Intel casually announces that Ice Lake-family 10th-gen Core chips are shipping now

Intel held its Computex keynote Monday night (US time), and if we can be completely honest, we weren't exactly blown away by the company's announcements. There was the Core i9-9900KS that you've no doubt heard about—more or less a Core i9-9900K with "multi-core enhancement" built right in—and some new Xeon E-series CPUs, as well as a few other topics we'll be talking about eventually.

The real news came almost as an afterthought at the end of Intel's show: the company is apparently shipping 10nm Ice Lake-family mobile CPUs to its partners as I write this. There are purportedly 11 different processors coming, ranging from Core i3s to Core i7s. All of the chips on the way at this time are Ice Lake-U and Ice Lake-Y models. That means "ultra low voltage" and "extremely low power" for those who aren't intimately familiar with Intel's nomenclature.

Even though we don't have a detailed list of the incoming CPUs, there's a lot to talk about. These chips, which Intel is calling "10th Generation Core," are built using Intel's second go at 10nm, the so-called 10nm+ process. (Recall that the barelythere Cannon Lake CPUs were the first attempt.) Rather than being yet another rehash of the Skylake architecture, these chips are based on the Sunny Cove core design that the company has been talking up for a while. The graphics in Ice Lake are based on Intel's Gen11 design, and the chips have quite a few other tricks, too: integrated Wi-Fi 6, Thunderbolt 3, and an on-die MIPI CSI 2 interface for mobile camera devices.

Hopefully at some point we'll get to do an architectural deep dive on Sunny Cove as we have in the past. For now, we'll have to be content with what Intel tells us.  The company claims that Sunny Cove offers an average 18% uplift in instructions-per-clock (IPC) compared to Skylake, which is quite significant indeed. Rival AMD's Zen 2 cores are supposedly a tremendous leap over that organization's previous-generation hardware, and those chips are "only" purported to step things up by some 15%, clock-for-clock. 

IPC isn't everything, of course. Keep in mind those words: "per clock." The products on the way are low-power parts for sure, but even at 15 W, a 4.1 GHz max boost clock looks a bit anemic next to the Whiskey Lake-based Core i7-8665U's 4.8 GHz. It will be interesting to see if the slightly lower boost clocks are the result of the fatter Sunny Cove core design or simply a limitation of Intel's beleaguered 10nm fabrication process. We're also interested to see if the bump in IPC can overcome the relative deficit in clock rate.

TechPowerUp got some pics of Ice Lake. They have a better, close-up shot over there. You should check it out.

These chips support dual-channel DDR4 RAM running at up to 3200 MT/s out of the box—no overclocking needed. Yet more interesting is their ability to hook up to four 32-bit channels of LPDDR4X memory running at an eye-watering 3733 MT/s. That extra memory bandwidth should come in handy for feeding the chips' thirsty Gen11 graphics. Intel has made a whole pile of architectural improvements to its graphics parts since Skylake's launch, but none are likely as significant as the 50% increase in functional units. The "GT2" variant found in most "UHD Graphics" parts contains 32 execution units (EUs) in these Ice Lake parts (up from 24 on Gen 9), while certain Ice Lake-U models will come with Iris Plus graphics built with 48 or even 64 EUs.

1920×1080 resolution. Intel claims Gen11 is 80% faster than Gen9, clock for clock.

Intel claims the largest model is capable of 1.12 TFLOPs of single-precision compute, or 2.25 TFLOPs of half-precision. In combination with the CPU's DLBoost acceleration, that's quite a pile of performance for AI inferencing. The implications for graphics are less clear-cut—GPU compute doesn't cleanly correlate to graphics performance—but Intel also says that the rasterizer is capable of shading 16 pixels per clock and texturing twice that number. That puts the top-end Iris Plus configuration squarely in the realm of AMD's Vega 8 GPU built into the (65 W) Ryzen 3 2200G, which isn't bad at all for a 28-W chip.

Other interesting tidbits about the graphics in these Ice Lake processors include their official support (finally) for VESA Adaptive Sync, their "tier 2" support for DX12 variable-rate shading, and their trio of display pipelines that enable them to run three UHD 4K HDR displays simultaneously—assuming your system board has the requisite DisplayPort connections. Sadly, there's nary a mention of HDMI 2.1, but the display controller does support HDMI 2.0b's 10-bit formats.

