News

6U CPCI (PICMG 2.0, 2.1, 2.16) Intel Core I7-3555 LE, 2,5 GHz CPU module - the rugged heart of your sonar system

19.03.2015

6U CPCI (PICMG 2.0, 2.1, 2.16) Intel Core I7-3555 LE, 2,5 GHz CPU module - the rugged heart of your sonar system

Highly Efficient Multiprocessor Computers for Special Purpose Applications. The article describes a new rugged computer (-40C-+85C) for applications in sonar platforms, based on 6U CPCI (PICMG 2.0, 2.1, 2.16) Intel Core i7 CPU Module CPC503 developed by Fastwel.

INTRODUCTION

Modern computing technologies enable to develop high-performance solutions. Now it takes merely few hours to solve multiple tasks, which earlier required days. Large contribution to the computing performance was made by the introduction of parallel technologies and graphics cards (graphic accelerators) computing. However, the use of many up-to-date technologies for solving practical tasks in industrial and mission-critical applications is often problematic. This is stipulated by strict requirements of the conditions, under which the special-purpose computer is used. In particular, these are: mechanical limitations, consumed power and released heat. These limitations have been keeping us long from creating high-performance special-purpose computers, which are able to solve complex real-time tasks.

Architectures of industrial embedded systems have become very popular nowadays. They are based on special-purpose high-speed data buses such as CompactPCI and enable to create compact computing systems, corresponding to the strict requirements.

Fastwel Group develops multitasking computing systems, including those designed for mission-critical applications, as well as makes special-purpose software for such systems. Together with local system integrator, Fastwel has developed a multiprocessor compact scalable special-purpose computer,

Figure 1: Computing system architecture

designed for the use in civil sonar systems of new generation. The core features of these systems are their working range and possibility of monitoring location of passing transport, passenger and fishing vessels. The systems can also be used for obtaining information on seafloor topography required for geological research works.

ARCHITECTURE

The Box PC contains 6 CPU modules based on Inter Core i7 with 1.5 GHz and 4 GB RAM, connected to each other via CompactPCI bus (Fig.1). This bus is implemented though 2 CompactPCI/Ethernet switches, providing reliable operation of internal bus by redundancy and data transfer at the speed of up to 10 Gb/s. Computing modules specified as CPUMs (CPU modules) and PMs (processor modules), have similar architecture, but differ in terms of their functions. The CPUM is a control module receiving processed data, which then will be distributed over the PMs that perform parallel processing of such data. The computing system is implemented in 6U format based on Schroff 16-slot chassis, chosen due to its high reliability and proper ergonomic characteristics. External view of the system and one of its processor modules is shown on Fig.2

Main features of the computing system (Fig. 2b.) are as follows:

  • Intel Core i7-2610UE (1.5 GHz, 4 MB, 2 cores);

  • RAM: DDR3 SDRAM 4 GB (1333 MHz);

  • Wolf XMC-E6760-VO GPU accelerator;

  • Flash-drive: 4 GB (NAND, up to 100 Mb/s.);

  • HDD: 500 GB (SATA II).

Table1: Technical features of computing system

Feature Computing system
Central Processing Unit (CPU) and form of manufacturing Intel Core i7-2610UE Sandy Bridge with GPU. Serial manufacturing
Process technology 32 Nm
CPU frequency 1.5 GHz
Number of CPU cores 2
CPU performance 18 GFLOPS
Performance 1,2 TFLOPS
RAM DDR3 SDRAM 4 GB 1333 MHz
RAM size 4 GB on cPCI board,
24 GB in 6 x cPCI — 6U
Disc drive size SATA II
Availability to install a floating-point unit or graphics processing unit AMD Radeon E6760 GPU
Operating system MS Windows or Linux
Availability of standard x86/x86-64 programming tools Available (for CPU). GPU programming is carried out by using OpenCL libraries
ECM resistance High
Shock resistance Average
Power consumed by CPU 20 W
Power consumed by cluster 700 W
Dimensions 450×500×300 mm

The system is additionally equipped with a powerful workstation, which includes NVIDEA Tesla GPU accelerators with CUDA technology support.

In terms of its technical features (peak performance, consumed power, dimensions), the developed special-purpose computer is at the same level with the best domestic- and foreign-produced systems, and even outperforms them in some parameters.

Structure of the system additionally provides:

  • Automatic control of fans operation mode;

  • Ability to install rear I/O modules;

  • Time after start required to make system ready for operation: 2 minutes;

  • Weight of no more than 50 kg;

  • Dimensions (W x H x D): 500×450×300 mm.

A distinctive feature of the special-purpose computing system is the use of AMD/ATI Radeon E6760 GPU accelerators. This enables not only to output video images but also use them for data processing in parallel with the CPU, which sufficiently increases general performance. Now in such systems it is possible to use GPU accelerators due to the special line of the Embedded chips manufactured by AMD company, as well as mezzanine boards based on such chips. Architecture and design of the computing system, depending on a task, makes it possible to use different numbers of GPU accelerators: from 0 to 6.

SUMMARY

Using hardware solutions, described in this article, enabled to develop a powerful special-purpose computing system. This computer helped to simulate and improve data-processing algorithms for civil sonar systems, making them more accurate. Due to its high performance, the equipment managed to simultaneously process information incoming from a large number of sensors which made it possible to increase accuracy of the obtained data on the seafloor topography required for geological research works, as well as monitoring of passing transport, passenger and fishing vessels. This was possible only because of paralleling and GPU accelerators. This is a clear confirmation that the use of such architecture solutions with GPU accelerators is highly promising.


img