Specifications

Technology Overview

Xserve G5

Xserve G5 Cluster

An Xserve G5 cluster uses a dedicated server system as the head node,

which can manage hundreds of Xserve cluster nodes. Apple’s Xserve RAID

adds high-performance, high-availability storage. Each cluster node has

these robust features:

• Computational power. Dual 2.3GHz PowerPC G5 processors provide high-

performance computation and superior scalability.

• Fast PCI-X. Two PCI-X slots support cluster interconnect technologies, such

as InfiniBand and Myricom’s Myrinet.

• Dual Gigabit Ethernet. Two onboard high-speed interfaces enable inde-

pendent gigabit connections to the head node for network services and

to the cluster interconnect for loosely coupled cluster environments.

• FireWire 800. Two FireWire 800 ports support small clusters using TCP/IP

over FireWire. FireWire Target Disk Mode allows easy cloning of system

configurations.

• Advanced cooling system. An array of eight fans keeps air flowing over

high-performance processing components. Intelligent management of the

fan array keeps noise and power consumption to a minimum.

• Indicator lights. Xserve G5 provides continuous, at-a-glance information

about the health and status of the systems in the rack.

• Built-in sensors. Hardware sensors integrate with Server Monitor to provide

easy monitoring of hundreds of systems from a single, intuitive interface.

• Automatic setup. Mac OS X Server makes it easy to set up an entire rack

of systems automatically from a directory on the network or a USB or

FireWire disk.

• Startup options. Front Panel Mode allows the system identifier button

to be used for selecting startup options, such as booting from a network

server or restoring default system settings.

High-Density Cluster Node Configuration

With the compute performance of two superscalar 2.3GHz PowerPC G5 processors,

the Xserve G5 cluster node configuration is ideal for High Performance Computing

(HPC) in scientific and technical environments, as well as for workgroup clusters and

render farms. One dual processor node can execute over 10 billion double-precision

floating-point operations per second, or more than 10 gigaflops per U. That means

a rack filled with Xserve G5 systems can offer over 420 gigaflops of processing power.

Single-precision floating-point performance is also remarkable: With a dual-pipeline

Velocity Engine on each processor, the Xserve G5 cluster node can execute over 35

gigaflops. Best of all, by eliminating the cost of unneeded components, Apple has

made this high-density 1U system extremely affordable.

For more information about Apple solutions for computational clusters, see