AMD is one of Silicon Valley’s oldest semiconductor companies and we were excited to see our iconic AMD Radeon™ graphics cards featured in an episode of the hit HBO series “Silicon Valley”. In the show, the characters use graphics cards to build a high performance compute cluster in order to run their proprietary universal compression engine “800 times faster than on the CPU alone” – or so they claim.
What we do know is that AMD Radeon graphics cards are the ideal tools for gaming, providing the ultimate visual experience for gaming and virtual reality. Since the fictional team at “Pied Piper” created a compression solution to solve big data storage problems, not play games, it’s safe to assume they might see even more performance if they used AMD FirePro™ professional graphics cards for high performance compute.
The idea of using graphics cards is not new and certainly not confined to the world of television. In fact the world’s most energy efficient supercomputer, the L-CSC based at the GSI Helmholtz Center makes use of AMD FirePro S9150 server graphics cards and sits on top of the prestigious Green 500, (November, 2014 ranking), a list that ranks supercomputers by energy efficiency.
We’ve all been there; you need to build a high performance compute cluster for research, financial modelling or big data analysis but you don’t want to melt the ice caps or empty your wallet doing simulations. The good news is that this is not an isolated problem and thanks to AMD and our range of FirePro graphics cards, you can crunch numbers in an energy efficient manner while keeping your bank manager happy.
AMD FirePro graphics can run code that has been programmed in the industry-standard OpenCL™ language. This language is widely supported and has a highly active developer ecosystem and libraries such as clMath that include Fast Furrier Transform (FFT) and Basic Linear Algebra Subprograms (BLAS), enabling researchers to quickly accelerate their programs on AMD FirePro cards.
So what makes AMD FirePro such a good fit for high performance computing? Take the aforementioned AMD FirePro S9150 server GPU, a number crunching monster that devours datasets and produces data that scientists and researchers can use for research papers and engineering.
The chip inside the card is designed for energy-efficient double-precision calculations – with up to 2.53 TFLOPS of peak performance -- so work can be done to greater precision without sacrificing performance – and is absolutely critical when it comes to scientific research and financial modelling.
An AMD FirePro card needs data to work on and, unlike consumer graphics cards, the server-orientated AMD FirePro S9150 GPU has 16GB GDDR5 memory to store vast datasets. Alongside capacity you need reliability and AMD FirePro S-series cards use Error Correction Code (ECC) technology to ensure data integrity and the right results.
As AMD FirePro cards are meant to be used in harsh datacenter environments. They are designed to work without any direct cooling; instead, they use the cooling already in the server. This means there are fewer moving parts in the server; helping ensure users get the maximum uptime possible.
AMD FirePro cards are much more than brutal number crunchers. Having all of this compute power comes to nought if it isn’t energy efficient. Being energy efficient is not just about saving money on electricity bills; energy efficiency means being able to pack more compute per watt. AMD FirePro graphics cards have proven to be ahead of the pack, with the L-CSC cluster using AMD FirePro S9150 cards certified as the world’s most energy efficient high performance computing cluster.
It’s important to remember that you should use the right tool for the job. If you find yourself in a similar situation to Gilfoyle, co-CTO of your hot new Silicon Valley start-up, do your homework. The odds are you are looking for an energy efficient compute cluster, so go with what the pros use. Go with AMD FirePro GPUs!
Karl Freund is a Corporate Vice President at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites, and references to third party trademarks, are provided for convenience and illustrative purposes only. Unless explicitly stated, AMD is not responsible for the contents of such links, and no third party endorsement of AMD or any of its products is implied.