Sign in with
Sign up | Sign in

Intel's Ivy Bridge CPU Die Layout Estimated

By - Source: techPowerUp, PC Watch | B 41 comments

With Ivy Bridge a little over a month away, Hiroshige Goto, contributor for PC Watch, has estimated the layout of Ivy Bridge silicon.

As discussed at the IEEE International Solid-State Circuits Conference, Intel engineer Scott Siers announced that there will be four different Ivy Bridge die models. In addition, Ivy Bridge will carry up to 1.4 billion transistors that span over an area of 160 mm2, which is about 26 percent smaller than the comparable 216 mm2 Sandy Bridge die with 1.16 billion transistors. Ivy Bridge is built on 22 nm process, which is the "tick" process of Intel's Tick Tock Model.

Image Credit: Hiroshige GotoImage Credit: Hiroshige Goto

Taking a closer look at the Ivy Bridge's estimated die layout, the layout is similar in design to current-gen Sandy Bridge. The die is made up of three general sections, 1) CPU cores, 2) System Agent and 3) Graphics core.

The CPU cores are made up of four x86-64 cores with 256 KB dedicated L2 cache per core and shared 8 MB L3 cache. The System Agent holds the dual-channel DDR3 integrated memory controller (DDR3 1600), a PCIe interface (as a shared x16 port or two separate x8 ports), a DMI link, a display controller, power controller unit, and a FDI link. The Graphics Core has 16 programmable EUs that handle parallel processing loads for the GPU and can be programmed to perform GPGPU tasks. In addition, it holds the Multi-Format CODEC, which supports MPEG2, VC1, AVC and also MVC (multi-view video coding) for stereoscopic 3D. All the components are bound by a ring-bus that transports tagged data between the CPU cores, the graphics core, the L3 cache, and the system agent. 

Image Credit: Hiroshige GotoImage Credit: Hiroshige Goto

As Scott Siers announced, there will be four different variants of the Ivy Bridge die models.

  • 4+2: All four cores enabled, full 8 MB L3 cache enabled, all 16 shader cores (EUs) of the IGP enabled
  • 2+2: Two cores enabled, 4 MB L3 cache enabled, all 16 shader cores of the IGP enabled
  • 4+1: All four cores enabled, 6 MB L3 cache enabled, fewer shader cores of the IGP enabled
  • 2+1: Two cores enabled, 3 MB L3 cache enabled, fewer shader cores of the IGP enabled

   

Image Credit: Hiroshige GotoImage Credit: Hiroshige GotoImage Credit: Hiroshige GotoImage Credit: Hiroshige Goto

 

Display 41 Comments.
This thread is closed for comments
Top Comments
  • 23 Hide
    hardcore_gamer , February 27, 2012 9:32 AM
    1/3 rd of the die area is taken by the crappy GPU cores. They should have used it for additional logic or cache to speed up the CPU, atleast for the K series CPUs since the people who buy them use a discrete GPU anyway.
  • 20 Hide
    killabanks , February 27, 2012 9:48 AM
    memadmaxThe tick tock model is why intel will always be the innovator, and AMD the follower...

    more like billions of $$ means intel will always be ahead of amd
  • 11 Hide
    fuzznarf , February 27, 2012 10:12 AM
    hardcore_gamer1/3 rd of the die area is taken by the crappy GPU cores. They should have used it for additional logic or cache to speed up the CPU, atleast for the K series CPUs since the people who buy them use a discrete GPU anyway.