I wasn't kidding when I said that this news seemed like an afterthought. Besides the fact that it was a very quick announcement right at the end of Intel's keynote, Intel didn't talk about specific models, pricing, availability, or anything else, really. Intel had representatives from its biggest partners on stage to show off their upcoming hardware, but it's anyone's guess when you'll actually be able to buy one. Generally speaking, I'd rather have the technical details in hand and have to wait for the mundane stuff rather than vice versa, but it's pretty unusual that we can't even tell you when to look out for systems packing this hardware with any more precision than "later this year."

Comments closed
    • Ikepuska
    • 2 months ago

    Not commenting directly on the products themselves, but anyone else note how even Intel isn’t playing along with the USB-IF naming scheme. On their slides it’s still USB 3.1 (10 Gbps) and USB 3.0 (5 Gbps) rather then Gen 1 or Gen 2…. except under usb-c it’s 3.1 Gen2 so…. yeah. I hate this naming convention. I get why they do it (Has to do with the way they ratify and carry forward specifications) But I refuse to accept that as a consumer. And their so called Official Consumer Names for the specs are worthless and haven’t been used since 2.0 won out over SuperSpeed or whatever.

      • JustAnEngineer
      • 2 months ago

      My personal corollary to Hanlon’s Razor is that it’s very possible for people to be [b<]both[/b<] evil and stupid.

    • blastdoor
    • 2 months ago

    I’m more interested in the desktop, but I know mobile is the bigger market, so this is probably more important for intel than competing with AMD for desktop and HEDT.

    Also, moving something to 10nm frees up 14nm capacity.

    So… kinda boring for me, but good for intel

    • NovusBogus
    • 2 months ago

    Still waiting for the Xeon W-21xx series replacement. Updated Dell/HP/Lenovo engineering workstations are now about a year late, a particularly egregious situation given the core bump the dirty consumer peasants recently got.

      • ermo
      • 2 months ago

      “dirty consumer peasants”? Have my upvote!

    • ronch
    • 2 months ago

    I don’t know about you guys but AMD being competitive again and Intel responding in kind is really just fun to watch. And if you’re buying, then all the more good for us. Just like the good ol’ days of K7 and K8. Come to think of it AMD lost its way when Jerry rode off into the sunset with his cowboy hat and Ruiz took over. Remember, K10 and Bulldozer were started on his watch. Dirk probably just hoped for the best when he took over and thought the then-ongoing Bulldozer project was well underway and couldn’t just scrap it, and maybe he thought it will be just fine too when it launches.

    • anotherengineer
    • 2 months ago

    Ice just melted off the lakes 10 days ago here.

    I don’t want to hear of anything relating to cold or winter.

    for at least 6 months

    • tipoo
    • 2 months ago

    18% IPC improvement, 4.8GHz/4.1GHz = 117%…Sounds like it’ll just break even with Whisky lake, though hopefully be able to stay at full turbo longer in the same cooling.

    Also curious if their IPC gains are including speculative execution patch losses on older processors?

      • Klimax
      • 2 months ago

      Doubtful. Too variable across workloads and those mitigations can be disabled. Also it doesn’t look like they modified prediction mechanisms beside fixing flaws.

      • DragonDaddyBear
      • 2 months ago

      I’m not that concerned with the performance being even. The real benefit is probably the lower clock speeds and, thus, probable lower power consumption. If the performance breaks even that’s going to be worth it on the mobile side. Let’s just hope the laptop makers don’t go looking at the trend of phones and keep cutting battery sizes or this product will be far less exciting.

        • tipoo
        • 2 months ago

        Oh I’m sure many will take any power savings as an opportunity to cut the battery and make them thinner.

          • Spunjji
          • 2 months ago

          Gotta cut those heatsinks down too! Who wants a moderately slim, quiet, and cool device with solid battery life, when you can have a razor that hisses angrily at you as it boils and seethes? πŸ˜€

          • DavidC1
          • 2 months ago

          That might be true in some circumstances.

          The integration and the dual-FIVR allows for potentially thinner systems and smaller board sizes, so there will be systems with greater battery capacities. When Haswell came out, the Dell XPS 12 increased battery capacity by 6%, while reducing weight by 0.1lbs. We have more choices than ever before! Isn’t that a good thing?

