Boston Limited

Take our GPUs for a free test drive

Posted on 12 April 2012

While the new of Intel E5-2600-series processors deliver a considerable performance increase over previous generation CPUs, there's a revolution in processing going on in the background - the switch to a hybrid computing model.

The concept behind hybrid computing is to utilise the previously untapped computational capabilities of GPUs in concert with traditional CPUs to deliver a performance increase an order of magnitude greater than that possible using CPUs alone.

For example, a compute node consisting of a pair of Intel Xeon CPUs plus two Nvidia Tesla M2070 GPUs runs the popular molecular dynamics simulator NAMD a staggering 100% faster than a similarly specified CPU-only compute node. NAMD also scales extremely well across multiple GPUs, with eight nodes performing 200% better than eight CPU-only nodes.

To find out how much GPUs can speed up your code, we offer a free service which provides you with the ability to test drive our GPU cluster - the Nvidia SimCluster. To find out more, and being your test drive, check out our SimCluster.

Comments

Posted by SpecialK on 12 April 2012 at 14:44

Will this work with Fortran and OpenACC ok if so where would I find information on this please?

Posted by James Gorbold @ Boston on 12 April 2012 at 16:20

Hi SpecialK, the Nvidia Tesla GPUs in our SimCluster so indeed support OpenACC and Fortran. In fact, if you're an existing Boston customer we offer free training courses to help you get up to speed with OpenACC. If you email me, james.gorbold@boston.co.uk I'll be able to provide you with more details.

Posted by SpecialK on 12 April 2012 at 16:41

Thanks James for your help on this the training courses would certainly be of interest do you have the ability to let me run my own code on your systems for benchmarking or can you loan me some systems so I can try and run my own code and obtain what I would call real world results? Also, are you aware of PGI's compilers at all and do Boston have experience in support these types of configurations?

Posted by James Gorbold @ Boston on 16 April 2012 at 09:58

Hi SpecialK, the main point of the SimCluster program is to give you the opportunity to run your own code on our GPU cluster, so you can find out for yourself how much faster it will run. If you follow the SimCluster link in the blog post you'll be taken to the webpage where you can book a time slot on our GPU cluster. Re PGI compilers, we have had other clients running applications written in CUDA Fortran on our GPU cluster, so I'm sure our tech team will be able to help you out with any specific queries you have.

Posted by SpecialK on 16 April 2012 at 17:00

Thanks I've just signed up - do you know when you may have Intel's MIC available and also NVIDIA's Keplar based Tesla available I saw this online and found it very interesting? http://blogs.nvidia.com/tag/keplar/

Posted by James Gorbold @ Boston on 20 April 2012 at 04:45

That's great news SpecialK, I'm sure you'll fine it very informative trying out our GPU cluster. I'm afraid I can't confirm when Intel's MIC and Nvidia's Keplar cards will be available, but as Boston is a major partner with both companies, we'll be happy to share more info as soon as we can via this blog.