Phenom II X4 965 slips into more modest thermal envelope

Some folks scoffed when AMD elected to introduce the Phenom II X4 965 processor back in August with a 140W thermal design power (TDP), because that pushed beyond AMD’s customary peak of 125W for desktop processors. The move wasn’t without precedent, but it signaled that AMD’s 45nm chips couldn’t reach the 3.4GHz mark without a bit of extra juice.

Now, some months later, AMD has introduced a revised version of the Phenom II X4 965 Black Edition processor (say the name in a single breath for extra credit, kids) with a 125W TDP. The thermal improvement comes hand in hand with a couple other tweaks. For one, the memory controller has been altered to support the use of four DDR3 DIMMs at 1333MHz.

Also, an enhanced halt state, C1E, is now supported in hardware. Naturally, implementing power-state switching in hardware is much faster than software, and AMD says that means power management schemes should now have “virtually no impact” on performance. AMD processors have had a bit of a rocky history on this front in recent years, with power-management features potentially compromising performance in certain situations. Sounds like they’re cleaning that up. Still, I wouldn’t expect much of a difference in regular use. We tested with power management schemes enabled in our last major CPU review, of the Intel Lynnfield processors, and we did some spot checks along the way with power management disabled, too. Generally, we found that enabling and disabling power management had very little to no effect on most tests.

The reduction in the processor’s TDP rating comes with a real reduction in power use. Here’s a quick glance at power use from the Damage Labs test bench.  These results show full-system power draw following the test methods and configurations outlined in our Lynnfield review.

There’s virtually no change to power draw at idle, but power draw under load is down by 9W versus our example of the earlier 140W version of the X4 965. That’s good, but then in this particular application, the X4 965’s performance just barely surpasses that of the Core i5-750. So the power efficiency picture isn’t altered much. Here’s the proof.

Still, moving into the 125W thermal envelope ought to allow the X4 965 to find its way into more systems from big PC makers, along with enabling broader compatibility with existing AMD-oriented mobos and CPU coolers.

AMD suggests it may have also bought itself a little more overclocking headroom with this new rev of silicon. And yes, this is an unlocked Black Edition processor. Sadly, we haven’t yet had time to test that theory with our X4 965 125W sample. I do feel a CPU comparison coming on, though. Perhaps soon.

Oh, and if you’re looking to purchase the 125W version of the X4 965, look for a part number ending in M, as in HDZ965FBK4DGM, or modest or meek. That’s the 125W model. The part number for the 140W version ends in I, as in HDZ965FBK4DGI, or incinerate or inferno. Just sayin’.

Comments closed
    • Kaleid
    • 11 years ago

    What voltage this this new version run at?

    • Welch
    • 11 years ago

    Just FYI….. These are available on Newegg now… not sure when they came in but it was very recent. $199.99

    • maroon1
    • 11 years ago

    Did anyone notice that 125w TDP version did consume the same power as 140w TDP version at idle ? Did any one notice that difference in power consumption at full load is small

    Thats barely an improvement in power consumption

      • Meadows
      • 11 years ago

      Comment #22 might have a point.

    • clone
    • 11 years ago

    I’d like to see some undervolting to see how stable they can keep the cpu.

    I had an Opteron 170 overclocked back in the day to 2800 Mhz using 1.25 volts rock stable for 8 months before I sold it……. I wonder if their is any room to move to reduce temps a little more.

    • zagortenay
    • 11 years ago

    Scott Wasson says: “AMD processors have had a bit of a rocky history on this front in recent years, with power-management features potentially compromising performance in certain situations. Sounds like they’re cleaning that up. Still, I wouldn’t expect much of a difference in regular use.”
    I own an X4 955 (with Asus Crosshair III), checked the efficiency of C&Q by running a variety of benchmarks in different conditions (default Voltage-default speed, undervolted-default speed, maximum speed with default Voltage, even in an overclocked state at 3.7 GHz.) and I could not record any significant performance degradation we can speak of. All I can say is; “C&Q technology works like a charm”. If it is working for me and not working for some others, then that is either a Bios issue or there is a mobo implementation problem.
    Then what is this? “…in recent years…” Isn’t October 2009 recent enough? I assume the above statement is either a deliberate “low blow” or unintentional “brain farting”. Ha ha!
    On the other hand the new stepping is real disappointment. Only a 9W of power reduction? Does 140 minus 9 make 125? And they waited for almost a year to release the new, er…stepping/revision. Is this the new C3 stepping? I am confused.

      • MadManOriginal
      • 11 years ago

      So when you ran benchmarks did you do the simplest thing and test with CnQ on and then with it off? Because not doing that and just running different voltages and speeds wouldn’t test whether CnQ affects performance whatsoever.

