Everyone knows that a high-end computer is usually a gaming computer. To run the best and newest games, you need the latest Intel i7 processor and a monster of an Nvidia or ATI graphic card. However, while unknown to most gamers, today’s video cards can be technically tweaked to provide performance that is at least 10x better than a CPU. Enter the era of hybrid CPU-GPU Computing.
Most gamers have probably never noticed the speed and complexity inside video card technology. The RAM chips used in video cards are usually a step more advanced than the current DDR2/DDR3 technology. The processor component inside the video cards is literally a hundred times more in number compared to the high-end quad core CPUs of today.
The picture above is an analogy of a quad-core CPU in comparison with a hundred-core GPU.
A quad core CPU has 4 “solid” cores of powerful engines made best for handling operating systems, multitasking software, and generic computations. However, today’s GPU inside video cards – contain hundreds of smaller processors. They may be less powerful in capacity, but their sheer number and speed will literally outgun a quad core in computing speed.
Nvidia has released its “Tesla” series of high performance GPUs.
These Tesla GPU’s were designed to accept special programming instructions that simply outpace the equivalent and high-end GeForce gaming cards. Nvidia introduced the “CUDA” technology. This is basically a programming foundation that software manufacturers can use to harness the power of the high performance GPU cores, and combine the power of both CPU and GPU.
Hybrid Computing is a CPU working together with a GPU to quickly process complex mathematical functions. Through this technology, tasks that require the management and multitasking is handled by the CPU. Computations that are mathematically similar but quite lengthy to process is broken to hundreds of pieces and are assigned to the GPU. The GPU’s output is then combined and organized by the CPU.
Through hybrid computing, important scientific video processing needs are accomplished X times faster. In the picture below, Hybrid Computing allows for 5x faster digital content creation in Adobe products (imagine, Adobe Photoshop to be 5x faster!). Weather modeling that is usually done by multiple connected computers, can be done 80x faster with GPU technology. Even more surprising and usually our misconception – is that number crunching analysis is accomplished much much faster with a GPU rather than a CPU.
More applications are soon to be supporting the technology of Hybrid Computing. Programmers will soon be able to harness the combined power of multi-core CPU and GPU processors. 3D games and applications are starting to switch to the CUDA technology. Its only a matter of time before consumer games, applications, and drivers are released to take advantage of the CUDA enabled technology of Nvidia GPUs.
The link below is a list of GeForce GPUs that have CUDA capability. Upcoming games (or future game patches and updates) can definitely take advantage of Hybrid Computing. CUDA GPUs