Intel 45nm Penryn Details Announced

24

Comments

  • Reply 21 of 64
    irelandireland Posts: 17,798member
    Quote:
    Originally Posted by mwswami View Post


    Even a low double digit gain is nothing to sneeze at.



    Point is it's a very vague statement to make. They say it's the biggest advancement in the chip in over 40 years, if we only see 10% gains, then thier level of bullshit would be 100% IMO. Let's wait and see, I was just looking for some more definite statements to come from them that's all.
  • Reply 22 of 64
    When will we see these chips in Macbook Pros?
  • Reply 23 of 64
    irelandireland Posts: 17,798member
    Quote:
    Originally Posted by CoolHandPete View Post


    When will we see these chips in Macbook Pros?



    Tomorrow.
  • Reply 24 of 64
    rolorolo Posts: 686member
    So, Santa Rosa C2D iMacs, MBPs and MBs this spring and Penryn-based iMacs, MBPs and MBs in the fall, right? When do we get new case designs? Santa Rosa or Penryn? Spring or fall?



    I'm in the market for a new MBP but will wait a couple months for Santa Rosa (post Leopard) but if it's only a minor update, I'll wait until fall for the 45nm job.



    Am I correct in assuming that C2D will perform better with Santa Rosa and that Penryn will do better with Nehalem? If that's the case, buying a MBP either this spring or next would make the most sense.
  • Reply 25 of 64
    murkmurk Posts: 935member
    Intel's constant if you think our chips are great now, just wait six months announcements have got to drive Steve batty.
  • Reply 26 of 64
    Quote:
    Originally Posted by Ireland View Post


    Point is it's a very vague statement to make. They say it's the biggest advancement in the chip in over 40 years, if we only see 10% gains, then thier level of bullshit would be 100% IMO. Let's wait and see, I was just looking for some more definite statements to come from them that's all.



    I wouldn't count on a whole lot more. New instructions and low-level features typically account for 10-20% improvements except in very specialized cases.
  • Reply 27 of 64
    Quote:
    Originally Posted by murk View Post


    Intel's constant if you think our chips are great now, just wait six months announcements have got to drive Steve batty.



    On the contrary, it must make him very happy.
  • Reply 28 of 64
    Quote:
    Originally Posted by Ireland View Post


    Point is it's a very vague statement to make. They say it's the biggest advancement in the chip in over 40 years, if we only see 10% gains, then thier level of bullshit would be 100% IMO. Let's wait and see, I was just looking for some more definite statements to come from them that's all.



    As I understand it, Intel is employing newer techniques (high-k dielectric and the metal gate) in the 45nm process to bypass the limits that were being reached with technologies used at 65nm. It also now makes it possible to go to 32nm and 22nm and maybe beyond in the future.



    Intel hasn't provided all the CPU specs yet but we can draw some conclusions based on the data they have made available.
    • 20% increase in transistor switching speed and 30% reduction in switching power => Faster and/or cooler CPUs. I am calculating top of the line DC mobile CPU to be 2.8 GHz at TDP lower than the current 35W for Merom ... maybe mid 20s. The ULV DC versions might even go below 10W (imagine that inside a touch screen Mac Tablet!) For the DC desktop chip (Conroe replacement) , the speeds may even reach 4GHz while sticking to 65W. For the DC and QC new Xeon/Penryn chips, the speeds will be somewhere in the middle while keeping TDP at or below the current levels.



    • 50% increase in L2 cache from 4MB to 6MB for Dual Core and from 2x4MB to 2x6MB for quad core => Further improved performance and Intel probably able to defend against the up coming AMD native QCs (Barcelona and Agena) and buy time till Nehalem for their own native QC.



    • 30% decrease in die size (from 155 to 111 mm^2) => Less Cost



    • New SSE4 instructions => Good further boost for multimedia applications



    • Multi-die Package Quad Core => Higher Yield

    So basically much faster, cooler, and cheaper CPUs ahead ....
  • Reply 29 of 64
    backtomacbacktomac Posts: 4,579member
    Quote:
    Originally Posted by Ireland View Post


    Point is it's a very vague statement to make. They say it's the biggest advancement in the chip in over 40 years, if we only see 10% gains, then thier level of bullshit would be 100% IMO. Let's wait and see, I was just looking for some more definite statements to come from them that's all.



    I think moving forward users will have to be happy with more modest gains when new chips are released. The stark contrast between pentium and core 2 was due to pentium sucking so bad. The comps with core 2 won't be as easy.
  • Reply 30 of 64
    nvidia has pro cards that can do sli
  • Reply 31 of 64
    murkmurk Posts: 935member
    Quote:
    Originally Posted by Programmer View Post


    On the contrary, it must make him very happy.



