Eclipse/GP
Tesla Powered Personal SuperComputer
The EclipseGP is a fully-integrated GPU Personal SuperComputer with up to 4 NVIDIA Tesla C2070 or Quadro FX6000 GPUs. Utilizing the latest DDR5 memory and 1792 GPU cores, the EclipseGP has a peak throughput of nearly 4Tflops. It is essentially a desk side supercomputer at a fraction of the cost, size, and electrical power. The EclipseGP is available as a desk-side workstation or rack-mount.
Key Features
Overview
- Up to 1984 cores (Tesla C2070+visualization/video output graphics card GTS-450)
- Up to 24GB of DDR5 GPU buffer
- Up to 96 of system memory (recommended ratio of ≥4GB per GPU)
- Up to 24TB of hardware RAID disk storage
- CUDA Toolkit pre-loaded
- Linux or Windows
- Desk-side, 3U or 4U rack-mount
CUDA™ Toolkit
Each EclipseGP GPGPU powered system comes loaded with the latest version of nVIDIA's CUDA™ Tool kit for Linux. The CUDA™ Tool kit includes the following:
- nvcc C compiler
- CUDA FT and BLAS libraries for GPU
- Profiler
- gdb debugger for GPU
- CUDA runtime driver
- CUDA programming manual
Specifications
GPUs
1-4 nVIDIA Tesla C2070 each with 6GB DDR5
RAM
Up to 96GB DDR3-1333MHz ECC system memory
CPU
1-2 quad / six core Intel 5500 series processor(s) (support for 1-4 GPUs)*
Disk Storage
1-8 SATA2 / SAS / SSD
73GB to 3TB sized drives, certain sizes only available in certain form factors, 3TB only available in SATA
Networking
(2) 100/1000 Gigabit Ethernet on-board
High-performance Interconnect Support
40Gb/s QDR & 20Gb/s DDR Infiniband*
10Gig Ethernet*
(* a maximum of 3 cards, 1536 cores, with the use of IB or 10G)
Graphics
GTS-450 (single slot card)
on-board VGA**
Management
Optional IPMI Remote Management Card
OS Support
64-bit Linux or Windows
Power
1400w auto-ranging low-noise power supply
100-240V AC, 1x plug
Software
CUDA™ Toolkit
Each EclipseGP GPGPU powered system comes loaded with the latest version of nVIDIA's CUDA™ Toolkit for Linux. The CUDA™ Toolkit includes the following:
- nvcc C compiler
- CUDA FT and BLAS libraries for GPU
- Profiler
- gdb debugger for GPU
- CUDA runtime driver
- CUDA programming manual
