‘Phenom II X4 955’ sample surfaces, is overclocked to 4GHz

The Phenom II X4 955 exists, and it’s a pretty nice overclocker. At least, that’s the impression you might get after reading a thread at the XtremeSystems Forums, in which a member has posted CPU-Z screenshots of the unannounced AMD processor running overclocked in the Windows 7 beta.

As you might remember, we last wrote about this purported Phenom II X4 955 two weeks ago. Back then, word was that the 3.2GHz quad-core CPU would hit stores in April, and that it would be a Socket AM3 part with a 125W thermal envelope and support for DDR3-1333 memory. That chip was also said to be a replacement for a 3.1GHz Phenom II X4 950, which AMD supposedly canceled because of better-than-expected yields.

Dumo from the XtremeSystems Forums shows the 955 engineering sample running at the stock 3.2GHz speed with a 1.25V voltage setting. At 1.45V with air cooling, the poster claims he was able to hit a cool 4GHz by pushing the HyperTransport speed from 200 to 250MHz—and he even has a screenshot to prove it. That screenshot shows Dumo also overclocked his memory to 1666MHz, which netted over 11GB/s of read bandwidth according to a Lavalys Everest benchmark.

Comments closed
    • Samlind
    • 11 years ago

    Interesting at least one site is claiming 6ghz on Ln2 with a Phenom II.

    • lycium
    • 11 years ago

    > /[

    • jdaven
    • 11 years ago

    This is a little off topic. But I just went to Alienware.com to see their latest stuff and they don’t have an AMD system over a Phenom 9950. Where are all the Phenom II systems? Me thinks Dell might be holding them back.

    • UberGerbil
    • 11 years ago

    ……….

    • BoBzeBuilder
    • 11 years ago

    I’m still waiting for the Nehalem crushinator.

      • asdsa
      • 11 years ago

      I don’t see that happening for a long time. Intel is too much ahead. Maybe those arabs can give a couple of more billion to R&D department to speed things up.

        • BoBzeBuilder
        • 11 years ago

        Gas prices are down. Arab don’t have couple of more billion.

          • UberGerbil
          • 11 years ago

          Yes they do. They banked it when times were good (and even at current prices, which are a problem for some producers, oil is still profitable for the Gulf states). The Abu Dhabi Investment Authority — the largest sovereign wealth fund in the world — has $875B.

      • donkeycrock
      • 11 years ago

      do the math

      2900xt = sucky
      38xx = not bad
      48xx = kick ass

      phenom = sucky
      phenom2 = not bad
      phenom3 = ?????

        • Waco
        • 11 years ago

        Hey now, I liked my 2900XT 1 GB. 😛

          • Meadows
          • 11 years ago

          Some people liked their GeForce FX Ultras too.

        • khands
        • 11 years ago

        Sort of, the 3870’s were about on par with the 2900XT’s, only they cost a heck of a lot less to produce (and could possibly come in X2 models). If they were going to do the same to with their processors, they’d have to have their high-end stuff have twice as many cores as Intels (for lack of a better technical analogy) along with much faster cache (for the GDDR5 analogy).

        • SPOOFE
        • 11 years ago

        If the Core i7’s were merely rebadged Core 2’s, your equation would be a bit more robust. 🙂

    • adisor19
    • 11 years ago

    The new Celeron 300 ?

    Adi

      • Jigar
      • 11 years ago

      No dum dum, it’s *[

        • adisor19
        • 11 years ago

        *sigh*

        Guess you’re too young to know what i’m referring to…

        Adi

          • atryus28
          • 11 years ago

          At 32 I don’t feel that old but maybe I am. I know what you are talking about adi.

          • Vaughn
          • 11 years ago

          you are correct about that one

          So Jigar2speed5095 is probably under 18years old =)

            • Jigar
            • 11 years ago

            Dude, i am married with a 2 kids…. 😉 but, yes i still like to act like one…

      • MadManOriginal
      • 11 years ago

      Nah, it’s not going 1.5x the speed and it’s starting out as a totl CPU.

      • Drive
      • 11 years ago

      you mean Celeron *[<300a<]* right??

        • Kurkotain
        • 11 years ago

        you mean celeron 450 A right? ; )

        • mr_greedy
        • 11 years ago

        I had a pair of 300a’s cookin in my abit BP6..

        NICE.

      • Rakhmaninov3
      • 11 years ago

      haaaah I had a Celery 400 overclocked to 600, it was awesome.

