The overall goal of this thesis is to evaluate the feasibility of FPGA based computer system in HPC. This works is performed within ExaNeSt, an EU funded project which aims to develop and prototype energy effcient solutions for the production of exascale-level supercomputers. As the matter of fact, the current computer architectures need to be re-thought in order to reach the exascale performance. Scale current technologies will make the power consumption so large, that to run and maintain such systems will be technologically and economically too much demanding. Let's take for example the system TSUBAME3.0, the #1 in the Green500 list of June 2017, which has an energy effciency of 14.11 GFLOPS/Watt. If we assume to be able to scale this system without loss of effciency to reach 1 exaFLOPS, the resulting system will have a power consumption of 70.9 MW, 500 times more than the actual power (142 kW). This makes clear that the power consumption issues demands the study of a new class of HPC systems based on energy effcient solutions.

FPGA in HPC: High Level Synthesys of OpenCL kernels for Molecular Dynamics(2017 Dec 18).

FPGA in HPC: High Level Synthesys of OpenCL kernels for Molecular Dynamics

-
2017-12-18

Abstract

The overall goal of this thesis is to evaluate the feasibility of FPGA based computer system in HPC. This works is performed within ExaNeSt, an EU funded project which aims to develop and prototype energy effcient solutions for the production of exascale-level supercomputers. As the matter of fact, the current computer architectures need to be re-thought in order to reach the exascale performance. Scale current technologies will make the power consumption so large, that to run and maintain such systems will be technologically and economically too much demanding. Let's take for example the system TSUBAME3.0, the #1 in the Green500 list of June 2017, which has an energy effciency of 14.11 GFLOPS/Watt. If we assume to be able to scale this system without loss of effciency to reach 1 exaFLOPS, the resulting system will have a power consumption of 70.9 MW, 500 times more than the actual power (142 kW). This makes clear that the power consumption issues demands the study of a new class of HPC systems based on energy effcient solutions.
18-dic-2017
Gorlani, Paolo
Brandino, Giuseppe Piero
Cozzini, Stefano
File in questo prodotto:
File Dimensione Formato  
Paolo Gorlani.pdf

accesso aperto

Tipologia: Tesi
Licenza: Non specificato
Dimensione 2.03 MB
Formato Adobe PDF
2.03 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11767/68036
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact