AMD Lights Up Broadway with Times Square’s Largest UHD Display

The graphic display moving across the 25,000 square feet of Ultra-High Definition digital signage at 1535 Broadway on New York City’s Times Square, incorporates nearly 24 million pixels, powered by AMD FirePro professional graphics cards.

The sign, with a playback system designed and managed byDiversified Media Group, is a single surface covering a city block in length and stands eight stories high. Driving the visual display are three AMD FirePro professional graphics cards, each powering six sections of the display for a combined resolution of 10,048 x 2,368 pixels.

Optimisation and Multitasking

Apart from high quality 2D and 3D graphics and video, multiple-display walls require high densities and a long life cycle. Many applications are now using the computer’s GPU not only to accelerate 3D graphics, but also to off-load computational-intense tasks such as rendering and simulations from the CPU, usingOpenCL.

The most recent generation of AMD FirePro GPUs withGraphics Core Nextarchitecture [see below] has been designed to efficiently balance compute tasks with 3D workloads. This enables multitasking from a single card, optimizing utilization and maximizing performance. Specifically, the cards are optimized for workstation applications supporting high-resolution geometry performance and smooth handling of complex designs and effects.

Each of the three graphics cards at the Times Square display is able to output high-quality visuals on three, four or six displays, from a single workstation or PC, through the use ofAMD Eyefinityprocessing. The Eyefinity system reduces overall system complexity, mainly due to Display Output Post-Processing, an AMD OpenGL extension that lets users of the cards control and manipulate the desktop image directly as a texture, and renders it through the OS in the usual way before outputting to a display.

Zone Synchronization

AMD FirePro graphics cards supportDisplay Port 1.2as well, which enables graphics and video of a high enough quality for applications that require UHD resolutions in 4K, 4096 x 2160 pixels, and an extended 10-bit colour spectrum. Low-latency, high-throughput links are also established between the various third-party SDI I/O components of video processing systems. Multiple graphics cards can then be combined and synchronized to create massive display walls with tens and sometimes hundreds of screens.

The individual display sections of the wall are synchronized across graphics cards and zones using theATI FirePro S400 synchronizationmodule, made for applications that require synchronization to external sources, genlock, or synchronization of 3D rendering of multiple GPUs in different systems, framelock. The combination of these capabilities with AMD FirePro graphics accelerators extends the use of the cards to more types of applications.

Based on a full hardware design with a dedicated processor, the ATI FirePro S400 can result in extremely accurate synchronization in most situations. The dedicated processor allows up to four attached GPUs to carry out the required rendering, without affecting sync.

Graphics Core Next Architecture

AMD FirePro graphics andGraphics Core Next, or GCN, architecture meet diverse software demands by executing a very high number of operations in parallel within a single computer clock cycle. Those operations, applied to billions of polygons every second, are what make a display wall or animation look real to the human eye.

GCN GPUs are built to handle workloads and programming languages normally confined to the main processor. They are also a good match for newer GPU-aware languages like C++ AMP and OpenCL.

Increasing the number of transistors in a GPU has a major impact on the potential performance of a graphics card, but the transistors also need an effective design to turn their potential into performance combined with image quality, that is, the clarity and accuracy of textures and effects. Therefore, GCN architecture employs three specialized techniques for this.

Realistic Textures

One deals with the fact that increasing the physical size or number of textures can have a negative impact on the performance of a GPU.Partially Resident Texturesor PRT have been developed to utilize massive texture files of up to 32 TB, with minimal performance impact by streaming small bits of textures into the GPU as needed.

Available on all modern GPUs,anisotropic filteringis another technique that assists the GPU in making sure textures remain sharp at different distances. However, GPU designs differ in the way the anisotropic filtering is actually executed, which the GCN architecture has been specifically optimized to handle.

Tessellationis a rendering effect that can dynamically generate higher detail in a scene on the fly to increase realism. Tessellation is not new to GPUs, but the manner in which it is executed can have a large impact on images. Because of this, the GCN Architecture has again been optimized to result in a measureable increase in tessellation performance. www.amd.com