Core PCB design skills driving success in high-speed electrical engineering include differential pair routing, controlled impedance design, and precise skew management. Engineers must master signal integrity principles to ensure accurate propagation with minimal degradation, especially given fast edge rates and high-frequency operation. Simulation-driven verification is critical for validating performance before production. Additionally, proficiency in building optimized PCB stack-ups—balancing signal, power, and ground layers—is essential for managing electromagnetic interference. A solid grasp of basic electronics, including calculating impedance and power distribution, forms the foundation. Finally, attention to detail and strong problem-solving skills are vital for addressing thermal and manufacturability challenges effectively.

Modern PCB design relies on ECAD tools like Altium Designer, Cadence Allegro, and KiCad for essential workflows including schematic capture, PCB layout, trace routing, and design rule checking. Key methodologies include optimizing component placement for signal integrity, defining layer stack-ups with impedance control, and performing 3D visualization to prevent physical conflicts with enclosures. Advanced practices involve constraint management, ERC checks, and MCAD integration for seamless collaboration with mechanical teams. These tools generate Gerber files, drill files, and BOMs for fabrication, ensuring designs meet high-frequency requirements and manufacturing constraints while enabling rapid iteration and error reduction. [1][2][3][5]

Current challenges in high-speed multi-layer PCB design for AI hardware include maintaining signal integrity at extreme bandwidths (often >28 Gbps) while managing crosstalk and impedance control across 16–32 layers. Engineers must also address severe thermal management issues, as AI accelerators dissipate 350–700W, requiring robust power distribution networks with ultra-low impedance. Additionally, escape routing from large BGA packages demands advanced HDI techniques like microvias and sequential lamination, since traditional vias are insufficient. Ensuring EMI/EMC compliance and minimizing signal degradation through back-drilling to remove stubs are critical. Finally, selecting low-loss materials (e.g., Megtron 6/7) compatible with high-volume fabrication remains essential for reliable performance. [1][3][4][5]

Tenstorrent integrates cross-disciplinary collaboration in PCB projects by requiring engineers to partner with cross-functional teams spanning electrical, mechanical, and thermal domains to ensure designs meet performance, thermal, and manufacturability goals [7]. Engineers act as collaborators who work closely with architecture, RTL, packaging, and software teams to optimize system-level implementations [1][3]. The company emphasizes working across disciplines from concept to production, driving innovation collaboratively while coordinating with ODMs and manufacturing partners for seamless handoffs [1][2]. Additionally, engineers engage in design reviews and shared problem-solving with ASIC packaging teams and internal layout groups to refine signal and power integrity [5][7]. This approach ensures reliable, high-density PCBs that push performance boundaries.

90–120 words: 108

The provided job description and search results do not specify unique design innovations Tenstorrent employs for its AI CPU PCBs. The content emphasizes that Tenstorrent designs high-speed, multi-layer PCBs supporting technologies like 100/400/800 GbE, GDDR6/7, and PCIe Gen5/6, but does not detail proprietary PCB innovations. Instead, Tenstorrent’s key innovations lie in its RISC-V CPU architecture and AI IP integration, including chiplet and heterogeneous compute strategies for combining RISC-V CPUs with AI cores [2][10]. PCB-related work focuses on implementing these advanced architectures rather than introducing novel PCB design methods.