        • zagortenay
        • 11 years ago

        I ran benchmarks with both C&Q enabled and disabled. Here are some examples: (16×200=3200 MHz./VID:1.275V)

        *[

          • MadManOriginal
          • 11 years ago

          Ok, just making sure because it wasn’t 100% clear from your first post.

          • Meadows
          • 11 years ago

          C&Q makes your WinRAR suffer, and it looks like you get a longer boot time and lose 1-3% CPU power depending on the task, too.

          Not significant, but the way people are (especially WinRAR users), it’s hardly acceptable.

            • zagortenay
            • 11 years ago

            We are looking at the same picture, but apparently we are seeing different things. IMO there is a certain amount of perfomance loss in certain applications when C&Q is enabled, but I wouldn’t use the word “suffer” because the subjective feeling is quite snappy and without running benchmarks you can not tell the difference. As you have said, performance loss is only around %2-3, which is worth the energy saving. Even 3 seconds of boot delay does not bother me, but this is just me…
            My objection was against the myth of “how crappy C&Q is, how it reduces the quality of usage, you have to disable it order to OC and so on(I love using my cpu OCed to 3.6 GHz. stable as a rock with stock voltage and C&Q Enabled. I disable C&Q for extreme OCing though, not to risk stability).

            • Meadows
            • 11 years ago

            The WinRAR tests show a nearly 30-second discrepancy. That’s a 10% slowdown and noticeable to almost any hillbilly user.

      • OneArmedScissor
      • 11 years ago

      r[

        • MadManOriginal
        • 11 years ago

        It’s also an at the wall measurement so it doesn’t take in to account PSU efficiency, the real CPU power draw difference is therefore a bit less than 9W.

          • Voldenuit
          • 11 years ago

          Yup. xbitlabs measured power draw from the 12V line and found a 7.2W power difference using LinX.

          ยง[<http://www.xbitlabs.com/articles/cpu/display/phenom-ii-x4-965-c3_3.html<]ยง No doubt AMD will say that 125W is 'ACP' and we ought to measure it when the computer is running Excel, Visual Studio or Windows Screensaver. No doubt something like Cinebench is a 'power virus'. :p I've never been a fan of ACP (or intel fudging TDP, but let's leave that one alone for now), but with its 125W 965, AMD seems to have slipped from 'erroneous' to 'fraudulent'. :/

        • zagortenay
        • 11 years ago

        “…but nobody ever said Cinebench is what it’s measured by.”
        OK you got me!
        Thanks for the explanation.

      • Shining Arcanine
      • 11 years ago

      You need to account for the inefficiency of the PSU to get the actual reduction. It is more like 7 watts, assuming about 80% efficiency.

      • albundy
      • 11 years ago

      yup, i love my 955! C&Q was defaulted on my 790fx board, but i changed that when i overclocked my ram to much tighter timings. cpu temp went up a few degrees, but i’m currently running 3.4ghz with CL5 timings at a balmy 28 degrees idle.

    • comet
    • 11 years ago

    This is all terribly missing the mark.

    Anyone who’s willing to buy a processor with higher than 90W TDP obviously doesn’t care about this issue.

    As for myself, processors are fast enough nowadays that I refuse to use anything with a larger than 65W TDP. That’ll be downgraded to 45W in 1-2 years and even less when ARM kicks Intel’s ass on the low power front.

      • Krogoth
      • 11 years ago

      They do care about thermal issue.

      Less watts consumed = less heat generated = fans rev slower = quieter system.

        • Lazier_Said
        • 11 years ago

        If silence is the goal, 125 against 140 watts is a distinction without difference. Crap is crap.

          • Krogoth
          • 11 years ago

          Silence computing is an oxymoron. You are going not to find anything near it with any of performance CPU/GPU combinations.

            • flip-mode
            • 11 years ago

            Where have you been the last few years? The 5870 is near silent. The GTX 2xx cards are all near silent. My 4850 is near silent. My Ninja heatsink that’s cooling an X4 955 is just one of about 50 coolers that provide cooling at near silence. Jeez, your statement is 100% incorrect.

            • Krogoth
            • 11 years ago

            There is no performance card with reference coolers that are completely silent when loaded. They are within 40-50dB range according to TR reviews. The lack of proper air circulation can make it worse by forcing the GPU fan to rev at higher speeds. The sound level (35-40dB) at idle is quiet, but still audible to people who have a good set of ears. It is different matter if it is bothersome to the user. ๐Ÿ˜‰

            Performance CPUs are not in a better position either unless they are undervolted and underclocked.

            • flip-mode
            • 11 years ago

            Maybe my components just work better than yours. My CPU cooler is inaudible regardless of load or idle, for the stock clocked x4 955. My GPU is inaudible at idle, and inaudible at load due to the fact that just the crunch of my feet while playing a game is louder than any fan noise. Dunno. Don’t much care. Just wanted to call into question the notion that high performance computers cannot be made practically inaudible.

