This 3rd edition of the Green500 List contains a plethora of interesting takeaways, of which the following five are most notable.
- The first sustained petaflop supercomputer - Roadrunner,
from DOE Los Alamos National Laboratory - exhibits extraordinary
- Roadrunner, the top-ranked supercomputer in the TOP500, is ranked #3 on the Green500 List. This achievement further reinforces the fact that energy efficiency is as important as raw performance for modern supercomputers and that energy efficiency and performance can coexist. (For comparison, the last two supercomputers to top the TOP500 are #43 and #499 on the Green500.)
- The top three supercomputers surpass the 400 MFLOPS/watt milestone
for the first time.
- Three of IBM’s Cell-based QS22 supercomputers at IBM Germany, Fraunhofer ITWM, and DOE/NNSA/LANL achieved 488.14, 488.14, and 437.43 MFLOPS/watt, respectively.
- Energy efficiency hits the mainstream.
- The energy efficiency of a commodity system based on Intel’s 45-nm low-power quad-core Xeon is now on par with IBM BlueGene/L (BG/L) machines, which debuted in November 2004.
- The FLOPS/watt distribution is still skewed from top to bottom but
- While the energy efficiency of the top-ranked Green500 supercomputer has improved by almost 131,000,000 Flops/watt since November 2007, the bottom-ranked Green500 supercomputer has improved by only 390,000 Flops/watt. Overall, the MFLOPS/watt distribution of the Green500 is skewed with only 28% of the machines acheiving more than 100Mflops/watt (up from 14% last November).
- Each of supercomputers in the top ten from this edition of the Green500 List has a higher FLOPS/watt rating than the previous #1 Green500 supercomputer.
Listed below are the Green500's Top 10 most energy-efficient supercomputers in the world as of June 2008. Please see the sidebar above for the complete June 2008 Green500 list in both online and offline viewing formats.
As actual consumed power can be much less than peak power, we explicitly note which form of power was used to rank the system and encourage all system stakeholders to report measured power -- systems where peak power is used are shown in "gray", while systems where measured power is used are shown in "black".
|Green500 Rank||MFLOPS/W||Site*||Computer*||Total Power (kW)||TOP500 Rank*|
|1||488.14||IBM Germany||BladeCenter QS22 Cluster, PowerXCell 8i 3.2 Ghz, Infiniband||22.76||324|
|1||488.14||Fraunhofer ITWM||BladeCenter QS22 Cluster, PowerXCell 8i 3.2 Ghz, Infiniband||18.97||464|
|3||437.43||DOE/NNSA/LANL||BladeCenter QS22/LS21 Cluster, PowerXCell 8i 3.2 Ghz / Opteron DC 1.8 GHz , Voltaire Infiniband||2345.50||1|
|4||371.75||Argonne National Laboratory||Blue Gene/P Solution||31.50||304|
|4||371.75||Dublin Institute for Advanced Studies/ICHEC||Blue Gene/P Solution||31.50||305|
|4||371.75||Science and Technology Facilities Council - Daresbury Laboratory||Blue Gene/P Solution||31.50||306|
|7||371.67||RZG/Max-Planck-Gesellschaft MPI/IPP||Blue Gene/P Solution||94.50||52|
|7||371.67||Stony Brook/BNL, New York Center for Computational Sciences||Blue Gene/P Solution||63.00||75|
|7||371.67||ASTRON/University Groningen||Blue Gene/P Solution||94.50||51|
|7||371.67||IBM - Rochester||Blue Gene/P Solution||126.00||37|