Supermicro New Zealand Channel Newsletter
Powerful GPU Servers
30 September 2011
This newsletter explains two issues:
- What is Tesla M2090 GPU?
- How to incorporate M2090 for HPC?
Many of the Supermicro Supercomputing server systems integrate GPU computing modules from NVIDIA. The latest Tesla M2090 GPU module, for example, is the world's fastest parallel computing processor for High Performance Computing (HPC). Supermicro servers powered by Tesla GPUs deliver up to 10x higher application performance and are ideal for seismic processing, biochemistry simulations, weather and climate modeling, signal processing, computational finance, CAE, CFD, and data analysis.
The M2090 is the top model of its generation. The other models are M2070 and M2050 and differences are outlined below by NVIDIA.
The basis of building an HPC is provided by Supermicro platforms available in 1U, 2U and 4U/Full Tower form factors, each capable of accomodating up to 3, 4 and 4 double-width GPU cards. The Supermicro 6016GT-TF-FM209 is a good example:
• Dual Quad-Core Intel Xeon processor 5600/5500 series
• Up to 192GB of RECC DDR3 memory
• FM209: 2x Tesla M2090 GPU + PCI-E 2.0 x4 (in x16 slot LP);
FM109: 1x Tesla M2090 GPU + PCI-E 2.0 x4 (in x16 slot LP)
+ 2x PCI-E 2.0 x8 slots
• 3x 3.5" hot-swap drive bays
• Matrox G200eW graphics controller
• IPMI 2.0 with virtual media over LAN and KVM-over-IP
• Dual LAN with Intel 82576 Gigabit Ethernet controller
• 1400W Gold Level high-efficiency power supply with PMbus
Please contact us if you need more information on this technology and class of products.
Supermicro New Zealand
Managed by Compucon New Zealand