          Also, while we can continue to increase battery life, having it last through the workday is more than acceptable to a vast majority. Significant portion of the population are also influenced greatly by advent of science fiction entertainment that shows smartphones and laptops that are paper-thin.

          I also think 10 hours is a good enough point, and most systems can do this. Personally, I’d like a lighter device in the 2.5lbs or lower as lots of premium models also have the 2-in-1 form factor and makes the tablet portion much more usable.

    • Mr Bill
    • 2 months ago

    Finally! We have something we can cancel!
    .
    .
    .
    I’ll see myself out…

    • Sahrin
    • 2 months ago

    Remember to deduct 16% performance from these numbers for when Intel releases a patch correcting future security flaws.

    • Goty
    • 2 months ago

    I like the fine print in Intel’s IPC slide:

    [url<]https://images.anandtech.com/doci/14436/Blueprint%20Series_May%2016-2019_COMBINED%20FINAL_AnandTech%20%281%29-page-029.jpg[/url<] [quote<]Performance results are based on testing as of dates shown in configuration and may not reflect all publicly available security updates.[/quote<] I'm pretty sure this is just CYA language, but wouldn't it have been better to say something like, "Performance results... reflect reflect all publicly available security updates at the time of publication?" It makes it seem a little shady, which doesn't really make sense given that these chips should conceivably have current security issues fixed in-silicon and that there's no reason to expect Sunny Cove should be impacted to any greater degree than the Skylake-derived architectures by any of the changes.

      • UberGerbil
      • 2 months ago

      I’d bet that’s the language their Legal Dept told them to use. It doesn’t matter what you write: once Legal gets done with it, it always sounds shady.

        • Voldenuit
        • 2 months ago

        Well, they’re adding up to twice as many shaders, so it should sound up to twice as shady.

          • Srsly_Bro
          • 2 months ago

          Well the 5 GHz 9900KS was put off until later this here or in Intel speak #soon. There is a possibility it can be cancelled. We have all year to speculate.

      • POLAR_IT
      • 2 months ago

      [url<]https://www.intel.com/content/www/us/en/benchmarks/benchmark.html[/url<] "Benchmark results were obtained prior to implementation of recent software patches and firmware updates intended to address exploits referred to as "Spectre" and "Meltdown". Implementation of these updates may make these results inapplicable to your device or system." Really?

        • psuedonymous
        • 2 months ago

        If anything, that would show Ice Lake in a [i<]worse[/i<] light than reality (where the chips they are comparing to would have lower performance after patching, while Ice Lake with hardware mitigations would not). Unintentional sandbagging.

        • Redocbew
        • 2 months ago

        It’s boilerplate. Only the most zealous of fanboys will trust vendor supplied benchmarks, so nothing’s really changed here. This time we just happen to know exactly why not to trust them.

      • Srsly_Bro
      • 2 months ago

      And if some dork forgot to install one patch, the original claim is false with wording to include all security patches. The language used is cya and appropriate for the circumstances. Easier to write that than it is too be 100% sure you have all patches installed. Not nefarious as I would usually claim, but playing it safe.

        • techguy
        • 2 months ago

        It’s also forward-looking. If Intel needs to release further patches, whether to address new security vulnerabilities OR IMPROVE THE PERFORMANCE OF EXISTING SOLUTIONS – this language covers these scenarios.

    • Srsly_Bro
    • 2 months ago

    Intel’s word for all things:

    SOON

      • shaq_mobile
      • 2 months ago

      Better than a valve soon!

        • Pwnstar
        • 2 months ago

        You mean “never”, right?

    • The Dark One
    • 2 months ago

    Not sure it makes sense to compared this chip’s IPC gains vs Skylake, a nearly-four year old architecture, to the gains between Zen+ and Zen2.

      • DancinJack
      • 2 months ago

      Welcome to marketing speak. Any advantage they can make by blurring the lines, they will.

      Having said that, Skylake (and Kaby/Coffee) were VERY similar in arch. But I agree, it’s dumb.

        • Goty
        • 2 months ago

        There were literally NO IPC changes between Skylake and Kaby Lake and, while I can’t find tests to support this, supposedly there were none with Coffee Lake either, so Intel doesn’t really have anything to gain by saying Skylake vs Coffee Lake or Kaby Lake.

