Most supercomputers are aimed on only the speed of processing. Consider the new Summit system by DOE that is now the most powerful supercomputer in the world. It has more than 27,000 NVIDIA Tesla V100 GPUs and 9,000 22-core IBM Power9 chips. But processing speed is not everything. In 2017, HPE (Hewlett Packard Enterprise) declared The Machine, its sample for a supercomputer developed around scorching speedy memory. It is meant to mix through tons of information, although it can manage its fair share of HPC (high performance computing) tasks.
Now, HPE is converting that dream into a reality in the form of Astra, the biggest ARM-supported supercomputer ever developed. Designed mutually with the Department of Energy, it is being used by the Sandia National Laboratory as a tentative new platform for nuclear study. As it is fueled by ThunderX2 Cavium ARM chips, it is significantly denser and more power efficient (meaning it can integrate additional hardware) in comparison to an analogous x86 system. Remarkably, that ARM processor also provides 33% faster memory speeds in comparison to most of the x86 CPUs.
“The energy required to shift information around a system is greater than the power required to calculate that information,” claimed VP of HPE’s Advanced Technologies group, Mike Vildibill, to the media in a statement. As a result, the requirement for more well-organized information transfer is one factor why HPE is entering into memory-powered systems.
Speaking of HPE, the firm at its yearly Global Partner Summit declared winners of the 2018 HPE Partner of the Year Awards in Las Vegas. Winners were rewarded for accomplishments and outstanding performance as well as for boosting meaningful business outcomes for shared users. “As a longstanding and leading worldwide distribution associate for HPE, we are offering the business resources, value, and expertise our associates expect from Ingram Micro,” claimed executive vice president at Ingram Micro, Paul Bay.