Full Transcript
Loading transcript…
CASPA2022 SIGII Advantest SonnyBawari
Watch with subtitles, summary & AI chat
Add the free Subkun extension — works directly on YouTube.
- Watch
- Subtitles
- Summary
- Ask AI
Semiconductor engineers, data scientists, and industry professionals interested in AI, edge computing, and advanced packaging solutions.
TL;DR
This presentation introduces Advantest's approach to integrating disparate data sources in semiconductor manufacturing, akin to overcoming the Tower of Babel. By leveraging edge computing and AI/ML, they aim to enhance yields and quality in advanced packaging for chiplets, moving beyond point solutions to a holistic data integration strategy.
Key Takeaways
- The semiconductor industry faces challenges similar to the Tower of Babel, with fragmented data, diverse formats, and global distribution hindering collaboration and insight generation.
- Integrating data across the entire semiconductor lifecycle, from chip design to final packaging, is crucial for improving performance and supply chain efficiency.
- Point solutions are insufficient for modern semiconductor challenges; a holistic, horizontally integrated approach is necessary to avoid surprises and achieve desired outcomes.
- Edge computing, by offloading intensive computations to dedicated servers, enables near real-time data processing and faster algorithm execution in semiconductor testing.
- Leveraging machine learning models directly in their native languages on edge servers allows data scientists to work independently of test programming languages.
- This edge computing approach enhances security by creating a zero-trust environment, isolating sensitive computations from potential operator access on test equipment.
- Advanced packaging for chiplets presents significant challenges, requiring new methods to ensure overall system performance when individual components cannot be tested in isolation.
In This Video
- 00:00Introduction to Sonny Bawari
Sonny Bawari, VP of Advantest Cloud Solutions, has 30 years of experience in semiconductor leadership.
- 01:08Speech Topic: Hyper Edge Computing
Bawari's speech focuses on hyper edge computing, AI, and machine learning driving yields in advanced packaging.
- 02:06Advantest: A Global Leader
Advantest is a global leader in the ATE industry with 30,000 systems installed worldwide.
- 03:26The Tower of Babel Analogy
The Tower of Babel illustrates challenges in semiconductor manufacturing due to dispersed data and different languages.
- 05:13Vision: Integrating Semiconductor Data
The vision is to integrate all semiconductor data, from design to test, for better analysis and decision-making.
- 06:48Test Ecosystem Challenges
Challenges in the test ecosystem include automation needs, complexity of chiplets, and thermal variations.
- 08:21Edge Computing Solution
An inference server offloads computation to the edge, enabling near real-time performance and faster ML model execution.
Questions & Answers
What is Advent Cloud Solutions and who is Sonny Bawari?
Advent Cloud Solutions is part of Advantest Corporation. Sonny Bawari is the VP of Advent Cloud Solutions, with 30 years of international leadership experience in semiconductor manufacturing and R&D validation.
What is the main challenge in semiconductor manufacturing today?
The main challenge is integrating data from globally dispersed manufacturing processes, different data formats, and varying languages, making it difficult to analyze and improve performance.
How does Advent Cloud Solutions address data integration challenges?
They aim to horizontally integrate all data from chip design, fab processing, and test insertions to enable analysis and decision-making similar to how tech giants use data.
What is edge computing in the context of semiconductor manufacturing?
Edge computing involves using an inference server with GPUs connected to the tester to offload complex computations, allowing for near real-time analysis and faster algorithm execution.
What are the benefits of using edge computing with an inference server?
It enables faster computation, allows data scientists to use native machine learning languages, improves security with a zero-trust environment, and separates the work of data scientists and test programmers.
What is the significance of chiplets in advanced packaging?
Chiplets are individual semiconductor dies that come together on a complex package. Ensuring the entire package works together and meets performance needs is a challenge, especially without testing each component individually.
Key Terms
- Chiplets — Individual semiconductor dies that are combined on a single package to form a larger system.
- Edge Computing — Processing data closer to the source of generation, in this case, using an inference server connected to a tester for real-time analysis.
- Inference Server — A server equipped with GPUs designed to run machine learning models and perform complex computations, offloaded from the main tester.