        Maybe they went with Skylake because of Skylake-X in the HEDT parts? I dunno, I’m really grasping at straws with that one.

          • Krogoth
          • 2 months ago

          Kaby Lake just updated the iGPU and added SGX support. Coffee Lake was just adding more cores and expending the L3 cache pool. Coffee Lake R implemented some hardware-level mitigations for Meltdown/Spectre.

          Ice Lake is the first considerable change on Intel’s desktop-front since Skylake.

          OTOH, Skylake-X is bit of a different beast with its meshed topology and UPI.

      • deruberhanyok
      • 2 months ago

      Skylake was their previous generation though. They just re-used it a bunch of times and added a + on the end.

      • frenchy2k1
      • 2 months ago

      Skylake is the last “real” architecture from Intel.
      Remember, Ice lake was supposed to be out years ago.

      Reportedly, intel has a very strong coupling between architecture and process. Skylake was supposed to be the last architecture on 14nm and everything after that was to be on 10nm.
      One of their engineers claimed he asked intel to port Ice Lake to 14nm as 10nm was repeatedly delayed, but they refused, as management thought 10nm was “just around the corner”.

      So, all the CPUs since have been Skylake derivatives, mostly trading additional space (more core, more graphics) for more performances, but on the same architecture.

      In that regard, the IPC comparison is warranted.

        • Klimax
        • 2 months ago

        Ice Lake is quite unusual in that it increased a lot of large OOO resources and L1. Especially SRAM cells for L1 are massive (usually 5 or more transistors per bit). Might be one of reasons why Intel went for very aggressive 10nm node that bit them hard.

      • tipoo
      • 2 months ago

      IPC has been the same since Skylake. It’s the clock speeds that rose.

      • Zizy
      • 2 months ago

      I agree – Intel took 4 years to improve IPC by 18% and dropped clocks by 20%. AMD took 1 year to improve IPC by 15% and kept clocks essentially the same. (numbers according to AMD and Intel = I wouldn’t trust either, waiting for independent tests).

        • ermo
        • 2 months ago

        On the other hand, the increased IPC for the architecture will likely be a boon once intel gets its 10nm process under control and can begin to make tweaks and refinements that allows it to boost clocks again.

    • DancinJack
    • 2 months ago

    Exactly like I said with the AMD slides, I care next to nothing about Intel’s either.

    Hopefully someone (even just a site visitor) can get a hold of one of these bad boys and give us some real numbers.

      • DancinJack
      • 2 months ago

      oddly, no downvotes on this one, heh.

        • LoneWolf15
        • 2 months ago

        Have you considered the downvotes are for attitude instead of analysis?

          • DancinJack
          • 2 months ago

          Yeah, sure, people on the internet get angry when they see things, factual things, they don’t like. boo hoo. downvote me more, mates!

            • Spunjji
            • 2 months ago

            “Am I so out of touch?”

            “No, it’s the children who are wrong.”

            • derFunkenstein
            • 2 months ago

            I have done my level best to not be that way but sometimes the children really ARE wrong. πŸ˜†

            • Spunjji
            • 2 months ago

            I recently discovered that once you reach a certain age, it becomes impossible to tell the difference xD

        • ronch
        • 2 months ago

        You sure? πŸ™‚

        • Srsly_Bro
        • 2 months ago

        You would be surprised, apparently.

        +1

      • ronch
      • 2 months ago

      And the thread continues….

        • DancinJack
        • 2 months ago

        FINALLY!

          • ronch
          • 2 months ago

          I reckon the wait wasn’t that long from the last AMD article…

      • Spunjji
      • 2 months ago

      Notebookcheck have laid their hands on some Geekbench (ugh) numbers from the XPS 13 7390 that appear to show performance parity with the i7-8559U on the CPU side, with improved GPU compute numbers.

      I’m suspending full judgement until we get *reliable* benchmarks, but this sort of result isn’t out of bounds given the apparent clock speed trade-offs. I also think it’s probable that more sophisticated benchmarks will show greater benefits from the architecture tweaks.

