- Cluster Max™ Super G-xF GPU cluster is a turnkey solution powered by the NVIDIA® Tesla™ GPU computing platforms, based on NVIDIA’s Kepler K20/K20X GPU Computing Accelerators and the latest generation of Dual Socket Enterprise class 22nm Intel® Xeon® E5-2600 v2 or 32nm Intel® Xeon® E5-2600 family processors, delivering up to 497.70 Teraflops of single precision and 166.32 Teraflops of double precision performance per 42U cluster.
- In addition, the Cluster Max™ Super G-xF offers customers in the financial services industry tremendous value and lower total cost of ownership through high performance risk analysis & simulations, faster installation and deployment, ease of serviceability, minimal risk of downtime, as well as power and cooling buffers.
AMAX ClusterMax™ SuperG Key Features:
- Delivers up to 338,688 GPU cores, 497.70 Tera flops of single precision and 166.32 Teraflops of double precision performance per 42U cluster
- Up to 756GB dedicated GPU memory
- Delivers 3 times higher performance per watt than the previous “Fermi” compute architecture
- Delivers 1 petaflop of computing in just 10 server racks.
- Supports dual socket 22nm 12-core Intel Xeon® E5-2600 v2 or 32nm 8-Core Intel Xeon® E5-2600 family processors on host systems
- Supports QDR/FDR InfiniBand fabric & real time InfiniBand diagnostics
- Faster communication with Infini Band using NVIDIA GPUDirect™ technology
- TCC driver reduces CUDA kernel overhead and enables Windows Remote Desktop and Windows Service
- Cluster management and GPU monitoring software, including GPU temperature monitoring, fan speed, and power, providing exclusive access to GPUs in a cluster
AMAX’s Cluster Max™ Super G Tesla preconfigured cluster comes with preinstalled with
- Red Hat Enterprise Linux 5.x/6.x, 64-bit or SuSe Linux Enterprise Server 11x, Red Hat Enterprise MRG
- CUDA 4.x, 5.x Toolkit and SDK
- Clustering software (IBM Platform Cluster Manager, Bright Cluster Manager)
GPU Software Development Tools
- Optional C-based software development tools and various libraries for GPUs
- CUDA compatible clustering software
Complete Cluster Distribution
- Free IPMI management suite
- Seamless standard and custom application integration & cluster installation
- Firmware upgrades & BIOS modification
- Multiple Interconnect Options: InfiniBand, Fibre channel, Ethernet (Gigabit, 10GbE)