Microway BioStack-LS Earns BioIT World 2011 Best of Show Award!
We're pleased to announce that the BioStack-LS, our modular, extendable managed Tesla GPU cluster, was named the winner of the IT Hardware and Infrastructure category. Even with tough competition from the likes of SGI and Isilon our winning combination of Microway hardware and technical expertise, NVIDIA GPUs, and easy-to-use Bright Cluster Manager software was extraordinarily compelling to the expert judges. The award-winning cluster included:
- 7 GPU Compute Nodes: each with 2 Tesla Fermi C2070s
- Delivered configured for life sciences software, including: AMBER, MATLAB, NAMD, and VMD
- 6272 CUDA cores with 14.42 TFLOPS of GPU performance (7.21 TFLOPS DP)
- 84 CPU Cores: 12 Intel Xeon CPU Cores per node at up to 3.06Ghz
- Microway 36 Port FasTree™ InfiniBand Switch
- Bright Cluster Manager™ with CentOS Linux installed and configured
- Full specifications and datasheet are available here
See A Video Interview with BioIT World on the Award
BioStack-LS can also be configured for other applications including simulation, CFD, FFTs, medical imaging, weather modeling, computational finance, and numerical analysis.
Boston Area GPU Programmers: HPC & GPU Supercomputing Group of Boston Meetup
Tonight, May 4th, at 6pm at Microsoft NERD Center, Cambridge, MA
at any given time, we have at least one WhisperStation with 2 or 4 Tesla GPUs installed available at our Plymouth, MA facility for remote login. See Eliot Eshelman or Mike Fried at the meeting for more information.
Did You Know About GPU Direct?
NVIDIA and Mellanox have partnered to optimize GPU compute
performance. NVIDIA GPUDirect technology allows a cluster of GPUs to
communicate across high-performance InfiniBand links without creating buffer
copies of the data. (Video)
New 56Gb/sec InfiniBand Cluster Interconnects with SwitchX Technology
Mellanox announced SwitchX, a protocol-independent ASIC for InfiniBand, 10G/40G
Ethernet and Fibre Channel. It will support the new FDR (56Gb/sec) InfiniBand standard. With up to 4Tb/s low-latency switching capacity,
SwitchX will power the next generation of HPC clusters. Prototype availability immediately with a general release later this year. See more information with this release