            • MadManOriginal
            • 11 years ago

            Your feet crunch?

            • SomeOtherGeek
            • 11 years ago

            No, he crunches feet?

            • flip-mode
            • 11 years ago

            Crunching /[

            • MadManOriginal
            • 11 years ago

            So what is it that’s crunching your feet? Time for a new desk maybe? Or time to vacuum up all the chip crumbs that are strewn beneath your desk Mr. fat-mode?

            • flip-mode
            • 11 years ago

            Krogoth… stuffed under desk… kicking him… crunching my feet…

            • Krogoth
            • 11 years ago

            Hey, I didn’t know you swing that way. ๐Ÿ˜‰

            Let me get the wax and whip ready.

            • Meadows
            • 11 years ago

            I didn’t know _[

            • Krogoth
            • 11 years ago

            Wanna join the party? I got a gag ready.

            • Meadows
            • 11 years ago

            Nay, I have other fetishes.

            • MadManOriginal
            • 11 years ago

            Kroger has a working ‘scouter’ and an orange martial arts outfit if that’s what you mean..

            • Krogoth
            • 11 years ago

            IT’S OVER 9000!

            • MadManOriginal
            • 11 years ago

            I’m not sure i[

            • Shining Arcanine
            • 11 years ago

            Silence means that it emits no noise what so ever. There is no such thing as being near silent.

            • MadManOriginal
            • 11 years ago

            q[

            • Welch
            • 11 years ago

            Wow…… Why the hell argue about semantics? By NEAR Silence.. or by being In-audible, they simply mean “THEY DONT FREAKING HEAR IT”… sure lets get out our most sensitive microphones and show that scientifically there is sound… who gives a flyingfox (sorry fox, your name is a great alternative). If the sound isn’t noticeable to the point that its annoying or interfering with the operation of you playing games, movies, surfing or whatever…. load or no load, who cares. If you care that much just buy a noise canceling or isolating headset.

            As for the OP, its a 965.. its their performance chip… who gives a damn if you pay $2.00 or $10.00 more on your electric bill. What does the extra % of power used REALLY equate to in dollars, which is what really counts.

            Crunching of the feet, I’m assuming he means his characters feet moving in the game, but who knows.. he may have someone under his desk.

            • MadManOriginal
            • 11 years ago

            I’m hoping you replied to my post because it was the last in the thread ๐Ÿ˜€ the content certainly seems more directed at #50.

            • Welch
            • 11 years ago

            Yeah it was in response to Shining Arcanine and anyone who got all up in Arms that someone would say their computer is near silent… Just continuing the posts from yours since you were the last comment on that sub-topic… no worries ๐Ÿ™‚

            • UberGerbil
            • 11 years ago

            Well, if you define silence as a binary state like “pregnant” that’s true. But silent is just a special name for zero db, and there are an infinite set of values that are near zero for any definition of zero (and any definition of “near” that is not “equal to”).

            • SomeOtherGeek
            • 11 years ago

            Hey, hey! No insulting me! ๐Ÿ˜‰ Being deaf everything is silent = golden. Go buy some ear plugs and you’ll understand.

            • Shining Arcanine
            • 11 years ago

            You obviously have never visited SPCR:

            ยง[<http://www.slientpcreview.com/<]ยง

    • gosh
    • 11 years ago

    Don’t you need new bios to get better idle power?

    • dpaus
    • 11 years ago

    “Pardon me while I slip into something a bit more comfortable…”

      • Welch
      • 11 years ago

      Great quote ahaha, reminds me of “Blazing Saddles” that is what your talking about right :D?

        • dpaus
        • 11 years ago

        I was merely talking about the new thermal envelope (he said innocently…)

        The “Pardon me while I whip this out” should be held back for the introduction of the 6-core Thuban chips ๐Ÿ™‚

    • ensoph42
    • 11 years ago

    Could someone clarify the statement “For one, the memory controller has been altered to support the use of four DDR3 DIMMs at 1333MHz.”? Did the older revision not support 4 DIMMs? Or was it that it did support 4 but only at 1066? Thanks for the help.

      • MadManOriginal
      • 11 years ago

      4 but only at 1066.

        • ensoph42
        • 11 years ago

        Fantastic. I just bought a 965 last week and I assume it’s the “Inferno” version. I’m not at home to check. Luckily I haven’t bought the memory yet but I can see my plan to go 2GBx4 at 1333 isn’t going to pan out I guess. This seems to be somewhat pertinent information. How is one supposed to know this tidbit of information without learning the hard way?

