Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/what-processors-and-memory-will-dominate-ai-ml.14475/
        )

    [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] => 2021370
            [XFI] => 1050270
        )

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

What processors and memory will dominate AI/ML?

Arthur Hanson

Well-known member
What are the opinions and views of the SemiWiki community on this area, which I feel will become a dominant factor in just about everything with whole new ecosystems most can't even imagine in just a few years? I also feel AI/ML will be data center-based and be sold as a service, with a few companies dominating. Any thoughts, views, and opinions on where these technologies will go and what companies will dominate will be appreciated. I have no doubt TSM, AMD, Nvidia and ARM will be among the many key companies in this ecosystem. Thanks
 
My guess is that domain specific chips will dominate AI/ML applications. Google, Amazon, and Facebook are making their own chips as are many other companies. I think most chips in the future will have AI integrated into it in some manner. TSMC will be a big benefactor of course since AI requires performance (leading edge process nodes) and AI will increase die sizes (less chips per wafers).

Right now AI chips are dominating the VC funding world. If you are raising money for a chip company you had better have AI in there somewhere, absolutely.
 
What are the opinions and views of the SemiWiki community on this area, which I feel will become a dominant factor in just about everything with whole new ecosystems most can't even imagine in just a few years? I also feel AI/ML will be data center-based and be sold as a service, with a few companies dominating. Any thoughts, views, and opinions on where these technologies will go and what companies will dominate will be appreciated. I have no doubt TSM, AMD, Nvidia and ARM will be among the many key companies in this ecosystem. Thanks
I see that in a few years, people will learn how to do these neural algos without GPUs, or ASICs.

The sparse matrix approach on CPUs have completely beaten the first gen "AI" ASICs.


This is a complete death sentence to anybody except ones on the bleeding edge. Nobody will pay for a not only more expensive, but also subpar way of doing things.

I remember a time when people were saying that real time audio was impossible without dedicated hardware. But just few years of computer science advances made for thousandfold reduction in latency, and CPU use by audio processing.
 
Back
Top