If I had a Hammer: Gelsinger details Nehalem

We’ve already discussed the new information Intel let slip today about its upcoming Penryn processor, the 45nm die shrink of the Core 2. Penryn chips are expected to enter production late this year. The next step after that is the processor code-named Nehalem. In accordance with the “tick-tock” development rhythm Intel has embraced, Nehalem will also be a 45nm part, but it will bring a major microarchitectural refresh.

Intel’s Pat Gelsinger unveiled some basics about Nehalem for the first time today, and in doing so, he called it the first truly dynamically scalable microarchitecture. We don’t yet know entirely what that means, but Gelsinger predicted the character of the architecture will become clearer as Intel discloses more about it. Here’s what we know about Nehalem now.

  • Has roots in the Core microarchitecture — To be precise, Gelsinger stated that the new architecture “leverages” the four-issue-wide, er, core of the Core microarchitecture.
  • An on-die memory controller with P2P interconnects — What the Hector? Yep, Intel admits it’s going for an integrated memory controller much like AMD’s. Nehalem’s memory controller will be designed to interface with DDR3 memory, which should be the standard in the market by the time these CPU products arrive. Intel will provide a version of its memory controller to support buffered DIMMs for server configs, as well.

    In addition, Nehalem will ditch the front-side bus for a series of point-to-point interconnects, presumably similar to AMD’s HyperTransport technology.

  • Intel’s own brand of fusion — Not content to ape AMD’s basic system architecture, Intel will also emulate AMD’s Fusion initiative by moving its integrated graphics core from the north bridge to the CPU socket for “mainstream” client PCs—think Centrino laptops and vPro corporate desktops. We don’t yet know whether the graphics core will be integrated directly into the processor die or placed in a multi-chip arrangement like Intel’s current quad-core processors are. We also don’t know how many CPU cores will potentially coexist with a GPU at once. Gelsinger would only say that graphics will be “in the processor socket.”
  • Hyper-Threading is back! — It may or may not be called Hyper-Threading, but Nehalem will have a similar simultaneous multithreading capability. Each processing core on the chip will have two front-ends, much like the Pentium 4 did. That means a quad-core version of Nehalem would be able to execute eight threads in parallel.
  • A multi-level shared cache architecture — I suppose one could say Core 2 Duo has a “multi-level shared cache architecture,” but reading between the lines, I believe Gelsinger may have been signaling the potential addition of an L3 cache level which could be shared between cores.
  • Something called “Performance enhanced dynamic power management” — Penryn’s dynamic acceleration is impressive in its own right, but Gelsinger said this one would involve “new, undisclosed technologies.” We don’t yet know what those are.
  • A native 45nm design — Intel claims Nehalem will “fully unlock” the benefits of its 45nm high-K fab process because it’s a native 45nm design, not a die-shrink.
  • Scalable configurations — The new architecture can scale from one to eight cores and, correspondingly from one to 16 simultaneous threads (at a peak of two threads per core). Its cache architecture is similarly scalable, allowing Intel to create a large family of products based on this design.

Gelsinger said Nehalem is currently “on track for production in 2008.”

Comments closed
    • wof
    • 13 years ago

    Oh no Hype-Threading is back πŸ™

    8 fake cores at 60% speed or 4 real cores att 100% ? I’d take the 4 anyday unless it’s for a server with tons of users.

    Why ? Because there’s nearly no software that scales well enough to actually get a net gain from those theoretical extra 20% of performance .

    Also as you optimise code the software scales worse as there are less spare resources to Hype-Thread width which makes it a pita for the few of us who still know how to write good code.

    Otherwise it seems nice and actually very similar to K8L or K10 or whatever it’s called.

      • IntelMole
      • 13 years ago

      Hey, you can still have your 4 cores at 100% if you want. Disable SMT, like people did with the P4. Or leave it running and watch as you don’t have enough threads at the moment to saturate a 4 core machine.

      y[

        • echo_seven
        • 13 years ago

        Regarding performance loss due to hyperthreading — the only reason I was aware of where hyperthreading reduces performance is due to inefficient use of cache, as it gets shared between threads. Is there any other major reason that someone would like to share with us? Wikipedia talks about slowdowns associated with the infamous “replay”, but that won’t be a problem in ICM architecture CPUs.

        It seems a moot point to me. By the time you’re up to that 5th thread do you really care if you could lose (potentially) 10% overall performance on 2 of the threads? If I cared I’d just stop running something so I’m back down to 4 threads…

          • wof
          • 13 years ago

          If your app only scales with 90% extra performance for each core you’ll loose overall performance going to 8 core smt over 4 core.

          If you look closely at the TR benchmarks of quad cores you’ll find that almost nothing scales to 4x performance on 4 cores.

            • echo_seven
            • 13 years ago

            Well, yes, nothing scales perfectly, but that’s different from saying that you’re going to lose overall performance simply by gaining the capacity to run more things simultaneously. When an application doesn’t scale perfectly, it implies unused resources on the 2nd, 3rd 4th core, because the programmer couldn’t find enough extra work to task them with. Isn’t that what hyperthreading would /[

        • wof
        • 13 years ago

        Most computer problems are incredibly hard to scale perfectly over multiple threads i.e. anything that isn’t “embarrassingly parallel”. However it’s usually easy to achieve say 60% more performance by using another cpu (cross cpu communication and locks etc will eat a lot of performance for many problems) if your second thread only gives 60% more performance you’ll effectively loose performance on SMT systems so you have to detect SMT cpu’s and write code to schedule things unless your software is aimed at freaks who manually sit and prevent tasks from running on fake cores.

        y[

          • Chryx
          • 13 years ago

          the pipeline bubbles being filled with a second thread is a totally different implementation to using the unused execution pipelines, Nehelam’s implementation will look like POWER5’s, not P4’s

    • Chaos-Storm
    • 13 years ago

    What? no “bitchin fast” feature?

      • Flying Fox
      • 13 years ago

      Well, Penryn is supposed to be bitchin fast already. Since “Nail-em” is the next gen, it is probably implied that it is /[

        • echo_seven
        • 13 years ago

        Bitchin Fast (BF) technology will be disabled in the Celeron version.

    • poulpy
    • 13 years ago

    [OT] speaking of next gen a bit more info on Barcelona: Β§[< http://anandtech.com/tradeshows/showdoc.aspx?i=2956&p=5<]Β§

    • d0g_p00p
    • 13 years ago

    I had high hopes for the K10, but this just seems like a slaughter. AMD needs to come up with something that is at least *close* for a reduction in price. The 486 DX133 vs the iP75, K6 vs iPMMX and Barton vs Coppermine P4 spring to mind.

    I really want to see what AMD has cooking. However, I think intel will be in my machine for quite some time.

      • echo_seven
      • 13 years ago

      did you mean Barton vs. Northwood P4, or Barton vs. Coppermine P3 ?

      edit: looking at the other two examples, you’re giving examples of when AMD came up close in performance, but didn’t quite beat Intel, I think. So that should be Northwood P4?

        • SPOOFE
        • 13 years ago

        I would bet he’s talking about the Northwood. From what I read, it’s considered the “best” (warning, subjectivity alert!) of the Netburst era… although I’ve heard people say that the 65nm Pentium D’s were good chips (don’t know myself, I wasn’t paying attention to Intel all that much at the time) but had a marred reputation due to Prescott…

        Oh, ignore me. I’m still trying to sober up.

    • Jive
    • 13 years ago

    I dont think the government would let AMD die anyways, would they? If AMD was in trouble wouldn’t they ask the government for help to bail them out anyways. Haven’t they done it before?

    • Forge
    • 13 years ago

    I really hope that even the rabid Intel fanboys realize what a huge disaster it would be to have AMD pull back to a K6 era position. I still remember the two-three speed grades of Pentium per year and the horrifying price gouging that we endured (2000$ Xeons, anyone? At the LOW end?).

    Please Gord, keep AMD alive and relevant.

      • just brew it!
      • 13 years ago

      Exactly.

      Since the beginning of the Athlon era, the enthusiast community has benefited *[

        • provoko
        • 13 years ago

        Yup, you could see this when the first AMD X2 dual cores came out at 500 dollars (intel seriously sucked), then AMD slashed prices to 300 after a long time.

