Munich, 7 May - NextSilicon Ltd., a deep-tech startup focused on developing a next-generation computing platform for the high-performance computing (HPC) market, and ParTec AG, a leading specialist in the integration, development and design of modular supercomputers and quantum computers, along with the development of accompanying software, have forged a collaboration with the goal of accelerating research workflows and improving energy efficiency using Maverick-1 - NextSilicon's first generation run-time reconfigurable hardware.

Leveraging intelligent software algorithms, NextSilicon dynamically reconfigures its hardware to optimize performance and power efficiency based on real-time application demands. Maverick-1 is a demonstration platform that proves the core capabilities of dynamic runtime optimization. The software-defined hardware platform is a novel approach to accelerating HPC without extensive code rewrites.

The Zuse Institute Berlin (ZIB) will be the first European customer to receive the brand new NextSilicon second generation technology, Maverick-2, to be delivered by ParTec. ZIB is one of nine high-performance computing centers in the National High-Performance Computing Consortium (NHR). By adopting the NextSilicon technology, ZIB will be further expanding high-performance computing capacities for research and development in Germany. In March 2024, ParTec, ZIB, and NextSilicon organized a two-day hackathon in Berlin, gathering a select group of HPC experts and technicians. The event served as an opportunity to introduce participants to NextSilicon's innovative approach to HPC. At the hackathon, participants engaged in hands-on sessions to learn about the programming environment, profiling analysis, and optimization methods. With early access to the Maverick-2 hardware, ParTec is poised to organize the second-generation Maverick-2 training workshops with ZIB in late 2024.

Expanding on this collaboration, ParTec will serve as NextSilicon's trusted partner in Europe for the distribution of their products and services. This strategic partnership will enable both parties to leverage their extensive expertise, thereby strengthening their market position.

ParTec quote:

"ParTec is delighted to partner with NextSilicon and support ZIB in their exploration of this novel architecture. Our recent accelerator hackathon, hosted in collaboration with ZIB and NextSilicon, was a resounding success, with participants gaining valuable hands-on experience with this novel runtime hardware," said Bernhard Frohwitter, CEO of ParTec AG.

NextSilicon quote:

"I am thrilled about our close relationship and technical exchange with ParTec in our collaboration with the Zuse Institute Berlin and the opportunity to integrate our next-generation computing solutions into their well-respected research environment. ParTec´s technical expertise and ZIB's commitment to pushing the boundaries of HPC using innovative architectures aligns perfectly with our vision. Our reconfigurable hardware architecture is designed to maximize efficiency and performance, enabling ZIB to unlock the full potential from their computational tasks. We look forward to seeing the remarkable outcomes that our combined expertise will undoubtedly yield," said Elad Raz, founder and CEO of NextSilicon.

ZIB quote:

"We are excited to work with the new innovative hardware and software technology developed by NextSilicon. For our work we can draw on the in-depth technical expertise offered by ParTec which is critical in this phase. With the increasing pressure to implement energy-efficient solutions, data flow architectures are becoming more attractive than ever. With NextSilicon's combination of dataflow engine and compiler-assisted hardware-software partitioning and code optimization, we are poised to take significant strides towards energy-efficient computing," said Thomas Steinke, head of the supercomputing department at ZIB.

Attachments

  • Original Link
  • Permalink

Disclaimer

ParTec AG published this content on 07 May 2024 and is solely responsible for the information contained therein. Distributed by Public, unedited and unaltered, on 07 May 2024 07:21:01 UTC.