Scientific Computing Using Consumer Video-Gaming Hardware Devices

📝 Abstract

Commodity video-gaming hardware (consoles, graphics cards, tablets, etc.) performance has been advancing at a rapid pace owing to strong consumer demand and stiff market competition. Gaming hardware devices are currently amongst the most powerful and cost-effective computational technologies available in quantity. In this article, we evaluate a sample of current generation video-gaming hardware devices for scientific computing and compare their performance with specialized supercomputing general purpose graphics processing units (GPGPUs). We use the OpenCL SHOC benchmark suite, which is a measure of the performance of compute hardware on various different scientific application kernels, and also a popular public distributed computing application, Einstein@Home in the field of gravitational physics for the purposes of this evaluation.

💡 Analysis

Commodity video-gaming hardware (consoles, graphics cards, tablets, etc.) performance has been advancing at a rapid pace owing to strong consumer demand and stiff market competition. Gaming hardware devices are currently amongst the most powerful and cost-effective computational technologies available in quantity. In this article, we evaluate a sample of current generation video-gaming hardware devices for scientific computing and compare their performance with specialized supercomputing general purpose graphics processing units (GPGPUs). We use the OpenCL SHOC benchmark suite, which is a measure of the performance of compute hardware on various different scientific application kernels, and also a popular public distributed computing application, Einstein@Home in the field of gravitational physics for the purposes of this evaluation.

📄 Content

PREPRINT

1

Abstract—Commodity video-gaming hardware (consoles, graphics cards, tablets, etc.) performance has been advancing at a rapid pace owing to strong consumer demand and stiff market competition. Gaming hardware devices are currently amongst the most powerful and cost-effective computational technologies available in quantity. In this article, we evaluate a sample of current generation video-gaming hardware devices for scientific computing and compare their performance with specialized supercomputing general purpose graphics processing units (GPGPUs). We use the OpenCL SHOC benchmark suite, which is a measure of the performance of compute hardware on various different scientific application kernels, and also a popular public distributed computing application, Einstein@Home in the field of gravitational physics for the purposes of this evaluation.

Index Terms—Scientific Computing, Accelerators, Parallel Computing, Supercomputing, GPU, GPGPU, OpenCL, SHOC, Physics

I. INTRODUCTION HERE is considerable current interest in harnessing the advancements made in multi- and many-core technology for scientific high-performance computing (HPC). An example of this trend is the rapid rise in the use of custom- designed HPC general purpose graphics processing units (GPGPUs) as “accelerators” in workstations and even large supercomputers. In fact, the second-fastest supercomputer today, ORNL’s Titan, makes use of Nvidia’s custom-HPC Tesla (Kepler series) GPUs to achieve petascale performance [1] and there are over 100 such accelerated systems in the top 500 supercomputers worldwide. One reason for this recent trend is that the large consumer video-gaming market significantly aids in “subsidizing” the research and development cost associated towards advancing these compute technologies, resulting in high performance at a low cost. In addition to their cost effectiveness, these GPU technologies

This article was submitted for peer-review on July 1st, 2016. This work was supported in part by the NSF award PHY-141440 and by the US Air Force agreement 10-RI-CRADA-09.
Mr. Glenn Volkema, is a member of the Physics Department and the Center for Scientific Computing and Visualization Research at the University of Massachusetts Dartmouth, North Dartmouth, MA 02747 USA (e-mail: gvolkema@umassd.edu). Dr. Gaurav Khanna is a Professor in the Physics Department and the Associate Director of the Center for Scientific Computing and Visualization Research at the University of Massachusetts Dartmouth, North Dartmouth, MA 02747 USA (e-mail: gkhanna@umassd.edu).

are substantially “greener” over CPUs, delivering higher computational performance per Watt of electrical-power consumed [2]. In a similar spirit, there is also the opportunity to utilize the commodity “off-the-shelf” video-gaming hardware itself for scientific computing. Here we are specifically referring to consumer-grade video-gaming cards and also game consoles themselves, as opposed to the custom-HPC variants. These often have similar high-performance characteristics and in addition, are sold at a significant discount (sometimes even below manufacturing cost) because the business model allows for making up the deficit through the sales of video games and other application software. Starting in 2007 the authors were pioneers in successfully utilizing a cluster of Sony PlayStation 3 (PS3) [3] gaming consoles for scientific computation [4]. They were able to demonstrate order-of-magnitude gains in efficiency metrics such as performance-per-dollar and performance-per-Watt as compared with traditional compute clusters [5,6,7]. Since then many universities and research groups took a similar approach and evaluated the value of such PS3 hardware for their own computation needs. The largest such system has been built by the Air Force Research Laboratory (AFRL) in Rome, NY. This system, named “AFRL CONDOR” utilizes 1,716 PS3s alongside traditional servers and Nvidia Tesla (Fermi series) GPGPUs to achieve 500 TFLOPS of computing power [8]. CONDOR has been demonstrated to be over 10× more cost-effective in similar metrics [9].

In this article we explore the capabilities of current generation consumer-grade, video-gaming hardware for scientific high-performance computing. Specific examples of the compute hardware that we consider interesting for this study are current gaming graphics cards like the AMD Radeon [10], Nvidia GeForce [11] series, and also the CPU-GPU “fused” or heterogeneous processor architectures like the AMD’s Accelerated Processing Unit (APU) [12] and the Nvidia’s Tegra System-on-a-Chip (SoC) [13]. The main advantage of considering such consumer-grade hardware for scientific HPC is low cost and high power-efficiency. Rapid advances and significant innovation is being enabled through major investments made by the gaming industry. This is driven by strong

This content is AI-processed based on ArXiv data.