    • WhatMeWorry
    • 2 months ago

    This is what bugs me about Intel: Compare the two side-by-side diagrams of the U-Series and the Y-Series. They are completely identical except for “10 lanes” versus “12 lanes” and the addition of “DDR4 3200” for the U-Series. Really Intel? You force me to visually inspect and compare 13 icons for only 2 minor differences. This isn’t “Where’s Waldo”. They could have had just one, larger diagram with one or two asterisks. But no. Intel has to make everything so needlessly complicated.

      • DragonDaddyBear
      • 2 months ago

      Have you seen their SKU lineup?

    • kcarlile
    • 2 months ago

    Oh geez, if it’s a new core architecture (lately known as tock…) why are they naming it with a *Lake name? That’s just … amd/nvidia level confusing.

    • derFunkenstein
    • 2 months ago

    Not that these processors are going to wind up in phones, but the image processing unit is ALREADY behind. 1080p120 and 4k30 is half of what an iPhone 8 can do, and the A11 Bionic SoC is nearly two years old already.

      • DancinJack
      • 2 months ago

      good thing we have tons of content for those capabilities!

      /snark

        • derFunkenstein
        • 2 months ago

        That’s what I led with.

      • Zizy
      • 2 months ago

      The problem is that nobody tells you what image quality you can expect from the unit, and almost no one bothers to test. If this CPU offers 10% smaller files at the same quality, it could even have industry leading image encoding.

        • derFunkenstein
        • 2 months ago

        Isn’t file formatting basically always handled on the software side? That would include turning the raw input into a JPEG or whatever.

          • DragonDaddyBear
          • 2 months ago

          Good point. But the RAW data is still handled by the sensor processor.

            • derFunkenstein
            • 2 months ago

            Sure, and so Zizy’s point about IQ stands. It just doesn’t have anything to do with file sizes. Hopefully the raw data is [i<]larger[/i<], but at least some of that is going to rely on the sensor's capabilities. There are tons of moving parts here, and processing raw pixels is a big part of that.

          • psuedonymous
          • 2 months ago

          [quote<]Isn't file formatting basically always handled on the software side?[/quote<] For mobile: absolutely not, encoding is done with a hardware fixed-function block. Only practical way to get it done at low power draw. For laptops/desktops: not if you're using Intel's Quicksync, or Nvidia's or AMD's GPU-accelerated solutions. For non-mobile where you have the option of software encoding, you'd probably choose hardware encoding for laptops (massive power draw improvement) and for applications where you want to minimise CPU load (e.g. screencasting), and software where you wanted to optimise for encoded video quality.

          • willmore
          • 2 months ago

          Sorta yes, sorta no. The ISP (Image Signal Processor) block often does the JPEG compression and it’s a mixed bag of fixed function hardware and firmware. They tend to produce pretty bad JPEG files—a simple decompression and recompression with better JPEG encoder can produce a file with identical visual quality, but at a fraction of the size.

          Some image sensors have a rudamentary ISP on them and can actually provide JPEG formatted images to the host.

    • techguy
    • 2 months ago

    “Shipping now” means shipping to partners for validation. We’ll see products on store shelves in ~6 months.

      • derFunkenstein
      • 2 months ago

      All you gotta do is read the article and you’d realize this comment is redundant. πŸ˜‰

      examples:

      [quote<]the company is apparently shipping 10nm Ice Lake-family mobile CPUs to its partners as I write this.[/quote<] [quote<] it's pretty unusual that we can't even tell you when to look out for systems packing this hardware with any more precision than "later this year."[/quote<] Not to mention it goes without saying. Intel doesn't ship CPUs to end users.

        • Redocbew
        • 2 months ago

        And that end users typically don’t buy mobile processors. Is it even possible to buy a U series chip at retail? There might be some OEM reselling them like what happens with memory and SSDs sometimes, I guess.

        Edit: Ark lists the Core i7-8665U as having BGA packaging, so that rules that out.

          • psuedonymous
          • 2 months ago

          About the only way these make their way to consumer hands is via the Chinese market, where there is a weird penchant for taking mobile BGA chips, and then producing domestic substrates (and custom IHS!) to allow them to be installed into LGA sockets.

          [url=https://shop33776908.world.taobao.com/shop/view_shop.htm?spm=a1z09.2.0.0.67002e8dLn2Y1h&user_number_id=38452638<]Yes, really[/url<].

            • Redocbew
            • 2 months ago

            Yeah that is weird, and google translate never fails to be amusing.

