Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Saturday December 18 2021, @08:06PM   Printer-friendly
from the no-huawei! dept.

AMD Becomes TSMC's Third Largest Customer

AMD's focus to produce all of its most advanced products at Taiwan Semiconductor Manufacturing Co. (TSMC) and increasing orders have made the company the foundry's third largest customer, according to estimates from Bloomberg and DigiTimes. Apple is still TSMC's No.1 customer and will remain so for the foreseeable future. But AMD's position ahead of Broadcom, Nvidia, and Qualcomm enables the company to negotiate better business terms, work closer with the contract maker of chips, and have an influence on development of next-generation nodes.

[...] AMD's share in TSMC's balance sheet is poised to grow as the company increases adoption of the foundry's advanced packaging technologies as well as embraces more expensive N5 for its upcoming Zen 4-based processors. Furthermore, once AMD absorbs Xilinx, it will be a considerably larger semiconductor company in general and therefore will use more of TSMC's services (and will pay more money).

[...] Bloomberg and DigiTimes estimated that Intel's input to TSMC's revenue as of December, 2021, was around 0.84% (though they do not divulge the exact period they considered as normally Intel's contribution to TSMC's earnings is significantly higher). Meanwhile, once Intel begins to use TSMC's leading-edge N3 technology (which is a rumor for now) in 2022 ~ 2023, its contribution may skyrocket all the way into the Top 3 of TSMC's clients.


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 0) by Anonymous Coward on Monday December 20 2021, @05:57AM (1 child)

    by Anonymous Coward on Monday December 20 2021, @05:57AM (#1206583)

    thx for reply.
    "logic" exits nowhere but in our minds. we can make physical constructs that emulate logic.
    for binary logic a certain amount of "gates" is required for each operator.
    in the silicon "clockwatch" realm -aka- cpu design there's no magic involved. intel and amd can not "reinvent" binary logic, tho intel somehow found a way to cut-out a transistor or two for a binary logical "add" operator with the core-line methinks (which was the BIG leg up over AMD at the time) .. or suemthing.
    there is much new tech in the "cluster" or "a-big-bunch-of-transistor" realm waiting in the wings but at the moment it's a intricate dance around the "anti-trust-mine-field" (mostly a card held by non x86 players) and "shareholder value" -aka- "milk-the-customer-for-what-it-worth" between a required(!) duopoly.
    so methinks, "faster" CPU is just more Hz and smaller distances between them gates and as mentioned "waiting in the wings" synergy (syncing) between discreet ALUs.
    the "platonic logic gates" (in silicon) are no different between AMD and intel (or "other").
    amd and intel are not fighting "to the blood", rather they have R&D ready to go but need to keep a duopoly alive and shareholders happy so no-one can just "sprint ahead" and blow up the duo-market ... *shrug*

    also, this stuff is getting old and sure you can train new talent but somebody that grew up with a mobile-phone-as-pacifier just doesn't have the "right stuff" of the old guard that was there when the seedlings where planted behind the wood-works and worried about how the wood grain would suffer during each spring and autumn storm ... now it's just "oh look a tree, let's cut it down and make furniture! i'll design it with "crap-droid-cpu-designer-5000(tm)" over lunch :)

  • (Score: 2) by takyon on Tuesday December 21 2021, @01:44PM

    by takyon (881) <takyonNO@SPAMsoylentnews.org> on Tuesday December 21 2021, @01:44PM (#1206834) Journal

    They do have to compete against each other and the likes of Arm and Apple Silicon in laptops and servers. Even desktop is not necessarily secure, as Apple is there with Mac Mini and eventually Mac Pro, and Nvidia or another player could try a push for Arm high-performance socketed chips.

    The next big thing for x86, other than big.LITTLE, will be 3D stacking, packaging, and integration. AMD's Milan-X can boost gaming by 15% and application performance by up to 80% in some workloads, simply by tripling the L3 cache per chiplet. And that's a quick and dirty implementation of 3D. Once they move to 3D monolithic SoCs with memory located nanometers/microns away from CPU cores, there's 1-3 orders of magnitude of performance and efficiency gains to be realized.

    --
    [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]