"E4 Engineering and NVIDIA offer a cluster with 8 Tesla GPU at no charge, to experience the benefits of GPU Computing"
In different contexts we were able to analyze areas for utilization of the GPU of the latest video cards are not related to 3D graphics. These architectures are so powerful in the areas of parallel processing that it can be used to run code typically directed to the calculation on the CPU, obtaining sensitive decreases in processing times.
The segment of GPU computing, a term with which those processes are given of calculation related to 3D graphics that are performed using the computing power of GPU your system, is gradually taking shape in both the consumer and professional. All has obviously puts the greatest interest, since a reduction in processing time is always appreciated: it allows for equal time to perform more calculations or to get answers faster.
Have a package of GPUs for parallel processing is not sufficient to reduce the processing time. It is essential that the application is liable, by its nature, to be executed in parallel and that the code be adjusted so that they can be executed on a GPU. NVIDIA is working for some time in this direction through Cuda and its family of hardware solutions Tesla, based on the family GT200 GPU and available both as a dedicated video card as a server rack or workstation outside.
E4 Computer Engineering, a company engaged in the production of committed server systems particularly oriented to the HTPC, established a new initiative aimed at the dissemination of NVIDIA Tesla. A cluster of 8 GPU was made available, free of charge and with remote access, customers E4 Computer Engineering is to study the possible benefits for the parallel GPU via its code.
The hardware architecture includes making available two servers E5118 1 unit, combined with Tesla S1070 two servers each with 4 GPUs working in parallel. The processing power theoretical maximum available, in single precision, is 8 TeraFLOPs in parallel using all 8 GPUs NVIDIA made available. Here's the press release provided by the producer, referring to this website for all the insights.
The HPC has for many years one of the core business of E4 Computer Engineering, and the company wants to give their customers the opportunity to experiment in remote connection, the computing power of HPC clusters E4 based on NVIDIA Tesla consists of two 1U Twin server (4 knots) and two E5118 servers Tesla S1070. The cluster is able to ensure a level of performance of 8 TeraFLOPS calculation, up to 150 times higher compared to traditional systems and reduced energy consumption, thanks to the hybrid hardware configuration (CPU + GPU). This scalable solution is a very important part of E4 Computer Engineering, towards the promotion of hardware architecture for HPC applications that include heterogeneous computing technologies.
Bookmarks