2022年11月8日

越來越多的摩爾定律的腦袋,電腦芯片開發人員采用“chiplet”架構擴展硬件的處理能力。Chiplets像集成電路塊設計與其他工作,類似Chiplets形式複雜,可堆疊芯片提高性能,同時保持類似的物理足跡。Chiplets提供許多優勢傳統的設計。但裝配問題,以及挑戰在平衡成本、性能、功耗和投放市場的時間,經常困擾他們早期的階段。旨在克服障礙chiplet創建、白木Fajadrad, Syrus Ziai和帕特裏克Soheili Eliyan成立,chiplet互連的啟動,在2021年。Eliyan的技術——被稱為NuLink連接chiplet組件使用標準的芯片封裝,導致該公司聲稱是執行和更節能的芯片。”的重點是開發一個方法使更多高性能、低功率和轉型互連chiplet架構,而專家認為是唯一的道路繼續規模摩爾定律,“Farjadrad告訴TechCrunch在一封電子郵件采訪。“我們用我們的技術標準包裝,從而節省時間,成本和開發工作相比其他互連方案需要更先進的包裝。此外,我們的方法具有可持續性的好處通過減少材料成本和生產過程中的浪費,降低能耗對高性能計算芯片。“Eliyan的根是在之前啟動,Aquantia,邁威爾公司2019年收購。Farjadrad說技術自2017年以來一直在開發; he co-started Aquantia and served as the startup’s chief engineer for nearly 15 years. Prior to co-founding Eliyan, Farjadrad spent several years at Marvell as CTO and VP of the company’s networking and automotive division. Ziai is a former Qualcomm engineering VP, while Soheili was previously VP of business development at semiconductor firm eSilicon. While Eliyan hasn’t launched its technology commercially yet — it expects the first silicon to hit the market in Q2 2023 — the company claims to have achieved the last step before manufacturing, a tape-out, using semiconductor manufacturer TSMC’s 5 nm process. “Process” in chip lingo refers to an architectural platform; TSMC began mass-producing 5 nm chips in 2020. “Eliyan’s technology enables processors by allowing them to scale in performance and power to be more readily and practically manufacturable,” Farjadrad said. “The world will always need more computing power, and Eliyan is enabling a critical aspect of making sure scaling will happen for any type of high-performance computing application.” The fact that Eliyan’s tech has yet to reach market might give some would-be customers pause. But the startup has notable investors in the chip world behind it, including Intel and Micron, who alongside Cerberus and Celestra contributed to Eliyan’s $40 million Series A tranche that closed today. With the capital, Eliyan plans to continue chasing after a chiplet market that could be worth $50 billion in 2024 — specifically by ramping up testing and implementation. Farjadrad wouldn’t name clients, but said that Eliyan, which currently has a 21-person staff, is in discussions with “big semi companies, hyperscalers and AI processor startups.” “We’re dealing with the challenges and realities of physics in designing and manufacturing advanced chips … [but we’re] in a high-demand market,” Farjadrad said. “Our technology will ultimately lead to faster, more efficient and cheaper high-performance computing to run data centers, cloud computing AI, graphics and more.”