        The only reason the X2s are less than 150 is because Intel has leap frogged way beyond them with the Core 2 Duos.

        But this Penryn chips look as though they will beat out AMD’s quad core and 8 core cpus next year, if AMD will make any 8 cores next year…… they don’t even have 4 cores yet! WTF AMD, PICK UP THE PACE!

      • Shintai
      • 13 years ago

      Unless people are stupid, we all know that a monopol would suck. We can just look at MS…why does an OS needs to be 2.5-3.7GB compressed and 10-15GB installed etc.

      Its a no brainer that if AMD or Intel was alone, they would only do the required developement to keep anyone from starting up.

      However it still doesnΒ΄t change the financial facts.

    • flip-mode
    • 13 years ago

    Shin, make sure you really rack up your post counts on these two news releases.

      • Shintai
      • 13 years ago

      Do you have a trolling quota you need to forfill each day? You really rack up your postcount that way if you get it….hint hint!

        • Anomymous Gerbil
        • 13 years ago

        You know exactly what he means.

          • Shintai
          • 13 years ago

          Yes, thats why I wrote as i do.

            • l33t-g4m3r
            • 13 years ago

            I was wondering who the paid shills were, now I know. Thank you for admitting it.

            P.S.
            Don’t get too offended, theres several other “fanboys” here I have the same opinion of.

            nothing really set me off on this rant other than being tired and cranky from working close to 16 hours today, I’m tired of the usual constant fanboy/troll crap, and just felt like saying something.

            once in a while, its fine, but every friggin news article on a cpu or gpu or whatever. geez, enough people.

            • Shintai
            • 13 years ago

            Do you even know what a shill is? My behaviour is far from that. And if you wanna accuse someone you better back it up.

            I can prove to you what I work with and where. Can you prove the other thing?

            • l33t-g4m3r
            • 13 years ago

            my post probably was unwarranted today, like you said, I just noticed the one comment and agreed with what seemed to be the sentiment of the OP.

            anymore the front page on tr is just a plain troll-fest come news on a new gpu or cpu, and its just plain ridiculous.
            its like when porky was here, except 10 others popped up to fill his place.

            Just once I’d like to read the news without seeing a horrible fanboy flamefest going on below in the comments, thats all.

            • Shintai
            • 13 years ago

            I understand.

            But you will always more or less end in this. Its mainly due to the speculation and the discussion it follows. This is really no different that it was 2-3 years ago. It just increased alittle in volume. Mainly due to the media feeding more halfbaked rumours to speculate on. Then we get a longer time, with less reliable information to speculate on, discuss and argue. That always sooner or later end with personal comments due to the nature of the annonyme internet.

            This is not a news comment either, its also in MMOs etc this trend continues.

            Best advice I can give you, skip the comments.

            • Anomymous Gerbil
            • 13 years ago

            Oh dear.

            • moose17145
            • 13 years ago

            y[< Shin, make sure you really rack up your post counts on these two news releases.<]y Lol, funny thing is... it's working! HAHA!

    • PenGun
    • 13 years ago

    Damn the Pentium M is sure getting a workout …

    • slaimus
    • 13 years ago

    This could very well be like a Conroe + i965G on a single package.

      • Krogoth
      • 13 years ago

      That is far too much of an oversimplification.

      It is more like step from Pentium 2 to Pentium 3.

        • roadrunner
        • 13 years ago

        The first Pentium 3 was just a Pentium 2 clocked 50MHz faster with SSE support…not much of a change.

        Later P3’s moved the separate, half-speed 512KB cache on chip, and full speed, but only 256KB. Still faster, though…

    • Drewstre
    • 13 years ago

    Hrm. Maybe it’s about time to retire the old PII-450…

    • Krogoth
    • 13 years ago

    Nehalem looks more then sexy enough to skip Pernyn is its kin.

    • just brew it!
    • 13 years ago

    Yikes… a next-gen 45nm “Core 2” style architecture, *[

      • Krogoth
      • 13 years ago

      Nehalem is a long away from reaching retail channels. AMD main concern is getting the K10 onto table and compete against Penyrn-based chips.

        • Flying Fox
        • 13 years ago

        2008 is not exactly long, but delays can always happen. We will see if Intel hits into misfortune in fabbing this thing.