      • WaltC
      • 11 years ago

      The “new” Celeron 300? Gosh, where was the “old” Celeron 300 that clocked into the GHz straight from the factory?…;)

      I guess I’m getting a little tired of seeing this long-in the-tooth comparison. Talking about moving a 200MHz cpu to 400MHz by FSB adjustments, prior to Intel locking all of its cpu multipliers, of course, seems to me a trifle different than moving 2GHz cpus to 4GHz, unlocked multipliers or no. By comparison, the Celeron 300 was such an incredibly simple cpu that it really doesn’t seem fair to compare it with anything current. But, yes, I had one, and yes, I overclocked it…;)

      The truly memorable thing about the Celeron 300 to me is that at least partially because of the competition AMD brought into the cpu markets, the amount of money we paid for a Celeron 300 in those days will today buy a cpu that will running stock murder it in performance–and whether the Celery 300 was overclocked or not would make no difference at all…;)

    • Jigar
    • 11 years ago

    Well, I hope we have a HD 48** series in making here … This time they deliver in CPU sector.

    • Anonymous Coward
    • 11 years ago

    Sure does take a long time to leave 3ghz behind.

      • Dposcorp
      • 11 years ago

      Its all about the “IPC” these days.

        • eitje
        • 11 years ago

        The other metric is Performance per Watt.

          • Meadows
          • 11 years ago

          Metric, you say. How do you divide performance by watt? How much is 1 Performance?

            • Dposcorp
            • 11 years ago

            I can’t tell if you are being sarcastic or not, but I think eitje means something like this review, where the reviewers “…..focus on CPU performance-per-dollar almost exclusively in this article, but while reading, keep in mind that there are other factors to take into account.”

            §[<https://techreport.com/articles.x/14756<]§ On this page, they take into account power efficiency per dollar spent on power in what they call a "render energy test." They look at how much energy each system used to render a test scene in Cinebench. Its take into account how much the entire system, not just CPU, uses, but most of the other parts are the same. §[<https://techreport.com/articles.x/14756/9<]§ This is a excellent article if you haven't read it, and I expect more of the same as new CPUs emerge.

            • MadManOriginal
            • 11 years ago

            I hope you’re just being silly too. Performance/watt is pretty simple: take a performance measurement and divide it by the power used to complete the test.

            • Meadows
            • 11 years ago

            That doesn’t sound very standardised. I do know what you people are trying to tell though, of course.

            • A_Pickle
            • 11 years ago

            It’s… very empirical. A task will take some amount of power to complete, and it’s easily measured. The amount of power that that task will take to complete will depend heavily on the hardware in the system.

            The hardware that completes the task in less time, and using less power… has superior performance-per-watt. The hardware that completes the task in more time, using more power… has inferior performance per watt. And since tasks are generally measured in quantifiable metrics, and since power is a quantifiable metric, it’s really fricking easy to compute performance-per-watt.

            …capische?

            • Meadows
            • 11 years ago

            What if hardware computes it in more time but using less power overall? (I know it’s not related, but it’s a crack in your specific shield)

            • no51
            • 11 years ago

            I suggest you go to the ISO and lobby for a standardized unit. In due time maybe we’ll see the “Meadow” unit as a measurement for performance/watt. Your Core i9 only does 35 mdw? Well my Phenom 6 does 39!

            • Meadows
            • 11 years ago

            Cool stuff, finally my name will be imprinted on the world. But will that gain me enough fame and money to pay for the highest MDW computer?

            • TheTechReporter
            • 11 years ago

            Ha! Mdw’s!

            But seriously, I think there is an issue here that everyone has been avoiding. Namely, which benchmark(s) do you use to measure performance? The specific benchmarks chosen have a large impact on the performance results. Also, do you use synthetic benchmarks, or only real-world benches? Do you use different, optimized versions of a benchmark (ie Intel vs AMD optimized), or a single generic one?

            In the end, mdw’s wouldn’t be any more useful than GHz.

            But, back to the original point, AMD never needed to rely on large clock speed increases before, and still doesn’t.

            • khands
            • 11 years ago

            Probably a whole performance testing suiet. Likely, however, would be a specialized Perf/Watt in a general computing area. Game Perf/Watt, Render Per/Watt, etc.

            • UberGerbil
            • 11 years ago

            You multiply total time to do the work by total power used. This is the methodology Techreport used.

            For thorough methodology:
            §[<http://www.spec.org/power_ssj2008/docs/SPECpower-Methodology.pdf<]§

            • UberGerbil
            • 11 years ago
            • DrCR
            • 11 years ago

            There’s one bench, a rastizer or something, that give the performance a a multiple times the dev’s original rig, which as a 900MHz P3 or something. So that’s one way of measuring performance.

            Maybe some of you guys will know the bench I’m referring to.

            • tfp
            • 11 years ago

            1 P/W

      • UberGerbil
      • 11 years ago

      When designing a transistor there’s a general truism that the faster it is intended to switch the leakier it is (even when you’re running it slower). This is why P4s ran so hot and power-hungry compared to other CPUs on the same process. So while we’ve moved on to better process techs to keep leakage down, it’s still true that raising the clockspeed ceiling tends to blow up the power and cooling budget. As a consequence, CPU designers are looking everywhere else they can to find performance, because they can’t just rely on raising clocks.

Pin It on Pinterest

Share This