A triple-core Llano CPU might be on the way

In recent years, AMD has positioned triple-core versions of its Phenom, Phenom II, and Athlon II processors at the bottom of its desktop CPU lineup. It wouldn’t be much of a stretch, then, for the chipmaker to offer a triple-core Llano accelerated processing unit. According to Fudzilla, that’s exactly what’s going to happen: the site claims it has "been able to confirm" that such a chip will be out some time this quarter.

Reportedly, this unannounced triple-core A-series APU will resemble the A6-3600, one of AMD’s still-unreleased 65W desktop Llano chips. The A6-3600 has four cores, a base clock speed of 2.1GHz, a top Turbo Core speed of 2.4GHz, and Radeon HD 6530D integrated graphics with 320 ALUs and a 443MHz core speed. Fudzilla says you’ll see "pretty much identical" specs on the triple-core part, just with one fewer core. (There’s no word on whether the TDP will be lower because of the missing core, though.)

Producing such a chip surely wouldn’t be particularly difficult for AMD. The company’s previous triple-core offerings have been quad-core chips with one disabled core, and Llano is based on the same CPU architecture as AMD’s Phenom II and Athlon II processors. With the cheapest desktop Llano APU currently selling for $119.99, and signs pointing toward similar pricing for the A6-3600, triple-core variants may be Llano’s ticket into the sub-$100 price range.

Comments closed
    • ronch
    • 9 years ago

    My AMD Phenom II X3 720 never fails to make me get tired of it. Got it cheaper than a Core 2 Duo E8400, offers 20% higher aggregate performance (single threaded performance is close enough), and with the 4th core unlocked, gives me Core 2 Quad Q9300-level performance.

    Quad core performance for a price less than a dual core. If they can pull that off with Llano, then it’ll be a hit. Problem is, the Intel landscape will be hard for AMD to match nowadays, with Intel dual cores on parity with AMD quad cores, unless AMD prices 3-core Llanos really low.

    • derFunkenstein
    • 9 years ago

    Without ACC this is going to be pretty much shunned by enthusiasts who expect something for free (like me! Phenom II X2 550 BE running as a quad)

    • dashbarron
    • 9 years ago

    Are these quad-cores with one disabled, or quad-cores with one core defective and then disabled?

      • stdRaichu
      • 9 years ago

      Probably a combination of both, but mostly the latter unless yields are already very good.

      • Goty
      • 9 years ago

      I think it depends on a number of factors. You could expect that AMD would stockpile the chips with one defective core and then release a triple-core CPU when they had a significant supply, but later on they might release fully-functional quads with one core disabled once yields improve enough if only to fill the price point.

        • OneArmedScissor
        • 9 years ago

        They may very well already be selling some chips with a single faulty core as dual-cores. There’s a good chance that the idea is to start selling tri-cores once they are actually manufacturing dual-core chips and they have more options for binning. They need to offer dual-cores at a low price right out the gate to get OEMs to bite.

        Or maybe they really are stockpiling them, and there just aren’t very many because the borked chips tend to come out with more screwed up than one core. We’d never know without them outright saying so, and they never would.

        • dashbarron
        • 9 years ago

        Makes sense. It is a bit shocking that AMD or Intel for that matter would have that many consecutive cores going bad on them while producing their chips. Would it happen that it is the same core all the time or a different 4th core? I ask because I would think it would make some performance difference depending on which of the four cores were disabled/defective, no?

          • OneArmedScissor
          • 9 years ago

          Is it? CPUs have had hunks of cache disabled since the dawn of time. GPUs have all sorts of parts disabled, even when they’re the “good” chips.

          The denser and more complicated CPUs become, and the smaller each part is, the likelier it is that a defect will pretty much ruin the particular part it affects.

          • Goty
          • 9 years ago

          I wouldn’t be surprised if AMD’s yields are still a bit low with Llano, especially considering how man firsts it encompasses for the company (first 32nm chip, first GPU on SOI, etc), so it wouldn’t be too surprising to learn that AMD had a decent stockpile of less-than-functional quad cores. The defects on a wafer should be distributed randomly, so whichever core was non-functional (or didn’t quite meet specs for operating voltage or clock speed) would have to be random as well. Performance should be the same no matter which cores are disabled unless AMD did something silly like having each core connect in serial to the others.