      • Vaughn
      • 13 years ago

      u guys are like broken records. intel release new CPU whoa AMD is gonna die. AMD releases new CPU whoa Intel is gonna die. You do know it will take more than 1 or 2 successful products to put either one of these multi billion dollar companies out of business. And even more so for intel since they are sitting on so much cash.

      I worry for neither of them. Look forward to whats ahead and hope they both survive to keep prices low.

        • Shintai
        • 13 years ago

        Profit is the indicator we usually use. And one of them aint exactly making much if any money.

        It took 1 product to kill 3Dfx.

          • Krogoth
          • 13 years ago

          3Dfx was killed by a bunch of factors none of them were really a product mishap.

          The V5 6000 was never officially released only half-working prototypes were made.

          3Dfx’s fate was sealed before the depute of the V5 series.

            • moose17145
            • 13 years ago

            Well just let me say this then, if AMD doesn’t kick Intel’s butt, then they better be kicking NVidia’s butt. If they screw up on both fronts though… THEN we can start worrying. At the very least they better be dang competitive on at least one front (preferably both).

        • just brew it!
        • 13 years ago

        While I don’t think either is gonna die anytime soon, it isn’t too far-fetched to imagine that AMD could slip back into the mode they were in back in the K6-x days, where they are relegated mainly to the low end. That would be unfortunate for everyone.

        • UberGerbil
        • 13 years ago

        When AMD was thumping Intel among enthusiasts, nobody was saying that Intel was going to die (well, except perhaps for a few extreme fanboys who had no clue how to read financial statements, similar to the people who think the PS3 could kill Sony). Given the size disparity (by almost any measure Intel is 7 to 10 times the size of AMD), things are far more dire for AMD when Intel has the technological upper hand than when the situation is reversed.

        AMD took on a *[

      • sluggo
      • 13 years ago

      Maybe Intel’s pronouncing it “Nail ’em”. πŸ™‚

        • Flying Fox
        • 13 years ago

        Good one. πŸ™‚

    • Shintai
    • 13 years ago

    Just a note, desktop and laptop versions will NOT have an IMC.

    And the cache is still a shared L2, just 24way with other enchancements.

      • MadManOriginal
      • 13 years ago

      You know this for a fact how..?

        • Forge
        • 13 years ago

        Kickin Pat Gelsinger talks in bed?

        Jeez, I’m being rough on poor Pat.

        • Shintai
        • 13 years ago

        Because it would be wierd for the Nehalem desktop chipsets to carry a MCH if the CPU got it. The only thing thats up in the air is, if the extreme parts will have IMC or not.

    • PetMiceRnice
    • 13 years ago

    It’s unreal how quickly things are changing. But I’m holding on to my E6400 until at least the end of 2009 when I’ll be looking at a new system again.

      • packfan_dave
      • 13 years ago

      Same with my T7200. Probably looking a Westmere (the 32nm shrink of Nehalem) by then if Intel’s still sitting at the top of the apple cart for an upper-midrange notebook (assuming my relative disposable income doesn’t change much).

    • Dizik
    • 13 years ago

    Damn. Is it wrong to lust for a wafer?

      • Gungir
      • 13 years ago

      Hells no.

      I want to be loyal to AMD, I really do, but it keeps getting harder. πŸ˜‰ K10 can’t come fast enough, and from the sound of this, its harder-hitting, next-gen big brother (K11?) can’t either.

      • Flying Fox
      • 13 years ago

      A wafer always turns me on. πŸ˜€

    • 5150
    • 13 years ago

    I want this now. Please?

      • echo_seven
      • 13 years ago

      Perhaps its time to invest in a DeLorean πŸ™‚

        • Krogoth
        • 13 years ago

        You are also going need some plutonium or Mr. Fusion to power that Flux Capacitor which needs 1.44 Jiggawatts of power!

          • 5150
          • 13 years ago

          First of all, it’s 1.21 jigawatts, secondly, I have already purchased my DeLorean, installed ground effects, a coffee can muffler, and a Type-R sticker. I can now hit 88 miles per hour in 1.4 seconds. Einny approves.

Pin It on Pinterest

Share This