Purported Sandy Bridge sample pushed to 5.5GHz on air

An overclocker by the name of JCornell has purportedly gotten his hands on a Sandy Bridge CPU and overclocked the bejeezus out of it. The CPU in question, a Core i7-2600K, appears to have a peak turbo speed of 3.4GHz. JCornell takes the chip up to an impressive 5.5GHz, which is good enough to shorten its SuperPi calculation time from just over 10 seconds to a little less than 7.5.

According to SemiAccurate, the overclock was done with air cooling and less than 1.4V, making Sandy Bridge’s potential all the more intriguing. Of course, you’ll need a K-series CPU with an unlocked upper multiplier to get one of Intel’s upcoming CPUs running at higher than stock speeds. Sandy Bridge prevents motherboards from increasing the base clock speed to push the CPU clock, leaving K-series chips as the only option for overclockers.

Enterprising motherboard makers have found ways around Intel’s overclocking roadblocks in the past, but none have been successful with Sandy Bridge—at least not yet. History tells us that someone will eventually find a way to circumvent the locked base clock, though. Until that happens, we can only hope that K-series Sandy Bridge CPUs maintain the reasonable prices we’ve seen from Lynnfield-derived members of the same family.

Comments closed
    • ronch
    • 9 years ago

    Ok, I know someone’s bound to ask this sick question, so whoever he is, I’ll beat him to the punch…

    Will it run Crysis?

    So there.

    • Krogoth
    • 9 years ago

    I am not impressed until that chip can handle that speed with 24 hours+ of Prime95’s torture test.

    • eitje
    • 9 years ago

    Well, that’s certainly not as good as the 10 GHz I can get with a Pentium 4.

    • christopher3393
    • 9 years ago

    FYI, I think the turbo speed is 3.8 on this part.

    • sweatshopking
    • 9 years ago

    you got a camera in my room? seriously, that’s creepy.

      • UberGerbil
      • 9 years ago

      I am Google StreetView.

    • khands
    • 9 years ago

    Edit: meant to be a reply, sorry.

    • paulWTAMU
    • 9 years ago

    I know the ghz race isn’t everything…but the thought of a 5ghz 4 or 6 core processor makes me a bit happy in the pants department 😉

    • sschaem
    • 9 years ago

    “AMD VP hints at Llano graphics performance”

    Microsoft showed that same GPU demo and it reach 722 gflops on a 5870.
    So right now llano is 22 time slower then a 5870, matching a HD 5450.

    And in the demo the CPU was mostly using L1 cache. That number could drop in a real game where the CPU will use main memory (shared with the GPU) With a HD 5450, the GPU got a dedicated memory bus.

    At this time llano point to have exactly the equivalent of a HD 5450 with ddr3

    I have no idea why Fudzilla think its 5 time faster then the demo shows…

      • stmok
      • 9 years ago

      l[

      • willmore
      • 9 years ago

      Were we reading the same article?

    • Spotpuff
    • 9 years ago

    Doesn’t the CPU start to starve for memory bandwidth the higher it clocks? Or does the integrated memory controller clock faster too to reduce latency? Are the processors even using all the bandwidth available?

      • khands
      • 9 years ago

      Depends on how you OC it, probably memory starved at those clocks though, given the inherent difficulty with raising the base clocks on SB.

      • UberGerbil
      • 9 years ago

      Which is why they use SuperPi for crap like this, because it doesn’t use memory — it runs entirely out of L1.

      SuperPi is a terrible benchmark for anything other than epeen pissing matches. It’s like those ridiculous “db drag racing” competitions: the winners of those can’t even be used as actual car stereos, because all they are designed to do is move a db meter to a maximum value. SuperPi uses very little of the CPU, so it’s perfect for these kinds of stunts because you’re more likely to be able to claim a victory when you don’t actually have to have everything working at that speed. The results are fun numbers to throw around on the intratubes, but they’re not proof of anything (much less fully-functioning CPUs at those clockspeeds — I always have to laugh when someone suggests running SuperPi as some kind of test of a stable system.)