            • NeelyCam
            • 9 years ago

            I remember saying GloFo yields still suck (and getting voted down for it).. so…

            Was NeelyCam right… again?
            Answer: yes

            • Goty
            • 9 years ago

            I don’t recall yields sucking with Phenom II, yet we still got the X3s shortly after launch.

            Was NeelyCam right… again?
            Answer: maybe

            • NeelyCam
            • 9 years ago

            I meant the 32nm SOI. But, it was very possible that there were yield issues with 45nm as well… Back then with AMD controlling the fabs, I’m sure they kept any yield issues a secret so they wouldn’t freak everyone out… but there must have been a reason why X3s came out.

            Overall, I think it’s a clever use of crappy silicon – it’s much better to sell bad X4s as discount X3s than to scrap them… I haven’t seen Intel do such things, though (except maybe with the cache). Then again, if they had yield issues, Intel wouldn’t let anybody know about it unless they’re significant enough that investors need to be warned..

            • Goty
            • 9 years ago

            I guess the question to ask would be “how bad do yields have to be for AMD to get a decent stockpile of three-core chips by whatever time they release them?” Of course, the answer would depend on a number of factors, including how aggressively AMD is binning the chips.

            • OneArmedScissor
            • 9 years ago

            “…but there must have been a reason why X3s came out.”

            People will buy them. See below:

            “I haven’t seen Intel do such things, though.”

            Uh, excuse me? Where did the single-core Core 2s come from? Or the quad-core Dunningtons? Or the single and dual-core Bloomfield Xeons? Or the quad-core Gulftown Xeons? Or the quad and six core Becktons? Or the six and eight core Westmere EXs? Or the quad and six core upcoming Sandy Bridge E/EN/EXs?

            They’re likely selling some quad-core Sandy Bridge chips as dual-cores, as well. It’s not like they’re going to throw them away.

      • ronch
      • 9 years ago

      Some will be the former, and some will be the latter. Simple as that. Being able to unlock the 4th core depends on your luck and the motherboard you’re using (and of course, whether AMD even allows it in the first place).

    • mutarasector
    • 9 years ago

    Interesting. The key factor of interest to me is the TDP, as it I would think it should have a lower TDP if and when all three cores are in use. What I’m wondering is if/when a triple core variant of the 100W A8-3850 Llano will emerge. Those might be of more use in an HTPC.

      • OneArmedScissor
      • 9 years ago

      It’s not going to lower the TDP figure they quote, but that doesn’t mean it doesn’t use less power under load. That’s different for every chip, though.

      TDP is just a ball park figure for OEMs to gauge cooling requirements. If you will notice, desktops with a tower case have used CPUs in “steps” of 65w, 95/100w, and 125/130w for years. It’s really just three different general levels, with the rare deviation that drops into laptop territory, like 45w, being for all-in-ones, not towers.

      They’re not going to adjust it a few watts because the heat output is largely dictated by clock speed and voltage, and it’s not an exact figure, anyways.

      • jensend
      • 9 years ago

      The 65W versions (A8-3800 and A6-3600) already look like they can be put to good use in an HTPC: [url<]http://www.xbitlabs.com/articles/cpu/display/amd-a8-3800_21.html[/url<]

      • Farting Bob
      • 9 years ago

      Why would you need a 100w chip in a HTPC? Unless you play games (at a lowish resolution) its overpowered for what you’d need. Hell my e-350 is powerful enough for 1080p/blu-ray.

        • mutarasector
        • 9 years ago

        I’m not thinking of the 100W so much, but the higher clock.@ 2.9Ghz but on fewer cores and at what price point. For the application I have in mind, I thought it might make more sense to have a lower power draw overall on just two cores operating, with one disabled, and a 3rd primarily idling. Of course, I don’t know how Llano will be power gating on a 3 core chip if at all.

          • jensend
          • 9 years ago

          If you’re really that worried about the power consumption from the extra cores then you should wait around until they introduce the two-core A4s (half the transistors-> definitely less power usage at same clocks). But they’ve done a good enough job at power gating all four cores that I don’t know how much of an advantage that’ll be.

            • mutarasector
            • 9 years ago

            More waiting… First, waiting on mini ITX Llano boards, and now a dual core @ 2.9+Ghz. Grrrr……….

Pin It on Pinterest

Share This