In a bid to outrun Chinese super-computing tech, the US plans to unveil an IBM supercomputer that will roughly double the computing power. The latest supercomputer named — Summit — will be installed at the Oak Ridge National Laboratory(ORNL) and will be rated at 200 petaflops.
In the latest release of the bi-annual Top 500 List, Chinese computer Sunway TaihuLight was discerned as the fastest super-computer in the world. It can achieve a theoretical peak speed of 124.5 petaflops, while the Linpack Benchmark used by Top500 saw it achieve an operational speed of 93 petaflops.
This result trumps China’s own fastest supercomputer Tianhe-2, which was the prior champion for three straight years. The TaihuLight is about three times faster than Tianhe-2, which operated at a peak speed of about 54 petaflops. This caused a ruckus in the supercomputing industry, making U.S blush red with embarrassment.
And this release reported by Computer World, that it is now building a supercomputer in partnership with IBM to rival the TaihuLight couldn’t come at a more perfect time.
In Summit, as disclosed by the Department of Energy’s ORNL will employ IBM’s Power 9 processors in conjunction with Nvidia’s Volta GPU’s. One completed, the Summit will be able to deliver over five times the operational speed of United States fastest supercomputer — Titan. Summit will only employ about 3400 nodes to process data as compared to 18,688 nodes used by Titan, which is currently placed at #3 in the Top 500 list.
Each node of the ‘Summit’ will offer ‘over half a terabyte’ of so-called non-coherent memory coupled with 800 GB of non-volatile RAM that will serve as a memory buffer.
The Supercomputing Supremacy Battle
In addition to Summit, the DOE which controls the super-computer used by the U.S government for scientific research also has two more supercomputers in pipeline for a 2018 launch.
One of them is called Sierra and is planned to operate at a 150 petaflop speed. This is also being developed by IBM and will be placed at the Lawrence Livermore National Lab. While the other one, named Aurora will be a Cray(an older supercomputer system) and Intel system couple will be located at the Argonne National Laboratory.
Even after U.S had banned the sale of Intel high-end microprocessors to China’s four main supercomputing centers, the country has still been able to overpower U.S and build its own microprocessor industry. The most notable fact about the fastest super-computer in the world is also that it does not employ any U.S patents in its development.
It is described as a “domestically designed supercomputer,” meaning the supercomputer was built in-house using its own ShenWei CPUs, developed by Jiangnan Computing Research Lab in Wuxi. These chips use a RISC-based architecture, and each processor has 260 cores capable of just over 3 teraflops.
Also, it is currently being used to run sizeable applications that include advanced manufacturing, earth systems modeling, life science and big data applications.
The DOE concluded their ambitious plan discussion by adding that,
It is pusuing the goals of the National Strategic Computing Initiative: Accelerating the delivery of exascale computing; increasing the coherence between the technology base used for modeling and simulation and that for data analytic computing; charting a path forward to a post-Moore’s Law era.