipulating computer graphics and image processing. Their highly parallel structure makes them better than general-purpose central processing units (CPUs) for algorithms that process large blocks of data in parallel. In an individual computer, a GPU could be present on a video card or embedded on the motherboard. Using CPUs, they are embedded on the CPU die.

Source: https://gpuhub.net/

The word "GPU" was coined by Sony in mention of the PlayStation console's Toshiba-designed Sony GPU in 1994. The word was popularized by Nvidia in 1999, who marketed the GeForce 256 as "the world's 1st GPU". It had been presented as a "single-chip processor with integrated transform, lighting, triangle setup/clipping, and rendering engines". Rival ATI Technologies coined the word "visual processing unit" or VPU with the release of the Radeon 9700 in 2002.

GPU companies

Many companies have produced GPUs under several brand names. In '09 2009, Intel, Nvidia and AMD/ATI had been the marketplace share leaders, with 49.4%, 27.8% and 20.6% marketplace share respectively. However, those numbers include Intel's integrated graphics solutions as GPUs. Not counting those, Nvidia and AMD control nearly 100% of the marketplace by 2018. Their respective marketplace shares are 66% and 33%. Furthermore, S3 Graphics and Matrox produce GPUs. Modern smartphones also use mostly Adreno GPUs from Qualcomm, PowerVR GPUs from Imagination Technologies and Mali GPUs from ARM.

Computational functions

Modern GPUs use the majority of their transistors to accomplish calculations linked to 3D computer graphics. As well as the 3D hardware, today's GPUs include basic 2D acceleration and framebuffer capabilities (usually with a VGA compatibility mode). Newer cards such as for example AMD/ATI HD5000-HD7000 even lack 2D acceleration; it must be emulated by 3D hardware. GPUs were at first used to accelerate the memory-intensive work of texture mapping and rendering polygons, later adding units to accelerate geometric calculations like the rotation and translation of vertices into different coordinate systems. Recent developments in GPUs consist of support for programmable shaders that may manipulate vertices and textures with lots of the same operations supported by CPUs, oversampling and interpolation ways to reduce aliasing, and incredibly high-precision color spaces. Because many of these computations involve matrix and vector operations, engineers and scientists possess increasingly studied the usage of GPUs for non-graphical calculations; they are specially suitable for other embarrassingly parallel problems.

With the emergence of deep learning, the need for GPUs has increased. In research done by Indigo, it had been discovered that while training deep learning neural networks, GPUs could be 250 times faster than CPUs. The explosive growth of Deep Learning recently has been related to the emergence of general purpose GPUs. There's been some degree of competition in this area with ASICs, most prominently the Tensor Processing Unit (TPU) created by Google. However, ASICs require changes to existing code and GPUs remain very popular.