NVIDIA Tesla K20 - 5 GB GPU Computing Accelerator Processing Unit Active Cooling Part Number: 900-22081-2220-000
The NVIDIA Tesla K20s graphic card was released on February 2013. This video card features a graphic processing unit stacked with 2496 cuda cores running at a clock speed of 706 MHz. The Tesla K20s comes with 5120 MB of GDDR5 memory. The memory transfer rate of this video card is 208 GB/s.
Performance And Power
The Tesla K20s graphic card can deliver a single precision computing power of 3524.35 GFLOPS. Furthermore, this video card can achieve a pixel fill rate of 28.24 Gigapixels/s and a texture fill rate of 146.85 Gigatexels/s. The maximum power consumption of the Tesla K20sgraphic card is 225 W.
The Tesla K20s graphic card can be plugged to a motherboard through the support of a PCI-E 2.0 x 16 bus. .
As per the software support, it can support the following programming interfaces and libraries: " DirectX 11.0, OpenCL 1.1, Shader Model 5.0 ". OpenGL is a cross language and multi-platform application programming interface for rendering 2D and 3D vector graphics. OpenGL is used to interact with a GPU to harness Hardware-accelerated rendering and eliminates the need to recompile shaders making it an efficient and performance oriented standard. Microsoft DirectX is an Application programming Interface for the Windows operating systems which Integrates well with AMD and NVIDIA GPUs. The Microsoft DirectX 10 is a significant step in 3D graphics. It features a highly optimized runtime, powerfull geometry sharders, and many other features. The Microsoft DirectX 11, introduced with Windows 7, brings numerous improvement over its precedessor DirectX 10. It is designed to be more efficient, supports sophisticated shading and texturing techniques and levearage the use of multi processors, resulting in smoother 3D animations. The open computing language (OpenCL) is programming interface used for general purpose programming across CPUs and GPUs and other processors. The standard gives developers the portability and efficiency of heterogeneous processing platforms.