Originally Aired - Thursday, January 30 • 11:15 AM - 12:00 PM Pacific Time (US & Canada)
Create or Log in to My Show Planner to see Videos and Resources.
Your account does not have access to this session.
{{chatBodyContent}}
Be sure to come back after the session starts to have access to session resources.
Location: Ballroom G
Title: Genetic Algorithm-Driven IBIS-AMI Optimization for Robust 200G SerDes Design
Session Handouts Available Upon Speaker Approval: 1
Description:
Copper connectivity is increasingly becoming a bottleneck in the rapidly evolving compute landscape that supports artificial intelligence (AI). The design of 200G SerDes faces aggressive timelines, necessitating a robust pre-silicon simulation methodology. A significant performance limiter in SerDes technology is the adaptation routine that determines the optimal equalization (EQ) for a given channel. The effectiveness of this routine, whether in laboratory settings or simulations, heavily depends on the robustness of the underlying algorithm. Traditional simulation models often fall short of accurately representing hardware due to runtime constraints and the necessity of understanding the channel response for EQ configuration searching. Consequently, SerDes electrical models expend considerable effort on this optimization task. This paper proposes enhancements to existing modeling techniques to enable a faster and more accurate representation of SerDes, thus narrowing the gap between model abstraction and actual silicon/firmware performance. By adopting proven optimization strategies, such as Genetic Algorithms, over traditional full-search methods for driving adaptation, we aim to improve the adaptation cost function. This approach promises to enhance performance prediction accuracy across a broader range of channels for the next generation of copper links, offering significant advancements in the field.
Type: Technical Paper Session
Create or Log in to My Show Planner to see Notes
