Ensemble 6000 Series Dual AMD GPGPU Processing OpenVPX GSC6201 6U GPGPU Module
5Pages

Ensemble 6000 Series Dual AMD GPGPU Processing OpenVPX GSC6201 6U GPGPU Module Over 4 TFLOPS of high-performance stream processing in a single 6U OpenVPX slot for rugged environments • Industry’s highest performance using latest generation AMD Radeon™ HD 6970M and HD 7970M GPGPU • Mercury’s StreamDirect™ for GPGPUs increases overall system efficiency and intelligence gathering • Industry standard MXM form factor ensures rapid access to latest generation GPGPUs, including over 4 TFLOPS in a single 6U slot with the AMD Radeon HD 7970M • Open standards OpenCL™ and Mercury’s Scientific Algorithm Library (SAL™) for optimized, platform-independent C/C++ application development • Rugged conduction-cooled module exceeds strict VITA 47 standards for shock and vibration to operate in the harshest environments The GSC6201 is a 6U OpenVPX™ carrier module that integrates two high-performance AMD GPGPUs for applications that can benefit from massively parallel processing on streams of high-bandwidth data. GPGPUs offer some of the highest GFLOP processing performance as well as the highest GFLOP/Watt performance efficiency in the industry. Meeting Today’s Processing Demands Commercial and defense applications such as radar, electro-optical/infrared (EO/IR), electronic warfare and cyber generate large amounts of raw sensor or network data that need to be processed in real-time to extract actionable intelligence. Each new generation of sensor arrays ushers in higher resolutions and frame rates; along the same lines, network traffic is also increasing exponentially. Consequently, the input levels from both of these data streams are now reaching “Big Data” proportions. By offloading compute-intensive operations to GPGPUs such as Fast Fourier Transforms (FFTs), matrix multiplication, Constant False Alarm Rate (CFAR), QR Decomposition (QRD), Synthetic Aperture Radar (SAR), video codecs (H.264, JPEG2000), pattern recognition or deep packet inspection, system architects can engineer solutions that can meet today’s processing demands — with room to scale for higher performance requirements in the future while preserving significant IP investment. StreamDirect for GPGPUs: Faster Intelligence Mercury’s StreamDirect writes data directly from the sensor straight into GPGPU memory, reducing data latency by 50 percent, thereby allowing the raw data to be processed immediately for analysis and intelligence. It does this by bypassing the CPU that previously served as traffic manager. Because the CPU is no longer the bottleneck, additional sensor/GPGPU processing slices can be added to the system to significantly increase processing capability while optimizing for SWaP (size, weight and power). The GSC6201 is Mercury’s 4th generation 6U dual GPGPU MXM module, backed with a history of field-proven deployments in a number of defense programs since 2008. Updates to the new GSC6201 include higher I/O bandwidth, StreamDirect for GPGPUs and support for both AMD Radeon HD 6970M and HD 7970M. Figure 1. StreamDirect reduces the time from raw data to intelligence by 50 percent by writing sensor data directly into the GPU’s memory for immediate processing

Previously, systems that used GPGPUs relied on host CPU memory copies to transfer data from the sensor to GPU. Incoming streams of sensor data had to first be written to the host CPU memory allocated for the sensor, and then copied to another buffer allocated for the GPU, and then written to GPU memory where the data is processed. Copies to and from the host CPU adds unnecessary latency between the sensor and GPU, which can affect time-sensitive applications such as electronic warfare. While subsequent software updates reduced the number of memory copies, reducing latency, the data still...

AMD Radeon HD 6970M and 7970M MXM AMD’s Radeon™ HD 6970M and newest HD 7970M deliver the highest performance and GFLOPS/Watt efficiency in the industry. The HD 7970M alone, built on 28nm semiconductor process technology, squeezes in a higher density of processors while maintaining the same power profile as its predecessors. This means even higher performance at over 2 TFLOPS per MXM and higher GFLOP/Watt while maintaining complete software intellectual property (IP) compatibility between product generations. Total GFLOPS in a 6U Slot With No Change to Power Consumption Random Vibration...

VPX-REDI The VPX (VITA 46) standard defines 6U boards with a modern, high-performance connector capable of supporting today’s highspeed fabric interfaces. VPX is most attractive when paired with the Ruggedized Enhanced Design Implementation standard – REDI™ (VITA 48). The GSC6201 module is implemented as a 6U VPX-REDI conduction-cooled board with an air-cooled variant available in the same VPX form factor for less rugged environments. ™ Targeted primarily for harsh-environment embedded applications, VPX-REDI offers extended mechanical configurations supporting higher functional density,...

Air-Cooled – Mercury Commercial OpenVPX™ System Specifications encompasses Temperature Operating: Storage: Compatible with VITA 65 VITA 46/48.1/48.2 (REDI) *Customer must maintain required CFM level Humidity Operating Vibration Shock Altitude perating O PCI Express® 10-90%, non-condensing 0.003 g2/Hz; 20-2,000 Hz, 1 hr/axis 20g, z-axis; 32g, x-, y-axes; 11 ms half-sine pulse 0-10,000 ft* *Customer must maintain required CFM level Conduction-Cooled – Mercury Rugged Level 3 Temperature Operating: -40°C to +71°C* Storage: -55°C to +125°C Humidity Operating 0-100% Vibration 0.1g2/Hz, 5-2,000Hz,...