Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
Intel Hardware

Overclocker Breaks CPU Frequency World Record with Intel's Raptor Lake Core i9-13900K (tomshardware.com) 50

Posted by EditorDavid from the in-the-chips dept.
Hardcore overclocker Elmor "officially broke the CPU frequency world record with Intel's brand-new Core i9-13900K 24-core processor," reports Tom's Hardware — by hitting "a staggering 8.812GHz using liquid nitrogen cooling, dethroning the 8-year reigning champion, the FX-8370, by 90MHz." That's right; it took eight years for a new CPU architecture to dethrone AMD's FX series processors. Those chips are infamous for their mediocre CPU performance at launch; however, these chips scaled incredibly well under liquid nitrogen overclocking....

Elmor accomplished this monumental feat thanks to Intel's new highly-clocked 13th Gen Raptor Lake CPU architecture. Out of the box, the Core i9-13900K can run over 5.5GHz on all P-cores while also hitting 5.8GHz under lightly threaded workloads. The 13900K is, by far, Intel's highest-clocking chip to date.
This discussion has been archived. No new comments can be posted.

Overclocker Breaks CPU Frequency World Record with Intel's Raptor Lake Core i9-13900K More

Overclocker Breaks CPU Frequency World Record with Intel's Raptor Lake Core i9-13900K

Comments Filter:
  • but nobody was making software for it. It wasn't until the PS4 & XBone forced game developers to come to terms with multi-threaded programming that we started to really see applications use it.

    The FX-8350 holds up better than the i7-2600 with newer games by a wide margin. But the i7 spanks the 8350 for any game around at the time. The only trouble is there's not a lot of reason to use a CPU that old these days since a $200 upgrade will outperform both of them.

    AMD gambled that software would come

    • Re:The 8300 serious had amazing multi-core (Score:4, Informative)

      by DamnOregonian ( 963763 ) on Saturday October 22, 2022 @01:58PM (#62988495)
      Amazing multi-core?

      The 8300 was a Bulldozer, which means it was a 4-core part marketed as an 8-core part, because it had 2 integer decode units in each core.
      Get the fuck out of here. It means there were some very bizarre niche workloads it was good at (great for certain synthetic benchmarks) while it was abysmal for the rest. It performed worse than HT in nearly all instances, which is why HT is used today rather than this boneheaded regime.

      The FX-8350 holds up better than the i7-2600 with newer games by a wide margin.

      Citation needed.

      • You're mostly correct. HT (SMT) wasn't that much better than AMD's CMT approach, it's just that the underlying core that AMD released was not that good. HT rarely added more than 30% performance advantage to Haswell, for example, while CMT provided 40-45% int performance to Bulldozer and Piledriver CPUs in scenarios where it could be used. HT was more-consistently useful, and the performance in 4t scenarios was so much better on Haswell that Piledriver didn't stand a chance, even in workloads that could

        • HT (SMT) wasn't that much better than AMD's CMT approach

          Yes, it absolutely was.

          it's just that the underlying core that AMD released was not that good.

          The core itself was also bad. It did have the benefit of being friendly to high clock speeds, though.

          In a traditional (and current) hyperthreaded core, execution resources are split out after the dispatcher in a fine-grained fashion.
          This allows for a hypothetical 100% usage of the execution resources of a single core. What this equates out to in terms of percentage performance increase isn't terribly relevant- what is important is that you're getting full use of your execution reso

          • Again, HT added maybe 30% to a Haswell core's throughput moving from 1t -> 2t. CMT added ~45% to a Piledriver module moving from 1t to 2t (on int workloads). CMT had its advantages. Piledriver modules just sucked, so adding 45% throughput wasn't good enough.

            • HT adds 30% throughput to a single core.
              CMT adds nothing to a single core, because they're "2 cores" (but it's really not)
              If bulldozer had been marketed as a 4 core part, then you could say, ok, it's smiliar to HT, just worse in nearly all workloads.

