• Re: Gallium Nitride Microprocessors

    From John Savard@21:1/5 to All on Sun Jun 29 21:58:49 2025
    However, even without Gallium Nitride, high speed computing may be coming.

    According to this article

    https://www.tweaktown.com/news/106021/amds-next-gen-zen-6-chips-could- launch-with-crazy-high-7-0ghz-cpu-clock-speeds/index.html

    there are rumors floating around that succest AMD's next generation of
    Ryzen processors may be capable of burst clock speeds as high as 7 GHz.

    Naturally, unverified rumor is to be taken with a grain of salt.

    But I can think of a way in which this would be technically feasible.

    Intel, for a while, made processors with P and E cores - Performance and Efficiency.

    AMD declined to go that route, as they feel their Zen cores have the
    ability to serve a wide performance range as the voltage is adjusted.

    But to get a clock speed of 7 GHz... well, one could use a BR core
    (for "bragging rights") based on the Bulldozer design. I mean, that's
    basically how IBM (which had the opportunity to use insane cooling
    solutions) got 5 GHz in its mainframes a while back.

    No doubt AMD is not doing anything so stupid, and if it does get to 6.7
    GHz or whatever they will come by it honestly.

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From John Savard@21:1/5 to All on Sun Jun 29 21:43:34 2025
    The lead researcher at MIT for bonding GaN to silicon is Pradyot Yadav, and
    the article is:

    https://www.bisinfotech.com/mit-team-creates-low-cost-gan-on-cmos- integration-process/

    The researcher developing a GaN microprocessor for NASA is Yuji Zhao, and
    the article is

    https://phys.org/news/2017-12-gallium-nitride-processornext-generation- technology-space.html

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Kerr-Mudd, John@21:1/5 to John Savard on Thu Jul 3 09:03:54 2025
    On Thu, 3 Jul 2025 02:55:01 -0000 (UTC)
    John Savard <[email protected]d> wrote:

    On Sun, 29 Jun 2025 20:41:05 +0000, quadibloc wrote:

    In a discussion on the subject, someone noted that GaN has a high defect density. I presume that it is not as high as that of Silicon Carbide, though, or the notion of using it to build a microprocessor would be too obviously stupid for words.

    I seem to have presumed wrong.
    Apparently the defect density of Silicon Carbide is 10^4 defects per
    square centimetre... and that of Gallum Nitride is, at best, ten times as high. (At worst, throw in another factor of 10^5.)



    I know next to nothing about this, but pshurely Gallium is a lot rarer (and thus more expensive) than Silicon?

    --
    Bah, and indeed Humbug.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From John Savard@21:1/5 to quadibloc on Thu Jul 3 02:55:01 2025
    On Sun, 29 Jun 2025 20:41:05 +0000, quadibloc wrote:

    In a discussion on the subject, someone noted that GaN has a high defect density. I presume that it is not as high as that of Silicon Carbide,
    though, or the notion of using it to build a microprocessor would be too obviously stupid for words.

    I seem to have presumed wrong.
    Apparently the defect density of Silicon Carbide is 10^4 defects per
    square centimetre... and that of Gallum Nitride is, at best, ten times as
    high. (At worst, throw in another factor of 10^5.)

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From David Brown@21:1/5 to John on Thu Jul 3 16:45:48 2025
    On 03/07/2025 10:03, Kerr-Mudd, John wrote:
    On Thu, 3 Jul 2025 02:55:01 -0000 (UTC)
    John Savard <[email protected]d> wrote:

    On Sun, 29 Jun 2025 20:41:05 +0000, quadibloc wrote:

    In a discussion on the subject, someone noted that GaN has a high defect >>> density. I presume that it is not as high as that of Silicon Carbide,
    though, or the notion of using it to build a microprocessor would be too >>> obviously stupid for words.

    I seem to have presumed wrong.
    Apparently the defect density of Silicon Carbide is 10^4 defects per
    square centimetre... and that of Gallum Nitride is, at best, ten times as
    high. (At worst, throw in another factor of 10^5.)



    I know next to nothing about this, but pshurely Gallium is a lot rarer (and thus more expensive) than Silicon?


    As an element, yes, it is much more expensive. I bought some gallium
    metal, because it is fun to play with - but it was not a cheap toy!
    However, the cost of making devices from a particular element or
    molecule is vastly more complicated than the rarity of the raw material.
    You can pick up a rock from your garden and you'll have enough silicon
    atoms for a wafer, but only traces of gallium. The cost of a silicon
    wafer (even before putting anything on it) will exceed the worth of its
    weight in gallium by a factor of 20 or so. It's the purification and
    growing virtually defect-free crystals that costs money - the cost of
    the raw material elements is almost negligible.

    GaN is seeing increasing use in power electronics - a GaN power FET can
    switch a lot faster, has lower resistance, and is physically smaller
    than equivalent Si transistors. But they cost several times as much,
    largely because of the economies of scale for making them. (Power
    electronics is a lot more tolerant of defects than CPUs, because the
    geometries are vastly bigger.)

