# Intel 野心勃勃的 18A 製程在接近量產階段面臨兩大挑戰：良率低於 10% 以及與台積電競爭的 N2 製程相比，SRAM 密度存在關鍵劣勢。

*semiconductor · news · 2024-12-05 · TechSpot*

## Key points

- Intel 的 18A 製程良率低於 10%，距離商業可行性尚遠。
- Intel 18A 的 SRAM 密度落後台積電 N2，僅達 31.8Mb/mm²，台積電為 38Mb/mm²。
- Intel 取消了面向客戶的 20A 製程，將策略與資源轉向 18A 開發。
- 若 18A 良率近期無顯著提升，Intel 2025 年產品路線圖將面臨重大風險。

Serving tech enthusiasts for over 25 years.TechSpot means tech analysis and advice you can trust In a nutshell: Intel's ambitious 18A node is grappling with two significant hurdles as it approaches production: yield rates languishing below 10 percent and a critical SRAM density disadvantage compared to TSMC's competing N2 process. These challenges could impede the node's deployment across Intel's next-generation CPU, AI, and custom chip portfolios. Recent reports indicate that Intel is facing significant yield challenges with its 18A node, potentially delaying its mass production timeline. According to the South Korean newspaper Chosun, current yield rates are below 10 percent, meaning that nearly nine out of every 10 chips manufactured are defective. This is a major issue, particularly as Intel has already canceled its 20A (2nm class) process node for Foundry customers and shifted resources to the 18A (1.8nm class) node. If the sub-10 percent yield rate proves accurate, it could render the node unsuitable for commercial production, at least until significant improvements are made. The challenge of packing transistors into increasingly dense layouts at these cutting-edge nodes is a formidable engineering hurdle affecting the entire semiconductor industry. Samsung's foundry yield for processes below 3nm is currently below 50 percent, with its Gate-All-Around (GAA) technology yield reportedly as low as 10 to 20 percent. There is, however, reason for optimism regarding Intel's 18A node, as the company still has several months to refine the process ahead of its projected 2025 production ramp. The potential payoff is significant, with 18A slated to power high-profile products such as Intel's Clearwater Forest server chips, Panther Lake mobile CPUs, and custom AI silicon. If Intel can rapidly improve 18A's yields to respectable levels – above 60 percent in the coming months – the stage could still be set for this node to drive the next generation of products from the company. That said, yield issues aren't the only challenge Intel faces with 18A. TSMC has reportedly gained an edge in another critical area: SRAM density. According to the ISSCC 2025 Advance Program, TSMC's N2 (2nm class) node shrinks high-density SRAM bit cells down to around 0.0175 μm², achieving a density of 38Mb/mm². In contrast, Intel's 18A node achieves 0.021 μm² and 31.8Mb/mm², which is closer to TSMC's previous-generation N3E and N5 nodes – a noticeable difference. As chip designs demand more SRAM, increasing the density of these tiny memory cells is vital for maintaining compact, efficient designs. This is where gate-all-around (GAA) transistors come into play. By controlling the channel on all sides, GAA transistors allow for tighter scaling compared to traditional finFETs. This tight control reduces leakage at small dimensions, enabling higher-density SRAM. Both Intel and TSMC are using GAA to shrink their SRAM bit cells, but TSMC has managed to pack them even more densely with its N2 node.

**Companies:** Intel, TSMC
**Countries:** South Korea, United States

[Read the full story on TechSpot](https://www.techspot.com/news/105844-intel-18a-node-reportedly-stuck-10-yields-sram.html)

---

Canonical: https://newsio.io/zh-TW/n/4c8b8fd6-bcd3-4bf3-acb6-d8f0c9684022/intel-18a-10-n2-sram
Summarized by Newsio from TechSpot. https://newsio.io/how-it-works