With all the data that needs to be processed, stored, accessed, and rewritten for advanced technology applications like artificial intelligence (AI), High Bandwidth Memory (HBM) is the key that drives these innovations forward.
To learn more, I spoke with Aaron Fellis, general manager at Lam Research, about advanced packaging breakthroughs that will enable HBM to power the AI era.
He explained that HBM is ideal for the high processing speeds demanded by AI because it allows for faster coupling between the processing unit and memory for both access and storage. Placing a larger memory device closer to the processing chip reduces latency, improves the speed of the entire system of devices, and reduces power usage and the amount of heat generated.
So, as the need for faster data access grows, HBM stacks will continue to rise in popularity, which brings a greater need for through-silicon vias (TSVs). As Fellis discussed, these TSV structures, or the vertical wires that connect stacks of chips, are critical for the future of AI and other advanced applications.
HBM chips are typically only used for applications that have robust demands that can’t be met any other way. But, as more data-intensive applications that need TSVs reach the mainstream, HBM should become more widely adopted.
After we discussed the importance of TSVs, Fellis went into detail about how Lam Research is investing in novel processes to enable advanced packaging for HBM. In some cases, this involves the construction of a tall, complex TSV structure, and in others, it involves the metallization of that structure. With this in mind, Fellis explained how achieving high aspect ratios (HAR) by means of etching and filling TSVs is essential for scaling to the demands of advanced technologies (Figure 1).
I learned that because advanced applications require more TSVs in their chip stacks to increase interconnect density, TSV openings must become smaller. When this happens, the relative height to that opening (aspect ratio) increases, which makes the fill process that much more challenging. Proper filling of TSV structures is essential to stack HBM chips.
As Fellis highlighted, Lam mitigates this issue through its Syndion® product family. Syndion enables highly precise etch formation of the TSV by coupling proprietary etch technology with a unique in-situ protective coating process. This allows the creation of precise TSV structures with the HARs that are needed to scale the size of TSVs. Lam’s SABRE® 3D copper electroplating tool further alleviates TSV problems by depositing a void-free fill. Together, Syndion and SABRE 3D deliver precision and accuracy in advanced packaging, driving yield improvement and cost efficiency. According to Fellis, they are currently used by all leading memory makers.
Leveraging its expertise in etch, deposition and clean processes, Fellis noted that Lam continually collaborates with chipmakers to address scaling challenges.
“We work very closely with our customers to understand where they’ll scale the aspect ratios and opening size of the TSVs, so we can adequately fill them with our copper interconnect films,” Fellis said. “It’s all about doing so in the most efficient and cost-optimized manner so we can deliver on our customers’ technical requirements with the fastest processing speed. This enables them to benefit from the most efficient use of capital equipment in their factory and accelerate their ability to deliver their HBM product roadmap.”
You can learn more about Lam’s industry-leading tools for next-generation of advanced chipmaking here.
Feature photo: : Image of Lam’s Syndion® etch tool. Syndion and Lam’s SABRE® 3D copper electroplating tool are used in the advanced manufacturing of high bandwidth memory by leading memory providers.
