Freescale Extends QorIQ Qonverge B4 Family to Address Industrial and General-Purpose Markets

Freescale’s advanced floating point technology is engineered to exceed the stringent requirements of computation-intensive applications that require high-precision numerical accuracy, such as radar, medical and media imaging, MAC-scheduler processing in Cloud-RAN, video analytics, security, industrial automation, and test and measurement systems.

To support these applications, Freescale added floating point arithmetic to the SC3900 processing unit, the industry’s highest performing fixed-point DSP core*. The floating point capabilities complement the broad array of advanced technologies already integrated into QorIQ Qonverge SoCs, including high-performance processing, security and network processing acceleration engines.

“Freescale’s breakthrough QorIQ Qonverge portfolio has experienced strong market acceptance and broad adoption throughout the communications and networking equipment space,” said Tareq Bustami, vice president of product management for Freescale’s Digital Networking business. “By adding advanced floating point arithmetic to support compute-intensive tasks, we are enabling a broader customer base to spark innovation with products leveraging the power, performance and cost advantages of QorIQ Qonverge technology.”

With advanced 28-nm process technology, the B4860 generates up to 14.4 GHz of total programmable processing performance and integrates four e6500 dual-threaded Power Architecture processor cores resulting in 61,000 processor CoreMark. Six fixed- and floating-point SC3900FP DSP cores yield 230.4 GMACS of fixed-point performance and 115.2 GFLOPS of floating-point performance. Combined with the four AltiVec SIMD engines embedded in e6500 cores, a total of 172.8 GFLOPS of IEEE 754 floating-point performance is attained. This combination of powerful programmable cores, together with built-in IP packet processing and security acceleration engines, delivers outstanding flexibility, exceptional processing performance and facilitates the development of a broad range of applications requiring massive processing performance for energy-efficient systems.

Freescale added floating point support to the production version of the B4860 and B4420 SoCs while retaining binary code compatibility with the existing SC3900 core. This enables continuity for customers already developing software for these SoCs.

Different from many other heterogeneous multicore architectures, the B4 family provides a full cache-coherent system keeping data consistency of the shared resource stored in local caches; not only between L1 and shared L2 caches within the core clusters but also between the different types of core clusters and external interfaces with no software intervention. This performance increase eases multicore programming and significantly speeds time to market.

High-speed and low-latency industry-standard interfaces are included to support the powerful SoC computing resources. The B4860 SoC incorporates a suite of high-speed standard interfaces pertinent for a multi-purpose SoC, including two 10 Gbps XFI/XAUI Ethernet controllers, four 2.5 Gbps/1 Gbps SGMII Ethernet controllers, two Serial RapidIO v2.1 controllers supporting eight lanes at 5G, and a PCI Express Gen2 controller supporting four lanes at 5G. It also includes two DDR3/3L controllers, 64/72-bit with ECC at 1.867 GHz, NAND/NOR flash controller, I2C, SPI, SDHC, USB and UART interfaces.

Development tools and enablement

Freescale’s development ecosystem for the B4860 includes the CodeWarrior Integrated Development Environment based on Eclipse technology, which provides a comprehensive multicore development platform. Available tools include the B4860QDS and B4420QDS development boards, C optimizing compilers for Power Architecture and StarCore cores, source-level multicore debuggers, device and core software simulators, optimized device drivers, a Linux SDK, and software analysis tools for profiling and program/data trace. In addition, Freescale and its partner network offer highly efficient MPU and DSP operating systems with BSPs and SDKs including optimized device drivers.