    Yeah, you're right. Intel probably has no idea what they are doing....
Other Comments
  • 2 Hide
    ojas , February 27, 2012 9:14 AM
    I wonder if they'd keep the die size the same as sandy bridge, wouldn't that let them stuff more things like cache and IGP shaders (or whatever IGPs have) in the same space?
  • 23 Hide
    hardcore_gamer , February 27, 2012 9:32 AM
    1/3 rd of the die area is taken by the crappy GPU cores. They should have used it for additional logic or cache to speed up the CPU, atleast for the K series CPUs since the people who buy them use a discrete GPU anyway.
  • 10 Hide
    Anonymous , February 27, 2012 9:47 AM
    @hardcore_gamer
    Ivy bridge and the rest of LGA1155 is ment for the low-mid range market segments, where people want the build-in graphics. Sure most people who buy the K CPUs probably wont ever use the GPU, but it would probably be too expensive to change the layout just for the K modules, not to mention that it would be a completely different design, which kinda goes against the tick-tock model.

    If you wanted to use discrete GPUs, you should go with the SNB-E, Ivybridge-E LGA2011 series, atleast following Intels logic :) 
  • 20 Hide
    killabanks , February 27, 2012 9:48 AM
    memadmaxThe tick tock model is why intel will always be the innovator, and AMD the follower...

    more like billions of $$ means intel will always be ahead of amd
  • 11 Hide
    fuzznarf , February 27, 2012 10:12 AM
    hardcore_gamer1/3 rd of the die area is taken by the crappy GPU cores. They should have used it for additional logic or cache to speed up the CPU, atleast for the K series CPUs since the people who buy them use a discrete GPU anyway.

    Yeah, you're right. Intel probably has no idea what they are doing....
  • -6 Hide
    gilbertfh , February 27, 2012 10:31 AM
    I am excited... I skipped Sandy since I considered her a 7. Hopefully Ivy will be a 10 and age well.
  • 1 Hide
    bartholomew , February 27, 2012 10:45 AM
    killabanksmore like billions of $$ means intel will always be ahead of amd

    :D  & Only the ones with lots of $$ can afford their best CPUs :p 
  • 0 Hide
    builder4 , February 27, 2012 10:47 AM
    hardcore_gamer1/3 rd of the die area is taken by the crappy GPU cores. They should have used it for additional logic or cache to speed up the CPU, atleast for the K series CPUs since the people who buy them use a discrete GPU anyway.


    They don't want to completely smash AMD by speeding up the CPU. Intel relies on AMD's continual existance to prevent the government splitting them up on monopoly grounds. As long as AMD is non competitive, intel won't increase the CPU performance of their mainstream chips.

    They do, however, offer a larger die and more CPU cores replacing the GPU on the 2011 platform for 3x the price.
  • 5 Hide
    mindless728 , February 27, 2012 10:50 AM
    Quote:
    1/3 rd of the die area is taken by the crappy GPU cores. They should have used it for additional logic or cache to speed up the CPU, atleast for the K series CPUs since the people who buy them use a discrete GPU anyway.


    not everynody buys the K series for gaming, I bought one for a minecraft server just to OC it to a pale 4GHz (don't need extreme), as it is a lot cheaper then buying server grade and also let me OC if need be
  • 6 Hide
    fuzznarf , February 27, 2012 11:00 AM
    For the mentally-inept who downgraded my previous post. the 'crappy' GPU is on the chip because it isn't just used for gaming performance.
    Intel has integrated the 2 for a common purpose... but i guess the internet-cpu experts on this forum know better than intel.

    "Zhou's solution was to have the CPU do the leg work by determining what data the GPU needs and then going and retrieving it from off-chip main memory. This in turn leaves the GPU free to focus on executing the functions in question. The result of this collaboration is that the process takes less time and simulations have found that the new approach yields an average improved fused processor performance of 21.4 percent."

    get a clue. http://www.tomshardware.com/news/CPU-GPU-APU-fused-processor-performance-boost,14653.html
  • 1 Hide
    tomfreak , February 27, 2012 11:02 AM
    ojasI wonder if they'd keep the die size the same as sandy bridge, wouldn't that let them stuff more things like cache and IGP shaders (or whatever IGPs have) in the same space?
    If u take a look at the measurement, they certainly have enough room for 2 more cores.

