32nm, six-core Gulftown prototype spotted

If there’s any truth to last month’s rumors, we may see Intel’s 32nm, six-core Gulftown processor in the first quarter of next year. As usual, though, some folks seem to get hold of these parts a little before everybody else.

Taiwanese overclocker JCornell has posted two images of a purported Gulftown prototype in the XtremeSystems Forums. The first image shows the underside of the CPU package next to a couple of Nehalem-based Xeon processors. (So one would think from looking at the capacitor arrangement and the "W5590" caption, at least.) The second image shows the prototype running Windows Vista with the Task Manager and CPU-Z open.

Barring some rudimentary Photoshop trickery, CPU-Z identifies the processor as a 32nm Gulftown engineering sample rated for 2.4GHz. The software also detects six cores, 12 threads, six chunks of 256KB L2 cache (1.5MB total), and 12MB of L3 cache. Vista’s Task Manager reported 24 threads, which suggests JCornell was testing not one, but two Gulftown samples inside a workstation or server build.

As we noted back in February, Intel may make this processor by cramming three dual-core dies onto a single package. That might make Gulftown relatively affordable to produce, although we wouldn’t exactly expect Intel to launch it at a bargain-basement price.

Comments closed
    • ish718
    • 10 years ago

    OK, so when will software catch up? O_O

    • JumpingJack
    • 10 years ago

    l[

    • JumpingJack
    • 10 years ago

    Double pst

    • cygnus1
    • 10 years ago

    My baseless conjecture is that Intel is taking a similar route as the Phenom X3, and ganging together 2 quad cores with 1 disabled core each.

    All they would need to do is use the spare QPI link of a regular nehalem to connect the second die and it would access memory through the first die. Or they could do some screwy NUMA stuff with the memory channels since the hardware is already there for that.

    • bcronce
    • 10 years ago

    Where’s the 6-core i7? And what’s AMD going to do about the i7 anyways? 4 core i7 is still winning/tieing many benchmarks over the 6 core AMD and uses less power

      • Game_boy
      • 10 years ago

      They’re not going to do anything. $250+ represents 1% of the market, and AMD is competitive in the remaining 99%.

        • MadManOriginal
        • 10 years ago

        Yeah unfortunately that $250+ market is where the good margins are.

          • accord1999
          • 10 years ago

          Plus, the only time periods where AMD has really made money is when it had the performance advantage; and therefore could sell products in the $250+ market.

      • vikramsbox
      • 10 years ago

      Well, they are going on about it, but they would have had to massively change the Bulldozer designs after the i7. Further, the notion would be to first get the existing products in line first and then hype the new ones.
      CPUs carry a large profit margin, the key is to sell all the cpus the fabs can make, and this is what wasn’t happening to AMD leadin to losses.
      The PhenomII would give AMD some breathing space, while it got its planning, design and finances in order. They can’t afford to make any more dramatic changes in design, and dramatic changes are not needed, as the Phen-2 has shown, to give a competitive CPU.
      Lets wait and see.

    • Saribro
    • 10 years ago

    hmm, forgot to press “Reply”. nm

    • flip-mode
    • 10 years ago

    So, probably a question that has been addressed two dozen times, at what point do you have diminishing returns with more threads – even taking VMs and mutli-threaded apps into account? Ever, never, soon, or not for a long time?

    • juampa_valve_rde
    • 10 years ago

    by the time intel sticks three dual core dies (with a nightmare to sync that l3 caches between), amd already has a functional, shipped monolithic 6 core processor. :S

    • Obsidian
    • 10 years ago

    Needz moar threadz.

    • dragmor
    • 10 years ago

    The increased turbo boost on these will be just as important as the extra cores.

    • Firestarter
    • 10 years ago

    I don’t want 50% more cores, I want 50% more megahurtz!