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Michael S@21:1/5 to David Brown on Thu Jul 3 22:59:27 2025
    On Thu, 3 Jul 2025 16:45:48 +0200
    David Brown <[email protected]> wrote:


    GaN is seeing increasing use in power electronics - a GaN power FET
    can switch a lot faster, has lower resistance, and is physically
    smaller than equivalent Si transistors. But they cost several times
    as much, largely because of the economies of scale for making them.
    (Power electronics is a lot more tolerant of defects than CPUs,
    because the geometries are vastly bigger.)


    My impression is that GanFET advantages are mostly at high voltages.
    Above 100V or something like that.
    At lower voltages at any given current rating best available GaNFETs
    tend to have higher capacitance than best MOSFETs.
    Also from 1st hand experience it appears that GanFETs needs more
    babysitting than MOSFETs.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Stefan Monnier@21:1/5 to All on Thu Jul 3 16:09:54 2025
    My impression is that GanFET advantages are mostly at high voltages.
    Above 100V or something like that.

    That seems sufficient in practice: most of the USB-C power bricks I see
    on the market (other than the very cheap ones) seem to use GaN nowadays,
    and that seems to allow noticeably more compact bricks.


    Stefan

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From David Brown@21:1/5 to Michael S on Fri Jul 4 09:26:59 2025
    On 03/07/2025 21:59, Michael S wrote:
    On Thu, 3 Jul 2025 16:45:48 +0200
    David Brown <[email protected]> wrote:


    GaN is seeing increasing use in power electronics - a GaN power FET
    can switch a lot faster, has lower resistance, and is physically
    smaller than equivalent Si transistors. But they cost several times
    as much, largely because of the economies of scale for making them.
    (Power electronics is a lot more tolerant of defects than CPUs,
    because the geometries are vastly bigger.)


    My impression is that GanFET advantages are mostly at high voltages.
    Above 100V or something like that.
    At lower voltages at any given current rating best available GaNFETs
    tend to have higher capacitance than best MOSFETs.
    Also from 1st hand experience it appears that GanFETs needs more
    babysitting than MOSFETs.


    They do have advantages at higher voltages, yes - but they also have
    advantages at low voltages, especially switching speeds and size. You
    are also correct that they need more babysitting - they need more
    advanced drivers, and protecting them from overcurrent is harder.

    You can now get devices that combine the drivers, protection circuits
    and GaN FET in the same package, which makes them a lot easier to use.
    I don't know the details of how these are constructed - maybe they have co-packaged silicon and GaN dies, or maybe they share a common
    substrate. It feels unlikely to me that the driver and
    monitoring/protection circuits are also GaN. My guess is co-packaging,
    but it's only a guess.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Terje Mathisen@21:1/5 to David Brown on Fri Jul 4 11:10:17 2025
    David Brown wrote:
    On 03/07/2025 10:03, Kerr-Mudd, John wrote:
    On Thu, 3 Jul 2025 02:55:01 -0000 (UTC)
    John Savard <[email protected]d> wrote:

    On Sun, 29 Jun 2025 20:41:05 +0000, quadibloc wrote:

    In a discussion on the subject, someone noted that GaN has a high
    defect
    density. I presume that it is not as high as that of Silicon Carbide,
    though, or the notion of using it to build a microprocessor would be
    too
    obviously stupid for words.

    I seem to have presumed wrong.
    Apparently the defect density of Silicon Carbide is 10^4 defects per
    square centimetre... and that of Gallum Nitride is, at best, ten
    times as
    high. (At worst, throw in another factor of 10^5.)



    I know next to nothing about this, but pshurely Gallium is a lot rarer
    (and
    thus more expensive) than Silicon?


    As an element, yes, it is much more expensive.  I bought some gallium metal, because it is fun to play with - but it was not a cheap toy!
    However, the cost of making devices from a particular element or
    molecule is vastly more complicated than the rarity of the raw material.
     You can pick up a rock from your garden and you'll have enough silicon atoms for a wafer, but only traces of gallium.   The cost of a silicon wafer (even before putting anything on it) will exceed the worth of its weight in gallium by a factor of 20 or so.  It's the purification and growing virtually defect-free crystals that costs money - the cost of
    the raw material elements is almost negligible.

    GaN is seeing increasing use in power electronics - a GaN power FET can switch a lot faster, has lower resistance, and is physically smaller
    than equivalent Si transistors.  But they cost several times as much, largely because of the economies of scale for making them.  (Power electronics is a lot more tolerant of defects than CPUs, because the geometries are vastly bigger.)


    I'd say it close to a watershed moment for power electronics, GaN power adapters have so much less loss that they don't get nearly as hot, while
    being much smaller.

    I picked up a 65W dual USB-C GaN wall wart, it happily drives my laptop
    as well as more or less everything else I have. (I have found one device
    that's incompatible with the power negotiation protocol, but that's a
    USB-C standards proliferation issue and not GaN.)

    Terje


    --
    - <Terje.Mathisen at tmsw.no>
    "almost all programming can be viewed as an exercise in caching"

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)