CNRG Beowulf Cluster
The CNRG Beowulf Cluster is dedicated to simulation of large scale neural networks. Funding for the CNRG Beowulf was provided by the Canada Foundation for Innovation and the Ontario Innovation Trust.
Hardware
Master Node
2.8 Ghz Intel Pentium 4 Processor
Broadcom Gigabit Ethernet Controller
4 x Intel Gigabit Ethernet Controller
64 MB ATI Radeon 9000 AGP Graphics Card
4 x 1024 MB 266 MHz DDR RAM
4 x 120 GB WesternDigital Hard Drive
Slave Nodes (64)
2 Ghz intel Pentium 4 Processor
1024 MB 266 MHz DDR RAM
Intel Gigabit Ethernet Controller
Network
3 x Extreme Networks Summit 7i Gigabit Switchs
Software
Operating system: Debian Linux
Message Passing Interface libraries: MPICH v1.2.6
Matlab v6.5
NetSolve v2.0
NESim
Performance
The performance of the CNRG beowulf cluster has not yet been fully characterized. The table below describes the run times obtained using the LinPack HPL Benchmarks compiled with g77 (version 2.95.4) and linked with MPICH library (1.2.6).
| HPL LinPack Performance | |
| Number of Nodes | Peak Performance |
| 12 | 8.8 Gflops |
All material on this site is copyrighted