T
techislife
Guest
Good morning Semiwiki experts!
This is my first post so hopefully it is eligible for discussion regarding semiconductor design and high-performance computing (HPC). This is not exactly breaking news with the European semiconductor industry falling into automotive obscurity, the ongoing pandemic and trade war, but I would be interested to hear your perspective on Europe's prospects to successfully develop advanced chip design and HPC architecture capabilities.
SiPearl, a French startup founded to design the CPU used on EPI's "common platform", has confirmed that its first generation CPU due in 2021 will license ARM Zeus IP. The company is headed by some people formerly employed by Atos and has so far raised €7 million with further financing in the pipeline pending proof of concept. ARM Zeus is apparently designed for N7P and fabrication by TSMC or Samsung thus seems likely.
EPI aims to use accelerators to achieve performance and energy efficiency goals. The common platform under development will cater to heterogenous needs in the future and use the RISC-V ISA. x86 is still clearly the most used HPC computing architecture but I have read ARM server chips have recently gained traction as Intel struggles bringing out denser process nodes.
I have read quite extensively about how the semiconductor industries in Asia formed - Samsung and TSMC are obviously perched high in the value chain today with their leading-edge process nodes. Samsung initially built up semiconductor fab capacity through DRAM and NAND technology transfers and its vertical integration provided ample benefits in the smartphone markets. Taiwan managed to reverse its brain drain with government incentives, the returning Taiwanese engineers and entrepreneurs set up a "hybrid Silicon-Valley system" with a network of constantly improving specialized firms. Most Asian countries entered the semiconductor sector in the OSAT segment, where low-cost labor drew foreign investments and the technological entry barriers were lower.
Now what are SiPearl's chances of success? With the initiative, the EU is tries to secure domestic know-how and supply of HPC tech deemed to be important to maintain European competitiveness in technology (read: the Automotive industry). There is some incumbent semiconductor players such as NXP and ASML but to me it seems they might not have capabilities relevant in HPC. Despite abundant financing at the EU level and long R&D traditions, I fear that there might not be a large stream of capable chip and architecture engineers available.
Thus, the EPI ecosystem will face hardships in the beginning until the roadmap for large-scale deployment of autonomous vehicles becomes clearer. A silver lining is that German car manufacturers have large coffers and extensive R&D experience. Taking a look at Tesla stock might, however, indicate that the proliferation of electric vehicles will instead be the next revolution in personal mobility. Nonetheless, Daniel Nenni recently posted that automotive will be the fastest growing IC segment despite only reaching a 10% market share in the near future.
When is it realistic for Europe to have a domestically designed HPC chipset and architecture to gain a spot in the supercomputing top 10?
This is my first post so hopefully it is eligible for discussion regarding semiconductor design and high-performance computing (HPC). This is not exactly breaking news with the European semiconductor industry falling into automotive obscurity, the ongoing pandemic and trade war, but I would be interested to hear your perspective on Europe's prospects to successfully develop advanced chip design and HPC architecture capabilities.
SiPearl, a French startup founded to design the CPU used on EPI's "common platform", has confirmed that its first generation CPU due in 2021 will license ARM Zeus IP. The company is headed by some people formerly employed by Atos and has so far raised €7 million with further financing in the pipeline pending proof of concept. ARM Zeus is apparently designed for N7P and fabrication by TSMC or Samsung thus seems likely.
EPI aims to use accelerators to achieve performance and energy efficiency goals. The common platform under development will cater to heterogenous needs in the future and use the RISC-V ISA. x86 is still clearly the most used HPC computing architecture but I have read ARM server chips have recently gained traction as Intel struggles bringing out denser process nodes.
I have read quite extensively about how the semiconductor industries in Asia formed - Samsung and TSMC are obviously perched high in the value chain today with their leading-edge process nodes. Samsung initially built up semiconductor fab capacity through DRAM and NAND technology transfers and its vertical integration provided ample benefits in the smartphone markets. Taiwan managed to reverse its brain drain with government incentives, the returning Taiwanese engineers and entrepreneurs set up a "hybrid Silicon-Valley system" with a network of constantly improving specialized firms. Most Asian countries entered the semiconductor sector in the OSAT segment, where low-cost labor drew foreign investments and the technological entry barriers were lower.
Now what are SiPearl's chances of success? With the initiative, the EU is tries to secure domestic know-how and supply of HPC tech deemed to be important to maintain European competitiveness in technology (read: the Automotive industry). There is some incumbent semiconductor players such as NXP and ASML but to me it seems they might not have capabilities relevant in HPC. Despite abundant financing at the EU level and long R&D traditions, I fear that there might not be a large stream of capable chip and architecture engineers available.
Thus, the EPI ecosystem will face hardships in the beginning until the roadmap for large-scale deployment of autonomous vehicles becomes clearer. A silver lining is that German car manufacturers have large coffers and extensive R&D experience. Taking a look at Tesla stock might, however, indicate that the proliferation of electric vehicles will instead be the next revolution in personal mobility. Nonetheless, Daniel Nenni recently posted that automotive will be the fastest growing IC segment despite only reaching a 10% market share in the near future.
When is it realistic for Europe to have a domestically designed HPC chipset and architecture to gain a spot in the supercomputing top 10?