    Ivy bridge would have been hexa core if bulldozer has SB-E performance. I want a Hexa core, but not socket 2011 expensive platform. IF I only wish AMD CPU engineer are as good as their GPU ones.
  • 1 Hide
    thomaseron , February 27, 2012 11:27 AM
    Ivy Bridge will be delayed.

    EDIT: It's true. Appearantly some production issues with the 22nM process...
  • 0 Hide
    EDVINASM , February 27, 2012 11:46 AM
    Nothing fancy in Ivy Bridge design IMO. I was hoping for huge increase in rendering and encoding performance but if the rumours are true mere 20% is not worth it. Better get myself 2600k and overclock it to bits :)  RIP Ivy. Will have to wait for next architecture to show me some 2x increase.
  • 1 Hide
    fuzznarf , February 27, 2012 11:59 AM
    edvinasmNothing fancy in Ivy Bridge design IMO. I was hoping for huge increase in rendering and encoding performance but if the rumors are true mere 20% is not worth it. Better get myself 2600k and overclock it to bits RIP Ivy. Will have to wait for next architecture to show me some 2x increase.

    This will only be the first production cycle of ivy. 3D gates are a phenomenal advancement. Plus much lower power consumpton and cooler running. and this is a ~20% increase over the 3900 series Sandys also...
  • 2 Hide
    kartu , February 27, 2012 12:05 PM
    memadmaxThe tick tock model is why intel will always be the innovator, and AMD the follower...

    Right. And also they know how to make offers no partner can refuse, P4 Prescott, that was:

    1) slower
    2) more expensive
    3) AND consumed much more power

    outsold Athlons like 3 to 1, or 4 to 1?
  • 2 Hide
    vittau , February 27, 2012 12:13 PM
    hardcore_gamer1/3 rd of the die area is taken by the crappy GPU cores. They should have used it for additional logic or cache to speed up the CPU, atleast for the K series CPUs since the people who buy them use a discrete GPU anyway.

    More cache doesn't necessarily mean better performance, in fact, it can even cause WORSE performance. You realize the processor has to search the cache when it needs information, right? There are many different algorithms for that, but to sum it up, the larger the cache the longer it takes.
  • 0 Hide
    maximiza , February 27, 2012 12:18 PM
    i am looking forward to it.
  • 3 Hide
    belardo , February 27, 2012 12:37 PM
    @hardcore_gamer: read what others have posted, especially fuzznarf. The GPU can be used by software for encoding video as well as decode. So what may take 4 cores 1 minute to do, with the GPU helping out - it may take only 30 seconds (guess example).

    @memadax: "The tick tock model is why intel will always be the innovator, and AMD the follower..."
    AMD was the leader during the P4 vs Athlon XP~64~x2 era in every technical way. Intel's huge market share kept AMD out of many markets (Dell) for which intel was sued and lost.

    When intel did this, they kept AMD from making profits which in turn means more R&D... intel illegally kept AMD out of the market. By the time Core2 came out, AMD was approaching 30% market share - go into Office supply stores and 4 out of 5 desktops were AMD.

    Core2 knocked the hell out of AMD and with its lower price, they kicked AMD in the balls... over and over.

    AMD's choice to do a P4 type tech in their FX Chips was stupid... and why many AMDers have gone to intel. AMD still makes good products... just forget the high end.

    @Kartu: Intel out sold AMD 4 to 1. Currently AMD still retains a 20+% share of the market.
    This is actually PRETTY damn good considering.
    A) AMD doesn't have a top performer
    B) AMD doesn't do ANY ANY Advertising... when was the last AMD TV Commercial? I think never.
    C) Intel is very active in adverting. Easy to catch 1-2 intel ads a day...
  • 0 Hide
    hardcore_gamer , February 27, 2012 1:02 PM
    belardo@hardcore_gamer: read what others have posted, especially fuzznarf. The GPU can be used by software for encoding video as well as decode.


    QuickSync is a separate fixed function hardware for video encoding/decoding, it doesn't use the GPUs in the processor.
Display more comments