Analog Devices Inc. ADSP-21160MKB-80

DSP Fixed-Point/Floating-Point 32bit 80MHz 400-Pin BGA Tray / IC DSP CONTROLLER 32BIT 400 BGA
Float Pt Digital Signal Processor -Dsp, 32 Bit, 80Mhz, Bga-400
Digital Signal Processor, 32-Bit Size, 48-Ext Bit, 33MHz, CMOS, PQFP240
The ADSP-21160M DSP is a SHARC processor that is optimized for applications such as telephony, medical imaging, radar/sonar electronics, communications, 3D graphics and imaging. The ADSP-21160M features a Single-Instruction-Multiple-Data (SIMD) architecture. Using two computational units (ALU, Barrel Shifter, MAC, Register files), the ADSP-21160M can have a five fold performance increase versus the ADSP-2106x on a range of DSP algorithms. It is code compatible with ADI's popular first generation ADSP-2106x SHARC DSPs. Like other SHARCs, the ADSP-21160M is a 32-bit processor that is optimized for high performance DSP applications. The ADSP-21160M includes an 80 MHz core for processing both 32-bit fixed-point and (32-bit, 40-bit) floating-point data types, a dual-ported 4 Mbit on-chip SRAM, an integrated I/O processor with multiprocessing support, and multiple internal busses to eliminate I/O bottlenecks. The ADSP-21160M is optimized for multiprocessing topologies; SHARC supports multiprocessing via link ports or an external port. In the diagram below, both cluster and link port multiprocessing is shown. Over the external port up to six 21160Ms may be connected together without any additional support logic. The arbitration for this shared bus is integrated on-chip. Each SHARC in the cluster has access to every SHARC's internal memory. Through the external port connection, optional external memory may be shared as well. The six link ports are another way to gluelessly connect a multiprocessing system. Both, the links and the cluster can be used simultaneously and systems with hundreds of SHARC DSPs can be developed using both peripherals. Maximize Speed, Minimize Size The ADSP-21160M is offered in a 27mm x 27mm 400-ball PBGA (Plastic Ball Grid Array) package 40% smaller than processors in this performance class. More Processing Per Watt In addition to small size and high performance, the 21160M's power dissipation enables new levels of MFLOPS per watt. The ADSP-21160M typical power dissipation is 2 watts, enabling 240 MFLOPS of performance per watt. This allows designers to use multiple processors on a PCI card within the PCI power limit of 25 watts. With eight 21160M DSPs, providing 480 MFLOPS each, a standard PCI card design can obtain 3.84 GFLOPS performance and leave 9 watts for other circuitry. The ADSP-21160M is supported by hundreds of readily available third-party hardware and software products and ADI's powerful VisualDSP development system, allowing for fast and easy development, debug and deployment.