review intel boosts ultrabooks and fanless pcs with more eighth gen cpus

Intel boosts ultrabooks and fanless PCs with more eighth-gen CPUs

Ahead of the IFA trade show in Berlin, Intel is announcing its latest update for its eighth-generation Core family of U-series and Y-series CPUs. Code-named Whiskey Lake and Amber Lake, these Skylake-derived processors promise mobile users better peak performance and better battery life, as well as a range of platform improvements. Whiskey Lake builds on the core-count boost that Kaby Lake Refresh parts delivered to ultrabooks , while Amber Lake parts represent the first refresh of Intel’s Y-series low-power CPUs since the arrival of Kaby Lake two years ago.

Of late, Intel has clearly defined where it sees the PC in today’s stable of computing resources: as the place where owners will do their most important, most focused, and most creative work. To back up that view, we can point to the performance increases the company delivered with its Kaby Lake Refresh CPUs for ultrabooks. Even as single-threaded performance at the architectural level has stagnated, Intel has been able to add more cores to its processors and boost single-core clock frequencies to get better performance in both lightly-threaded and demanding multithreaded workloads. Those improvements are key to what Intel defines as a generational improvement these days.

  i3-8130U i3-8145U i5-8250U i5-8265U i5-8350U i7-8550U i7-8565U i7-8650U
Base clock (GHz) 2.2 2.1 1.6 1.6 1.7 1.8 1.8 1.9
Boost clock (GHz) 3.4 3.9 3.4 3.9 3.6 4.0 4.6 4.2
Cores/threads 2/4 4/8
L3 cache size (MB) 4 6 8
TDP   15 W

First, let’s look at the three new 15-W parts Intel is launching today. As leaks suggested, the three CPUs launching in the Whiskey Lake family today boost peak performance above all. At the low end of the lineup, OEMs get a Core i3-8145U to play with. The i3-8145U establishes a familiar pattern among Whiskey Lake chips: it gives up 100 MHz on its base clock to the i3-8130U in exchange for 500 MHz of extra Turbo Boost peak.

Moving up the stack, the Core i5-8265U looks poised to supersede the i5-8350U as the best-performing 15-W Core i5 in Intel’s arsenal, despite its nominally lower-end model number versus the i5-8350U. It delivers a 300-MHz peak clock boost over the i5-8350U in the same power envelope, and it only gives up 100 MHz to that chip in the unlikely-to-be-seen-often base-clock department.

Similarly, the Whiskey Lake i7-8565U looks poised to displace the i7-8650U from its position atop the stack of 15-W chips. A 400-MHz boost in peak clock speeds seems well worth the i7-8565U’s 100-MHz deficit at base clocks compared to the i7-8650U. We’ll need to see where customer pricing for these chips end up, but it looks like life with an ultrabook is about to get a lot sweeter, regardless of the price point buyers choose.

Whiskey Lake CPUs, like their Kaby Lake-R predecessors, will support LPDDR3 memory running at up to 2133 MT/s and DDR4 memory running at speeds up to 2400 MT/s.

  m3-7Y30 m3-7Y32 m3-8100Y i5-7Y54 i5-7Y57 i5-8200Y i7-7Y75 i7-8500Y
Base clock (GHz) 1.0 1.1 1.1 1.2 1.2 1.3 1.3 1.5
Boost clock (GHz) 2.6 3.0 3.4 3.2 3.3 3.9 3.6 4.2
TDP (default), W 4.5 5 4.5 5 4.5 5
Cores/threads 2/4
L3 cache size (MB) 4

The Amber Lake Y-series CPU family also lets what used to be called Core m chips benefit from Intel’s under-the-hood tuning. The base Core m3-8100Y now has peak clock speeds closer to 2016’s Core i7-7Y75, and the clocks only climb from there. The i5-8200Y’s 3.9-GHz boost clock should chew through demanding web work much better than the i5-7Y54 and i5-7Y57 before it, while the i7-8500Y’s peak speeds deliver a similar 600-MHz speedup over the i7-7Y75.