          • MadManOriginal
          • 11 years ago

          Reviews I guess ๐Ÿ˜‰ Check the paragraphs after the third picture here: ยง[<http://www.techreport.com/articles.x/16382<]ยง This probably should have been mentioned in the later Phenom II reviews but I think they referenced the original review, it's probably also noted in motherboard specs. At the same time I'm not even sure that running the memory slower makes any real difference in performance (you should check reviews that might compare that) and if you can genuinely use 8GB of RAM and aren't getting it 'just because' then more slower memory would be better.

    • flip-mode
    • 11 years ago

    For some reason I thought the Phenom II x4-965 Black Edition processor used meaningfully more power than the Phenom II x4-955 Black Edition processor, but looking at these graphs, it seems not to do so at all. Idle power consumption – the more important issue in my opinion – is identical? Heck. I had said that I saw no reason for the 965, but I’ll have to flip that around and say I see no reason for the 955.

    This could mean a 3.6 GHz 975 is not very far fetched.

      • AlvinTheNerd
      • 11 years ago

      Considering the lack of Phenom II 9xx in OEM boxes (I can’t find any in the major manufactures) I don’t think AMD is going to push up on the current steppings.

      Phenom II debuted as a high end player and at high end, 140W is considered manageable. But since they have been pushed to middle ground, 140W and even 125W really aren’t acceptable. Since they can’t take the high ground for the current time, even with 3.6Ghz, I think their focus is lowering the power and becoming a better value in the middle ground.

      • dpaus
      • 11 years ago

      Go for the gusto: a 3.6 GHz Thuban… “Damn the torpedos, Mr Sulu, full impulse power!”

      • Welch
      • 11 years ago

      Heheheh……. You…. “FLIPPING” on that huh Flip ๐Ÿ™‚

      I thought the same thing when my friend built his system with his 965. 2 weeks ago it was $245.00, higher power consumption (on paper) and as far as everything goes the 955 was the same damn chip and according to many OCer’s out there could easily hit 3.4ghz matching the 965 in speed! For more than $60.00 less it seemed like a damn good buy to go for the 955.

      Now with the 965 being a mere 195.00 and the power consumption being, ehhhh, tweaked a very small amount, and the most important part…. support for 4 sticks at 1333, I too see not as much room for a 955.

        • Lazier_Said
        • 11 years ago

        Spending $245 for a PII for a new build, you should have “FLIPPED” him in the back of the head.

          • flip-mode
          • 11 years ago

          gah, I’m a boob.

          • Welch
          • 11 years ago

          Ehhh, not really. It was a little more than one could/should expect. But he wanted their top end unlocked CPU, and wanted it now. Because he was willing to jump on it when he did, he is one of those bastards that got lucky enough to get the 5850 for 255 or 260. So all things were equal. Not to mention they had a deal for the exact Asus MB he wanted as a combo deal, 35.00 off, so it brought it close enough to the 200.00 range.

          Considering the last time he and I built a brand new system, a flagship CPU would have ran you well over 400.00 maybe even 500.00 (memory isnt serving me right at this point, was too long ago) 245.00 for a flagship is fair.

          195.00……. even better ๐Ÿ™‚

    • dragmor
    • 11 years ago

    So this is the C3 revision? NDA due to life Nov 4.

    Can you show some undervolting tests? See if this is going to allow them to bring most of their CPUs out at 65w or less?

    • Prodeous
    • 11 years ago

    Drop in the order of 9w at Load …

    I was wondering, when you do your tests, how well the new CPU will undervolt.. maybe there is still some headroom to lower the power consumption …

    Doubt they would manage a 30W drop.. but still the really need to do something to be in the 95W range…

    I know about the 95W 3Ghz part, so an extra 30W for 400Mhz.. that is quite a lot…

      • grantmeaname
      • 11 years ago

      power consumption increases with the cube of frequency… a little bump in frequency means a huge jump in power consumption once you get up around 3GHz

        • glappkaeft
        • 11 years ago

        No, all things the same, power consumption increases with the square of voltage and linear with frequancy. Although if you start tweaking the process and fiddeling with with voltages to get the last Hz out of a part all bets are of.

          • Prodeous
          • 11 years ago

          That is what I still recalled about physics…

          But as you clearly identified at the end. to get the last hz.. all bets are off..

    • The Dark One
    • 11 years ago

    At this time of year, I wouldn’t mind having fifteen more watts of heat warming my feet.

    • DancinJack
    • 11 years ago

    Oh, and if you’re looking to purchase the 125W version of the X4 965, look for a part number ending in M, as in HDZ965FBK4DGM, or r[

      • Firestarter
      • 11 years ago

      or r[

        • indeego
        • 11 years ago

        ig{<โ„ข<}gย ย ย ย ย ย ย 

      • Prodeous
      • 11 years ago

      I agree, simply awesome..

Pin It on Pinterest

Share This