        • paulWTAMU
        • 9 years ago

        aaaw 🙁 You just clamped down my dreams

    • sweatshopking
    • 9 years ago

    well, I don’t think we’re going to be seeing any bulldozer cores hitting 5ghz on air. Yes, I know it’s supposed to be a “higher clocked chip”, but I simply don’t think it’s going to be as fast. It sucks, because I like amd more than intel, but I don’t know if they’re able to bring the same game they’re bringing on the gpu front.

      • sschaem
      • 9 years ago

      Dont we have the 40nm 6core AMD X6 1090T chip overclock to 4ghz on air?

      Is it crazy to think with new manufacturing, 4 core at 32nm, AMD cant reach a 25% faster clock ?

        • Voldenuit
        • 9 years ago

        And IBM’s POWER7 runs at 4.25 GHz stock on 45 nm.

        In fact, POWER6 had a max stock clock of 5 GHz, and IBM had prototypes running up to 6 GHz. And that was at 65 nm!

        Not dismissing SB’s prowess and headroom, but maybe the clockspeed stagnation on desktop CPUs is as much due to market factors as it is to engineering challenges and/or process limitations? With no real threat from AMD, intel may have simply not seen the need to drive up its costs by ramping clockspeeds and deal with lower yields and more elaborate cooling mechanisms

          • khands
          • 9 years ago

          I find this scenario far too reasonable to be real.

        • Voldenuit
        • 9 years ago

        Double post.

      • OneArmedScissor
      • 9 years ago

      I really don’t see how that matters. Things like memory latency from design to design have been much more important than clock speed variations for a while now, and it’s just going to become even more important as more and more tweaked designs show up targeting very specific applications.

      You have to remember that when you say “Sandy Bridge,” you are generalizing, as Intel is using significant variations of that one architecture to hit most every market. That can mean all sorts of things now. Look how much they bastardized Nehalem with the Westmere dual-cores – they DETACHED the memory controller and didn’t even shrink it to the same node as the CPU.

      This particular iteration of Sandy Bridge has already been shown to be pretty much Lynnfield with a few hundred MHz faster cache and a GPU. It doesn’t even run higher clock speeds, despite having moved to a new manufacturing node that’s been in use for over a year.

      Bulldozer vs. the Xeon/high end Sandy Bridge is still very much a game of wait and see.

      Running a lower end CPU at 5+ GHz is sure a good way to heat up your computer and waste electricity, but that’s about it.

        • sweatshopking
        • 9 years ago

        I’d also say it adds massive epeen. If i’m telling people I’m running a 5ghz cpu, i know i’m getting some. That’s all my wife needs to hear, and she’s off to bed. as it is, my q6600’s paltry 3.0ghz quad simply won’t work anymore.

          • OneArmedScissor
          • 9 years ago

          Lol that made my day.

          • UberGerbil
          • 9 years ago

          “and she’s off to bed”… to /[

          • MadManOriginal
          • 9 years ago

          It’s not the GHz of the transistors, but the electromigration of the silicon.

            • NeelyCam
            • 9 years ago

            Now I’m interested. Tell me more!

          • NeelyCam
          • 9 years ago

          l[

        • MadManOriginal
        • 9 years ago

        Depends on what programs you’re talking about. Some things are memory latency and cache size insensitive and realy almost solely on clock speed. Media encoding comes to mind which, not coincidentally, is just about the only semi-‘mainstream’ highly demanding CPU application remaining.

          • OneArmedScissor
          • 9 years ago

          But how many highly CPU intensive encoding applications are bound by clock speed and not four cores?

          Sure, a 5.5 GHz quad-core is going to do that fast, but faster than 4 GHz-ish octal cores that are actually taking advantage of all the new architectural changes?

          The memory latency thing was just a practical example, like how games might run “faster” at a higher clock speed, but you shouldn’t actually have an issue with the CPU slowing the game down at stock speeds if it’s the right one to begin with.

          It’s always going to be something like that. Within the coming months, it doesn’t look like anything on the PC side of things is going to be overlooked anymore. You will pretty much be able to buy a combination CPU and GPU or just CPU that suits your exact use. It’s no wonder Intel is limiting overclocking now.

Pin It on Pinterest

Share This