Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/chiplets-10-breakthrough-technologies-2024-mit-technology-review.19409/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Chiplets: 10 Breakthrough Technologies 2024 MIT Technology Review

Daniel Nenni

Admin
Staff member
Chipmakers are betting that smaller, more specialized chips can extend the life of Moore’s Law.

In response, manufacturers are turning to smaller, more modular “chiplets” that are designed for specific functions (such as storing data or processing signals) and can be linked together to build a system. The smaller a chip, the fewer defects it’s likely to contain, making manufacturing less expensive.

grid of chips

The US government aims to push microelectronics research forward. But maintaining competitiveness in the long term will require embracing uncertainty.

Companies including Advanced Micro Devices and Intel have been marketing systems based on chiplets for years. But whether chiplets can help the industry maintain performance gains at the pace of Moore’s Law will depend on packaging, which entails placing them side by side or stacking them, forming fast, high-bandwidth electrical connections between them, and encasing them in protective plastic.

Manufacturers are still working out the best way to balance cost with performance. The $52.7 billion CHIPS Act, the 2022 US legislation aimed at shoring up the nation’s chip industry, directs $11 billion toward “advanced semiconductor” research and creates a National Advanced Packaging Manufacturing Program to foster collaboration between academia and industry.

So far, chiplet adoption has been hindered by the lack of technical standards for packaging. That’s changing: the industry has embraced an open-source standard called Universal Chiplet Interconnect Express. In theory, standards will make it easy to combine chiplets made by different companies, which could give chipmakers more freedom in fast-moving fields like AI, aerospace, and automaking.

 
The MIT Technology Review articles have mostly disappointed me. This one does too. For one thing, UCIe is not an "open-source standard". UCIe is an industry specification promoted and evolved by a consortium of companies, and the consortium was founded by Intel. The specification, once approved and released, can be downloaded by non-members, so it is open to examination, but contributors to the specifications need to be members of the UCIe consortium, which has a membership fee, an approval process, and IP disclosure and licensing requirements. You must be a member of the consortium to be eligible for RAND IP licensing. So the specification is not "open source" by common industry definitions.
 
The MIT Technology Review articles have mostly disappointed me. This one does too. For one thing, UCIe is not an "open-source standard". UCIe is an industry specification promoted and evolved by a consortium of companies, and the consortium was founded by Intel. The specification, once approved and released, can be downloaded by non-members, so it is open to examination, but contributors to the specifications need to be members of the UCIe consortium, which has a membership fee, an approval process, and IP disclosure and licensing requirements. You must be a member of the consortium to be eligible for RAND IP licensing. So the specification is not "open source" by common industry definitions.

Agreed, it is an open standard not an open-source standard. Says it right on the website:

This new open industry standard establishes a universal interconnect at the package-level.

I'm just happy that semiconductor technology is popular now. After 40 years of semiconductor work I say it is about time. Even if they get it wrong. MIT Review should know better however. The author is an East Coast journalist, the worst kind of technology journalist. :cool:
 
Back
Top