Hardware Acceleration Solutions

The aXware and CIB solutions are unique SEMCAD X performance enhancing hardware options. The aXware Accelerator, a simple PCIe add-in card for your computer, and the aXware ClusterInABox, a multi-accelerator workstation, drastically reduce your application runtimes from hours to minutes without compromising accuracy!
Further reducing the runtime by factors of 10 to 50 or more, even Genetic Algorithm based optimization of CAD data becomes now feasible within a few hours only.
This dedicated hardware specifically optimized for FDTD and high bandwidth accelerator memory has proven to be more efficient and cost-effective than multi-processor workstations or multi-node clusters by heavily decreasing initial and long-term maintenance costs.

For more information on the exact performance figures, check the Performance section.

SEMCAD v14 Aletsch bundled with the latest Acceleware CUDA libraries can execute now in cutting edge NVIDIA Quadro technology.  Accelerator cards and Cluster-In-A-Box Solutions is now available in next generation Quadro (series 10)--- featuring larger memory size and bandwidth, higher solver speeds and huge simulation sizes. For extremely large simulation needs and even higher speeds, Linux clusters of NVIDIA Tesla S870 or S1070 GPUs are also available (C30-X and C40-X solutions).

NVIDIA Quadro FX 5800 Card and NVIDIA C1060 (A40):

Interface	PCIx16 express
Speed	up to 650 Mcells/s
Simulation Size (fully accelerated)	75 Mcells*
Memory on Card	4 GB
General Comment	The Quadro FX 5800 is the next generation of the Quadro FX 5600 (formerly Acceleware v3 card). It features higher memory and solver speeds and is an affordable solution for medium to large simulation sizes. The C1060 card has the same characteristics as the FX 5800, except it does not have video output.
Power Supply	1kW

Cluster-in-a-Box Solutions: CIB D40 (Dual) & Q40 (Quad)

Platform	64bit Workstation, Windows XP Pro 64 / Linux 64 -- Windows Vista coming soon
CiB D40	1 NVIDIA Quadro Plex 2200 Model D2
	Speed	up to 1150 Mcells/s
	Simulation Size (fully accelerated)	1-150 Mcells
	Memory on Cards	8 GB
	General Comment	The CIB D40 is the next generation CIB1000 (Quadro Plex 1000 Model IV formerly Acceleware D30) . Its NVIDIA Quadro Plex 2200 Model D2 substation offers improved performance and memory throughput. The CIB D40 is easily maintainable workstation with simulation performance equivalent to a multi-node supercomputer.
CiB Q40	2 NVIDIA Quadro Plex 2200 Model D2 (in parallel)
	Speed	up to 1800 Mcells/s
	Simulation Size (fully accelerated)	1-300 Mcells*
	Memory on Cards	16 GB
	General Comment	The CIB Q40 is the next generation CIB1500 (2 xQuadro Plex 1000 Model IV formerly Acceleware Q30) . The combined capacity of both NVIDIA Quadro Plex 2200 Model D2 substations is an ideal platform for highly complex industrial applications.

Top

The C30-X platform is a cluster of NVIDIA Tesla S870 GPUs that uses SEMCAD X and aXware to provide a powerful and scalable solution for large size problems. X indicates the number of Tesla devices, from 4 up to a maximum of 16. Each Tesla S870 contains four GPUs. Similarly, the C40-X is a cluster of next generation NVIDIA Tesla S1070 GPUs.

The C30-X and C40-X are optimum solutions for extremely large and complex simulations, at high speeds  without sacrificing accuracy.

C30-X	X number of Tesla S870 GPUs (up to 16)
	Speed	up to 2.7 Gcells/s (C30-4) to 14 Gcells/s (C30-16)
	Simulation Size (fully accelerated)	1-700 Mcells (C30-4) 1-2800Mcells (C30-16)
	Memory on Cards	6 GB per node
	General Comment	The C30-X is an extremely powerful and scalable simulation platform, capable of handling grid sizes in the order of Gcells and speeds in the neighborhood of 10 Gcells/s.

C40-X	X number of Tesla S1070 GPUs (up to 16)
	Speed	>15 Gcells/s
	Simulation Size (fully accelerated)	>6 Gcells
	Memory on Cards	16 GB per node
	General Comment	The C40-X is the next generation version of GPU-cluster solution C30-X.

The following plot shows the solver speeds and simulation sizes of the SAM+Generic Phone example in a C30-3 cluster.





Top