Warning: file_put_contents(/srv/users/serverpilot/apps/bitupdateus/public/wp-content/plugins/bulk-post-0.4-1/cache/sessions//n8JA7WNOe1HDHOXhAm0VGkTaKdhnHe0wbnnLYOfs): failed to open stream: No space left on device in /srv/users/serverpilot/apps/bitupdateus/public/wp-content/plugins/stupidpie-1.8.3-1/vendor/illuminate/filesystem/Filesystem.php on line 122
  Chinese Supercomputer Tianhe 2 Doubles Computing Performance Through New Accelerator Chip | Bit Updates
Home » bitcoin updates » Chinese Supercomputer Tianhe 2 Doubles Computing Performance Through New Accelerator Chip

Chinese Supercomputer Tianhe 2 Doubles Computing Performance Through New Accelerator Chip

Wednesday, September 20th, 2017 | bitcoin updates

  Chinese Supercomputer Tianhe 2 doubles computing performance by new accelerator chip



          20.09.2017 18:59 UhrAndreas Stiller

    China currently occupies the first two places in the ranking of the fastest computers.
                (Image: dpa, Top 500 Supercomputers Site)
              A homemade Chinese accelerator chip with 128 cores replaces the old Intel Xeon Phi processors in the supercomputer "Milky Way 2".
            At the International HPC Forum (IHPCF) in Guangzhou, the National Supercomputer Center (NUDT) at the International HPC Forum (IHPCF) learned that the old maps with Xeon-Phi-Knights-Corner (KNC) in Tianhe (Milky Way) and replaced it with a Matrix 2000 GPDSP accelerator developed in China.
 2.2 TFlops on 128 threads
The exchange is probably not yet through, but to the next Top500 list, which will be published in November to the SC17 in Denver, one will surely make it. Most of the information known so far about the system comes from the Tokyo professor Satoshi Matsouka, who was allowed to visit the plant together with Linpack creator Jack Dongarra and others. Jan Lin from the HPC Center at the University of Shanghai retook his tweets in English.

            The huge Marix 2000 chip, hioer in the hand of Linpack-creator Jack Dongarra

        Image: Satoshi Matsuoka / Jan Lin
The rumor mill says that the huge Matrix 2000 chips could be 128 in-order ARM chips with two 256-bit vector extensions. In the double-precision matrix multiplication DGEMM the Matrix 2000 with 1.2 GHz to 2.2 TFlops with 128 threads with 90.2 percent efficiency is to come to 240 Watt TDP. The old Xeon Phi 31S1P (with 57 cores) reaches only 840 GFlops in the DGEMM. For this reason, its memory bandwidth was up to 160 GB / s with Stream (Triad) a lot higher, the Matrix 2000 is 96 GB / s with 128 cores.
Theoretically up to 5.34 TFlops
As before, a node consists of two Xeon CPUs, but now expanded by two Matrix 2000 accelerators and equipped with 192 GB of memory. The theoretical peak power of a node is thus a total of 5.34 TFlops, of which 430 GFlops contribute the two Xeons – that is, the retention of the old Xeon 2692 processors with 12 cores and 2.2 GHz clock.
However, the node count is expanded from 16,000 to over 17,792. Thus, the complete Tianhe 2A computer reaches 94.7 PFlops peak power overall, compared to 54.9 PFlops before. Thanks to the proprietary 112 Gb / s network, the Linpack performance of the entire system is expected to be over 90 percent, which is expected to reach more than 80 PFlops – the predecessor system, the Tianhe 2, which headed the Top500 list for three years until 2016 to 33.9 PFlops.
The total energy consumption is still around 18 MW, which means that the energy efficiency has increased from 1.9 GFlops / W to more than 5 GFlops / W.




Comic Culture in Berlin:

Endurance, he has: When Ingo Römling was once booked by Disney as a

Groko Negotiations News B

                                  A dozen points on the dissent CDU, CSU and SPD began

Electronic Assistant at S

Man can not be displaced so easily. "Klingelingeling should I please here?", Asks

Range for digital radio D

                                                                          Antenna Television in Germany has