    • bogbox
    • 10 years ago

    Only 2 cores more ? Sorry not interested in a minor update of Nehalem .
    Waiting for 8 cores or 16 virtual cores.
    I wonder how can someone justify buying it. A quad is much more than you need.(I’m not referring to workstation ,server class CPU etc).
    My q9300 is 99% only when I boot into win7(1 minute) the rest of the time is 20% average.

      • Game_boy
      • 10 years ago

      The market for desktop people who need more than four cores/eight threads is tiny, certainly less than half a percent given that Core i7’s share is at 1% anway.

      What company is going to devote engineering, product qualification, motherboard design and marketing time to a market that small? Even Gulftown is only viable because the work’s been done for the server market already.

      AMD has Istanbul, but they’ve done market research that shows there isn’t a market for bringing that to the desktop.

      • Umbongo
      • 10 years ago

      A quad is much more than people need but you are waiting for 8 cores?

      As for justification in buying it, some people will always want and be able to use more power. If you want more than 4 cores I’m sure these will be a cheaper alternative than switching to a dual socket platform.

      There are workflows that utilize all the processoring power of the latest desktops, getting another 50% performance is nothing to be scoffed at.

        • bogbox
        • 10 years ago

        I was thinking at 8 cores are more interesting for me to watch , for benchmarks etc, but not to buy.:))
        Have you bechmark the CPU or read a review of Gulftown ? How do you know that the performents is going to be 50% better that Nehalem ? :))
        I hope Crysis will get that gain.// sarcasm :))
        ps Normal computing ,not workstation,or server. aka gaming ,firefox etc or alt least faster boot time.

          • Forge
          • 10 years ago

          The tasks you refer to CAN NOT be made faster by any CPU tech.

          Games are limited by GPU power.

          Firefox is limited by your internet connection, and maybe your ram.

          Boot time depends on your disk speed more than anything.

          This is like you saying ‘OMFG hybrid card si teh usless, they are slow like internal combustion. lul.’ It’s not addressing that, it’s designed around a whole other problem.

      • jjj
      • 10 years ago

      i wouldn’t call it a minor update when adding 2 cores gives it 50% more performance and the die shrink should give it another 10-15%.
      You are also contradicting yourself first you want 8 cores and yet 4 cores is “much more than you need”.

        • Game_boy
        • 10 years ago

        Die shrinks don’t increase performance.

          • poulpy
          • 10 years ago

          Well if you really want to be pedantic I guess they can very marginally increase performance as you will reduce on-die interconnects distance and will change the mask / have a new revision therefore squash a bug or two and/or improve/simplify some logic here and there.

      • srg86
      • 10 years ago

      Of course it’s only a Minor update of Nehalem. As by intel’s tick-tock scheem, you don’t get major architecture shifts at the same time as a new process tech.

        • ironoutsider
        • 10 years ago

        Agreed, I thought triple 3.4 ghz Phenom II was way overkill as well. With all the movement towards retrogaming and gudd nuff gaming, I’m not so sure what would make processors like these sell well? especially in today’s economy.

          • MadManOriginal
          • 10 years ago

          These are for servers or heavy duty time=money workstations that are loaded almost all the time. For the former virtual server consolidation is compelling for TCO and for the latter the payoff is obvious.

    • Game_boy
    • 10 years ago

    In the thread it’s posted that it failed to boot on an X58. That could be anything from it needing a BIOS update to it not actually working with it.

      • Umbongo
      • 10 years ago

      Probably just needs an update, it happened with the 65nm Core 2 quads and the move to 45nm.

      • JumpingJack
      • 10 years ago

      Or, it could simply be that it is a DP chip with 2 QPI links and not intended for a 1366 x58 board altogether. It would appear that he is working with a 2P server/WS set of chips and it shouldn’t work single slocket, it is not pinned out to do so.

        • Umbongo
        • 10 years ago

        The 5500 series Xeons with 2 QPI links do work in single socket boards. He said that the dual socket board needed an update to work too.

Pin It on Pinterest

Share This