What is AMD
What is AMD’s
Heterogeneous
System
Architecture?
By
Ralph Efftien
Polytechnia Institute CGS-1280C Computer Hardware
Berkeley Open Infrastructure for Network Computing (BOINC) evolved from the SetiAtHome screensaver program created by Dr. David Anderson at University of California – Berkeley. BOINC over the major revisions has gone from running scientific applications during a CPUs idle time to also running the apps on the GPU (BOINC, 2013). They first started with NVIDIA’s CUDA routines and added ATI’s CAL routines to enable projects applications to perform scientific calculations on their various GPUs. With the acquisition of ATI by AMD, AMD has since dropped support for the CAL routines and has gone with supporting OpenCL for running general applications of the graphics coprocessor. The following list are some of the projects that use the GPU:
1. http://einstein.phys.uwm.edu/
2. http://boinc.thesonntags.com/collatz/
3. http://milkyway.cs.rpi.edu/milkyway/
4. http://boinc.fzk.de/poem/
5. http://www.primegrid.com/
6. http://setiathome.berkeley.edu/
7. http://setiweb.ssl.berkeley.edu/beta/
8. http://moowrap.net/
These projects are the first that use the GPU to perform calculations on work units. They all started with applications that BOINC ran CPUs and the programmers at the various projects then adapted to run on the GPU. BOINC being a multi-threaded program is able to run one project on each core of a systems CPU and anywhere from one to four instances of an application on the GPU depending upon the GPU’s compute capabilities and the project’s implantation of the application. The Milkyway project is an interesting one, in that it requires the GPU to have double precision math capabilities and thus only runs on a limited number of graphics cards. The CPU have been designed to run general programming tasks that are serial in nature or tasks that may run in parallel. Whereas the GPU has now evolved to being very
Heterogeneous
System
Architecture?
By
Ralph Efftien
Polytechnia Institute CGS-1280C Computer Hardware
Berkeley Open Infrastructure for Network Computing (BOINC) evolved from the SetiAtHome screensaver program created by Dr. David Anderson at University of California – Berkeley. BOINC over the major revisions has gone from running scientific applications during a CPUs idle time to also running the apps on the GPU (BOINC, 2013). They first started with NVIDIA’s CUDA routines and added ATI’s CAL routines to enable projects applications to perform scientific calculations on their various GPUs. With the acquisition of ATI by AMD, AMD has since dropped support for the CAL routines and has gone with supporting OpenCL for running general applications of the graphics coprocessor. The following list are some of the projects that use the GPU:
1. http://einstein.phys.uwm.edu/
2. http://boinc.thesonntags.com/collatz/
3. http://milkyway.cs.rpi.edu/milkyway/
4. http://boinc.fzk.de/poem/
5. http://www.primegrid.com/
6. http://setiathome.berkeley.edu/
7. http://setiweb.ssl.berkeley.edu/beta/
8. http://moowrap.net/
These projects are the first that use the GPU to perform calculations on work units. They all started with applications that BOINC ran CPUs and the programmers at the various projects then adapted to run on the GPU. BOINC being a multi-threaded program is able to run one project on each core of a systems CPU and anywhere from one to four instances of an application on the GPU depending upon the GPU’s compute capabilities and the project’s implantation of the application. The Milkyway project is an interesting one, in that it requires the GPU to have double precision math capabilities and thus only runs on a limited number of graphics cards. The CPU have been designed to run general programming tasks that are serial in nature or tasks that may run in parallel. Whereas the GPU has now evolved to being very
References: BOINC. (2013). Retrieved from boinc.berkeley.edu: http://boinc.berkeley.edu/ George Kyriazis, AMD. (2012, August 30). Heterogeneous System Architure: A Technical Review. Pollice, M. (2013, March 3). Analysis: AMD Kaveri APU and Steamroller Core Architiectural Enchancements Unvieled. Retrieved from Bright Side of News: http://www.brightsideofnews.com/news/2013/3/6/analysis-amd-kaveri-apu-and-steamroller-core-architectural-enhancements-unveiled.aspx What is Heterogeneous System Architecture (HSA)? | AMD. (2013). Retrieved from developer.amd.com: http://developer.amd.com/resources/heterogeneous-computing/what-is-heterogeneous-system-architecture-hsa/