    With his penchant for secrecy? I doubt it. He may be happy about the progress, but would rather keep it under his hat until its time to ship. Everyone's waiting for dual quads and Photoshop. Why not wait a few more months for Penryn? I'm sure he hates that.
  • Reply 32 of 64
    jeffdmjeffdm Posts: 12,951member
    Quote:
    Originally Posted by Joe_the_dragon View Post


    what about more pci-e lanes in severs / workstations. 20-24 sucks next to NForce pro.



    You've harped on this, but I'm not really buying it now. When tests show at most a 5% difference between using 8 and 16 lanes on a graphics card, I'm just not seeing the need.
  • Reply 33 of 64
    addaboxaddabox Posts: 12,665member
    Quote:
    Originally Posted by murk View Post


    With his penchant for secrecy? I doubt it. He may be happy about the progress, but would rather keep it under his hat until its time to ship. Everyone's waiting for dual quads and Photoshop. Why not wait a few more months for Penryn? I'm sure he hates that.



    Whatever he thinks of it now, it can't be any different than what he thought of it when he pulled the trigger on the switch. This is standard Intel roadmapping.



    The only reason IBM/Freescale chip advances could seem "surprising" was that Apple was their only desktop customer and the advances were so few and far between.



    I'm sure Steve is comfortable with a tradeoff between that kind of "secrecy" and a schedule of rapid improvements that allow Apple to plan new products or form factors with a reasonable assurance that they will actually be able to bring them to market.
  • Reply 34 of 64
    Quote:
    Originally Posted by JeffDM View Post


    You've harped on this, but I'm not really buying it now. When tests show at most a 5% difference between using 8 and 16 lanes on a graphics card, I'm just not seeing the need.



    it more then that as the x16 lanes need by the video card only leave a few for other cards
  • Reply 35 of 64
    shanmugamshanmugam Posts: 1,200member
    good time for consumers!, hope AMD comes out with something surprising. That will heat up the competition further.



    Hope to see Quad Core in MacBooks and MacBook Pros ...
  • Reply 36 of 64
    irelandireland Posts: 17,798member
    Quote:
    Originally Posted by mwswami View Post


    [*]20% increase in transistor switching speed and 30% reduction in switching power => Faster and/or cooler CPUs. I am calculating top of the line DC mobile CPU to be 2.8 GHz at TDP lower than the current 35W for Merom ... maybe mid 20s. The ULV DC versions might even go below 10W (imagine that inside a touch screen Mac Tablet!) For the DC desktop chip (Conroe replacement) , the speeds may even reach 4GHz while sticking to 65W. For the DC and QC new Xeon/Penryn chips, the speeds will be somewhere in the middle while keeping TDP at or below the current levels.



    [*]50% increase in L2 cache from 4MB to 6MB for Dual Core and from 2x4MB to 2x6MB for quad core => Further improved performance and Intel probably able to defend against the up coming AMD native QCs (Barcelona and Agena) and buy time till Nehalem for their own native QC.



    [*]30% decrease in die size (from 155 to 111 mm^2) => Less Cost



    [*]New SSE4 instructions => Good further boost for multimedia applications



    [*]Multi-die Package Quad Core => Higher Yield[/LIST]

    So basically much faster, cooler, and cheaper CPUs ahead ....



    So now I think I get it. Does that mean that a 2.0Ghz one of these chips will be 10 or 20% faster than an existing 2.0Ghz? Or is it your point that we'll see more of a higher Ghz number in computers? Or is it both of these things? Please explain.
  • Reply 37 of 64
    mwswamimwswami Posts: 166member
    Quote:
    Originally Posted by Ireland View Post


    So now I think I get it. Does that mean that a 2.0Ghz one of these chips will be 10 or 20% faster than an existing 2.0Ghz? Or is it your point that we'll see more of a higher Ghz number in computers? Or is it both of these things? Please explain.



    It a combination of the two factors.
    1. In a same speed comparison between say Merom and Penryn, Penryn will win because of the extra L2 cache, SSE4, and some other minor micro-architecture changes that haven't been disclosed as yet.



    2. The Penryn family of processors will also be offered at higher speeds then the Merom/Conroe/Woodcrest.

    The bottom line is that for the same $ we will be able to get much faster and/or cooler processors across the various form factors.
  • Reply 38 of 64
    BTW, the L2 cache is 3MB total shared cache, not 6. This looks to be an evolutionary step forward. Looking forward to it, but not with baited breath.
  • Reply 39 of 64
    mwswamimwswami Posts: 166member
    Quote:
    Originally Posted by 1337_5L4Xx0R View Post


    BTW, the L2 cache is 3MB total shared cache, not 6. This looks to be an evolutionary step forward. Looking forward to it, but not with baited breath.



    Intel currently makes the Core 2 duos in 2MB and 4MB L2 cache variety. All the Core 2 duos in Macs are the 4MB variety. In Penryn these caches are being increased to 3MB and 6MB respectively. Penryn is definitely a big deal ...
  • Reply 40 of 64
    Whupsy, my gaff.
Sign In or Register to comment.