Accelerators for Technical Computing: Is It Worth the Pain? A TCO Perspective
Nowadays, HPC systems emerge in a great variety including commodity processors with attached accelerators which promise to improve the performance per watt ratio. These heterogeneous architectures often get far more complex to employ. Therefore, a hardware purchase decision should not only take capital expenses and operational costs such as power consumption into account, but also manpower. In this work, we take a look at the total cost of ownership (TCO) that includes costs for administration and programming effort. From that, we compute the costs per program run which can be used as a comparison metric for a purchase decision. In a case study, we evaluate our approach on two real-world simulation applications on Intel Xeon architectures, NVIDIA GPUs and Intel Xeon Phis by using different programming models: OpenCL, OpenACC, OpenMP and Intel’s Language Extensions for Offload.
KeywordsTCO heterogeneous architectures GPU Intel Xeon Phi programming effort OpenCL OpenACC OpenMP Intel LEO energy efficiency
Unable to display preview. Download preview PDF.
- 1.The Green 500: The Green500 List - November 2012: Heterogeneous Systems Re-Claim Green500 List Dominance (2012), http://www.green500.org/lists/green201211
- 2.Carlyle, A., Harrell, S., Smith, P.: Cost-Effective HPC: The Community or the Cloud? In: CloudCom 2010, pp. 169–176 (2010)Google Scholar
- 3.Turner, W.P., Seader, J.H.: Dollars per kw plus dollars per square foot are a better data center cost model than dollars per square foot alone. Technical report, Uptime Institute (2006)Google Scholar
- 4.Patterson, M., Costello, D., Grimm, P., Loeffler, M.: Data center TCO; a comparison of high-density and low-density spaces. Technical report, Intel Corporation (2007)Google Scholar
- 5.Koomey, J.: A Simple Model for Determining True Total Cost of Ownership for Data Centers. Technical report, Uptime Institute (2008)Google Scholar
- 8.Galorath, D.D.: Software Total Ownership Costs: Development Is Only Job One. Software Tech. News 11(3) (2008)Google Scholar
- 11.CAPS enterprise: Case study – Code Migration to CPU-GPU Hybrid: An Economic Approach (2010)Google Scholar
- 12.Wang, L., Khan, S.: Review of performance metrics for green data centers: a taxonomy study. The Journal of Supercomputing, 1–18 (2011)Google Scholar
- 13.Wienke, S., an Mey, D., Müller, M.S.: Accelerators for Technical Computing: Is it Worth the Pain? TCO Spreadsheet (2013), https://sharepoint.campus.rwth-aachen.de/units/rz/HPC/public/Shared%20Documents/WienkeEtAl_Accelerators-TCO-Perspective.xlsx
- 15.Brecher, C., Gorgels, C., Hardjosuwito, A.: Simulation based Tool Wear Analysis in Bevel Gear Cutting. In: International Conference on Gears, Düsseldorf. VDI-Berichte, vol. 2108.2, pp. 1381–1384. VDI Verlag (2010)Google Scholar
- 16.German Science Foundation (DFG): Personalmittelsätze der DFG für das Jahr 2013 (2013)Google Scholar
- 17.Community Research and Development Information Service: Audit Certificate Guidance Notes 6th Framework Programme (2005)Google Scholar