              It's an error to compare SMT with CMT.
              CMT is doubled up dysfunctional cores. It's a broken version of SMP, where your individual cores are only really half cores with shared resources.

              Again, again:
              In a CMT configuration, the most you can ever use of a "m
            • There has been a weird movement over time to try to paint the Bulldozer/CMT era as something other than a shitty marketing gimmick designed to rescue falling market share with false advertising, but that's what it was.

              Read this. [wordpress.com]
              The "use" of CMT wasn't to the user, it was to AMD. They needed a marketing W.
      • No serious, look up videos on YouTube of people comparing the 8350 to the i7 equivalent in more modern titles. Titles that make heavy use of multicore performance do much better on the 8350.

        The problem is just about every PC game back then was written to be ported to the Xbox 360 and the Xbox 360 had a lot more single thread performance then it's GPU needed so practically all games or written with single thread performance in mind.

        But if you happened to have a proper multi-threaded workload back whe

        • No serious, look up videos on YouTube of people comparing the 8350 to the i7 equivalent in more modern titles. Titles that make heavy use of multicore performance do much better on the 8350.

          Citation needed.

          Either way, I'll explain how it technically works for you.
          The Bulldozer line of chips were marketed as 8-core units.
          And for certain instruction streams- that wasn't too misleading.
          However, for most instruction streams, that was misleading.
          So your 8-core FX8350 could, on very integer-heavy workloads- perform like an 8 core part. Which made it literally 2 i7-2600s.
          But for most workloads, it performed as a 4-core part, which was less than an i7-2600, which got 4-cores +~30% via SMT for n

    • Re: (Score:2)

      by fazig ( 2909523 )
      Care to share the data that compares the FX-8350 to an i7-2600 in modern games?
      • Go to YouTube and search around for videos about the i7 and 8350 on recent games. It shouldn't take much to find the video but I'm on a phone so it's a pain to link it here. Basically modern games that make heavy use of multi-core and multi-thread support hold up much better on the 8350 than the equivalent i7.

        This isn't to say that you would actually want to use an 8350 with modern games. It would still heavily bottleneck them. But it gives you an idea of what AMD thought the future was going to be. The

  • Nice but.. (Score:3)

    by Z80a ( 971949 ) on Saturday October 22, 2022 @02:41PM (#62988595)

    That's probably the last gasp of the monolithic designs.
    At some point, all the companies will move to tiles/chiplets and clocks will go down, but chip sizes will go up quite a lot, maybe even literally up.
    With the freedom of assembling massive "total area chips", things like a a very wide, 16 instruction per clock cycle pipelines will be pretty viable.

  • Interesting but not useful. (Score:3)

    by Chas ( 5144 ) on Saturday October 22, 2022 @09:42PM (#62989467) Homepage Journal

    We gots us a gazillion yottahertz!

    With only one core!

    With most of the features turned off.

    And we have to have a dedicated team pouring in LN2 or it goes thermonuclear!

    And the one core actually eats the entire package's power budget!

    So you get a nice, single-threaded job done very fast.

    Or you can actually USE the system.

    From a purely data modeling, it's interesting that you got "A NUMBER".

    In the end, it's like snail racing.

    • Re: (Score:2)

      by Z80a ( 971949 )

      It's probably pretty good to run crysis

    • Yeah, all those extreme overclocks are pretty much not useful for actually using to do any work.
      It's just a dick waving contest that means nothing in the end.

      I just read the summay, so it doesnt even say how much was the power draw at that speed. Maybe someone will use liquid helium or something for a few more Mhz.

  • Unless you have work at a hospital for lab that has a supply of liquid nitrogen, it's just a pointless thought experiment. Alcohol and dry ice is be slightly more easily available, for the because they ship food items in dry ice, and also you can make dry ice from a CO2 tank which can be stored. It would get you -90C (-130F), not as cold, but sometimes, Safeway has everything you need.

Slashdot Top Deals

Somebody ought to cross ball point pens with coat hangers so that the pens will multiply instead of disappear.

Close