            [quote<]Magic change 4 generation i7 1150 CPU 4770HQ 4870HQ 4980HQ 4750HQ rotary [/quote<] A "magic change" is exactly what everyone wants, right?

            • Goty
            • 2 months ago

            I’d love to get one of those chips and find a mobo/BIOS capable of overclocking them. That would take me back to my AthlonXP-M days!

            • Spunjji
            • 2 months ago

            I loved my Athlon XP-M. One FSB tweak and I had the equivalent of an Athlon XP 3200+, didn’t even need to alter the voltage. It was great until I swapped it into a Micro ATX motherboard and cracked the die in the process. D:

            My next system after that used a Pentium M 1.6Ghz with a pin mod to get it running at ~2.2Ghz – I cooled it semi-passively under a modified copper Zalman Flower cooler. Truly, those were the days.

            • Usacomp2k3
            • 2 months ago

            I loved my XP-M 2400. Ran like a rockstar under the R120, one of the first AIO watercoolers on the market.
            [url<]https://pcper.com/2005/05/cooler-master-aquagate-mini-r120-water-cooling-kit/[/url<]

        • Srsly_Bro
        • 2 months ago

        SOON

      • Srsly_Bro
      • 2 months ago

      SOON

      • Spunjji
      • 2 months ago

      I thought this too, until I spotted that the Dell XPS 13 7390 is due out in early/mid July with these processors inside. It’s not clear whether that signals broader availability or just that Dell got the drop on the first batch, but it’s something.

      On the other hand, early performance indications for the CPU side of the i7-1065G7 look nearly identical to the i7-8559U, so this may not be any kind of big deal.

        • Spunjji
        • 2 months ago

        …the Dell laptop just got bumped back to August. Hard to say whether that’s due to CPU shipment issues or one of the myriad reasons a product like that can be delayed, but it’s still the only one I’m aware of with a release date.

    • UberGerbil
    • 2 months ago

    Better late than never?

      • Ikepuska
      • 2 months ago

      Better Lake than River?

    • Neutronbeam
    • 2 months ago

    Ice Lake? Cool!

      • ermo
      • 2 months ago

      Underappreciated comment of the day.

    • DragonDaddyBear
    • 2 months ago

    I’m actually more excited about this than Ryzen. Don’t get me wrong, Ryzen is awesome and a long time coming, but this has more applicability to my situation. I’m in the market for a new laptop this fall for a masters program I’m planning on starting. Having Thunderbolt built into everything is very appealing. Makes me think the rumors of a Surface with USB C are true. The XPS announced this week looks really nice, too.

      • UberGerbil
      • 2 months ago

      How late can you wait? We’re days away from Q3; for volume shipments in September — ie the beginning of Q4 — you’d expect there to be a lot more noise about this already (demo machines everywhere, leaks months ago).

      I agree that for the mobile segment — which is a big deal — these chips are more relevant than Ryzen 3xxx… at least until AMD puts a GPU chiplet under the hood (at a reasonable total TDP). But this architecture and node has been so long in coming that Intel might as well be in [url=https://statesymbolsusa.org/symbol-official-item/missouri/state-nickname/show-me-state<]Missouri[/url<].

        • DragonDaddyBear
        • 2 months ago

        I can wait a while. I can limp along with the Lenovo Miix I have but it’s going to suck because the screen is cracked and the touch pad is terrible, meaning a mouse comes along everywhere. But I’d rather suffer for now and make sure I get what I’ll like later on.

      • SecretMaster
      • 2 months ago

      Is this a funded M.S. or are you self-funding? If you are funded, see if your PI can get you a computer to use as a stop-gap. Alternatively some university libraries do laptop loans, and you might be able to get a semester loan.

        • DragonDaddyBear
        • 2 months ago

        Uhh, no idea what you’re talking about there. I’m GI Bill and going through WGU if that helps. (I did my BS degree from them and it was freeking awesome.) Really, I’ll be fine. I don’t have that much of a problem. It’s mostly papers at this point, so I could even use my work laptop to type stuff up if I HAD to.

        • shaq_mobile
        • 2 months ago

        Can confirm we do loaners, and lots of IT have their own loaners so you can Dodge the standard beat up ones.

Pin It on Pinterest

Share This