Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Intel Hardware

Intel Demonstrates 10nm Ice Lake Processor, Promises PCs Will Ship With it Later this Year (theverge.com) 80

Intel announced a major rethink of its chip design back in December, just before it finally delivers 10nm chips for PCs and laptops. At CES 2019 this week, Intel is demonstrating its first Ice Lake 10nm processor that's based on its new Sunny Cove microarchitecture. From a report: Intel is building in Thunderbolt 3, Wi-Fi 6, and DL Boost (deep learning boost) into these Ice Lake chips for laptops and PCs to take advantage of. Intel is now promising that PC makers will have devices with Ice Lake processors on shelves by the end of 2019. At its CES keynote today, Intel demonstrated ODM systems from Pegatron and Wistron, and Dell even joined Intel on stage to show off an Ice Lake-powered XPS laptop that will be available later this year. Dell didn't show the device powered on, but it appeared to be a 2-in-1 device that looked similar to the XPS 13. Intel is also looking to the future, too. The chip giant is planning to use Foveros 3D chip stacking technology to build future chips, a method that allows Intel's chip designers to stack extra processing power on top of an already-assembled chip die. These "chiplets" can be stacked atop one another to form a processor that includes graphics, AI processing, and more.
This discussion has been archived. No new comments can be posted.

Intel Demonstrates 10nm Ice Lake Processor, Promises PCs Will Ship With it Later this Year

Comments Filter:
  • by Anonymous Coward on Tuesday January 08, 2019 @03:10PM (#57926486)

    Come back when your crap a actually works.

    • Re: (Score:2, Insightful)

      by jellomizer ( 103300 )

      Here is the big question. Your current chips probably are vulnerable to the problem. If you are holding off until they fix the problem, you will still be volnernable, and running a slower, outdated, and perhaps non-upgradable computer.

    • It works just fine. Having an irrelevant bug doesn't make it not work, in fact it works quite a bit faster for having it.

  • by Anonymous Coward

    Unless IntelME has some low-level Ring-0 phoning home to do.

    • by jellomizer ( 103300 ) on Tuesday January 08, 2019 @03:17PM (#57926538)

      Overall reduce size of your technology, and power consumption. I don't like the idea of integrated everything chips myself. But they are advantages in integrated.

    • by crow ( 16139 )

      It used to be that a bunch of standard ports were part of the chipset, USB, parallel, serial, SATA, ethernet, etc. I assume that secondary chip is now part of the CPU, so they're now saying that WiFi is on par with ethernet, which probably makes sense. For laptops, it's a no-brainer. For desktops, the added silicon is negligible, and if you don't want it, don't connect the antenna.

      I could see some security concerns with high-security applications, but they're already worried about transmissions from vari

    • by AHuxley ( 892839 )
      One less chip for the brand making and paying for the motherboard to worry about.
      Wake on LAN is now Wake on wifi?
  • Really (Score:5, Insightful)

    by 110010001000 ( 697113 ) on Tuesday January 08, 2019 @03:31PM (#57926646) Homepage Journal
    "Dell didn't show the device powered on"

    Wow. Real impressive.
  • 10nm eh? I'll believe it when I see it. This announcement just reeks of a desperate attempt to try and take the steam out of AMD's 12nm chips that just launched.
    • don't worry, node size has just been a marketing term now for years.

      It does NOT mean half-pitch or gate length is 10nm.

      It's a term for another generation of smaller chip size than the last one. And yeah, the 10nm components on one manufactures chip could be bigger or smaller than competitors 10 nm....or 12 nm.....

      great times we live in, marketing things by buzzwords and hype instead of hard data and specs. (marketing wanks and sales choads should be lined up and summarily shot, to usher in a new age of le

    • It is not that Intel cant do 10nm, its that Intel cant do it economically. So far every 10nm part they have sold has come at a large loss, and this will probably be no different.
      • It's also that they just can't do high power chips with the original 100 mtr/mm2 10nm. Not even uneconomically. They have to rework all their cells to lower the density.

        • Thats an issue for another day for them. With yields so terrible it doesnt matter how great the working chips are.

          Normally things like power/heat issues are dealt with first by binning the parts. Refinement comes after large numbers of parts are made and the binning statistics that go along with it are considered.

          In the case of 10nm, Intel cant even do binning because the yields are so bad.

          What Intel needs to and will do is chiplets. Its the only way forward (only way to deal with the bad yields) and
    • It's a low power chip and relatively small to work around yield issues that Intel apparently still has with 10nm. It would seem that high power chips for desktops and servers just don't work at the 100 megatransistor per square mm density that Intel originally targeted and they will need to dial that back 20-40% to make serious chips.

      IOW, the shitshow continues.

  • The new process involves "sandwiching" an iGPU core into the CPU core. In the past iGPUs were severely constrained because you wanted to keep them relatively simple to keep yields up (since if the iGPU is bad you just bricked the CPU in most cases). It should mean much, much faster integrated graphics. e.g. entry level gaming laptops in the $400-$500 range.
  • by jmccue ( 834797 ) on Tuesday January 08, 2019 @04:59PM (#57927192) Homepage

    Does it still have "ME" ? If so can it be fully disabled ? If disabled can it be validated ?

    If not, "Thanks but no thanks"

    • by tlhIngan ( 30335 )

      Does it still have "ME" ? If so can it be fully disabled ? If disabled can it be validated ?

      If not, "Thanks but no thanks"

      Every one has ME. Even the disabled ones still require ME.

      ME is required to boot the processor (it's not easy to boot modern complex processors with variables clocks and power sources.

      The "ME disabled" firmware simply does enough to boot the processor up and halts, so you lose functionality like dynamic voltage and frequency selection (DVFS), important for mobile processors, as well as t

    • If not, "Thanks but no thanks"

      Have you not bought a CPU since 2007 or are you not aware that every vendor has an ME equivalent running beside their main processor?

  • by Anonymous Coward

    "just before it finally delivers 10nm chips for PCs and laptops."

    That remains to be seen. So far they are only claiming that they will deliver them later.
    Just like they've claimed repeatedly.

  • Dear Intel,

    can you "deep learn" from your Meltdown and Spectre mistakes? Seriously, I have a Celeron from about 2005 with these vulnerabilities.

  • But when will they demonstrate a hot shit processor? What are these stupid fucking names for anyway? It tells me nothing about what this processor is and I have to go look for that info elsewhere
    • by Anonymous Coward

      Answer to your question is the same as the answer to most other questions about why something is no longer fun or intersting or funny. "Lawyers" Really.

      At one point you could come up with cool internal project names. Then someone named a project "Jedi". Internal project name, not external marketing name. Lucasfilm sued them anyways. So now the only names allowed for any project have to come from a map. So it's rivers, lakes, towns, mountains only, etc.......... :(
      So unfortunately your "hot sh

  • This appears to be an enhancement to the AVX instructions, and a 16-bit "brain floating point". Are there any real applications for this on a PC, or is this just a marketing buzzword for toy applications? Personally, if Intel has transistors to burn, I'd rather they burn them finding ways to speed up context switching. That would make for noticeable improvements in PC responsiveness even with clock speeds remaining the same, particularly as the number of cores increase.

Truly simple systems... require infinite testing. -- Norman Augustine

Working...