Y-series CPUs tend to find their way into premium Chromebooks and other fanless PCs where light and bursty workloads are the order of the day, so the Amber Lake family will likely prove a fine shot in the arm for the performance of those systems.

Unlike Kaby Lake Y-series processors, Intel says that Amber Lake parts will only support LPDDR3-1866 RAM rather than LPDDR3 and DDR3L alike.

Both families of chips integrate a new platform controller hub (PCH) that brings the improvements from Intel’s most recent desktop chipsets to mobile systems. The Whiskey Lake and Amber Lake platforms have integrated controllers for wireless connectivity in their PCHes, meaning that notebook OEMs will be able to choose among Intel’s Integrated Connectivity (CNVi) Wi-Fi modules to let their machines talk to the world. Whiskey Lake and Amber Lake systems will also benefit from native USB 3.1 Gen 2 connectivity.

One might see the Thunderbolt 3 dot above and think that Intel is integrating a TB3 controller into the Whiskey Lake PCH, but that’s not the case. Despite the rest of the orbiting dots being standard equipment in Whiskey Lake and Amber Lake systems, Thunderbolt 3 is only included in the above diagram as a statement of capability rather than a sign that the company is bringing its TB3 silicon on-package. Several editors were led astray by this diagram during our briefing, so caveat emptor.

Since Whiskey Lake and Amber Lake chips include support for CNVi wireless modules, Intel will highlight systems configured this way with an “Optimized for Connectivity” badge similar to the image above. The reason for this is that the top-end Intel CNVi module, the Wireless-AC 9560, supports 2×2 MIMO and 160-MHz channel widths with compatible routers to potentially deliver over a gigabit per second of raw bandwidth. Just how many notebook makers opt to include that module remains to be seen, but the “Optimized for Connectivity” badge should at least allow those looking for the latest and greatest notebooks to pick them out of a crowd.

Intel has compiled a range of things that Whiskey Lake and Amber Lake systems might do better than a five-year-old PC, presumably a starting point for many in the slowing pace of PC upgrade cycles. Foremost among these for many will be battery life. Intel claims its new products will allow systems to run on battery for up to 16 hours (and possibly more with vendor optimizations), a number obtained by testing 1920×1080 local video playback. More mixed workloads might result in lower battery life figures, but Whiskey Lake ultrabook and Amber Lake fanless systems could still deliver more than a workday’s worth of battery for on-the-go computing.

Intel also highlights improvements in video-encoding performance, integrated-graphics performance, wireless network speeds, and features like Amazon Alexa integration as reasons to plunk down the money for notebooks with Whiskey Lake chips inside. Notebook buyers should, of course, evaluate Intel’s performance claims against their own needs, but it should go without saying that a five-year-old mobile PC likely won’t hold a candle to the latest and greatest ultrabooks and fanless systems, in our experience.

Despite those various improvements, the simple fact that Amber Lake and Whiskey Lake chips both deliver substantial increases in peak boost clocks will likely be the single change that will make mobile PC users the happiest. Single-threaded performance improvements are precious and difficult to come by these days, so the fact that Intel has managed to provide those increases without changing the processors’ thermal envelopes much or at all seems like a win-win to us. We’ll doubtless hear about many notebooks with Whiskey Lake and Amber Lake chips inside as IFA kicks off this week, so stay tuned.

0 responses to “Intel boosts ultrabooks and fanless PCs with more eighth-gen CPUs

  1. Have to admit to being nonplussed. Its so confusing. Plus sizing in clothing means its bigger. I blame the developers of C++

  2. Because they are ~5W CPUs using high performance cores.
    If they packed 4 in there with HT can you imagine how low the sustained full load speed would be?
    I have a dual core Y series that under full CPU load with no GPU load manages around 1.2 to 1.5GHz.
    A quad core would be rated at ~1GHz I imagine and at full load with a Turbo speed on 1 core of 4GHz or so.
    You’ve gotta draw the line somewhere and this is sensible.

  3. There you go. Exact same situation as low end phones getting 8 or more A53 cores while high end ones sometimes have less. It takes well under half the die area to perform half as well.

  4. A Skylake core is greater than 4x the size of a single Goldmont core. You can fit 4 Goldmont core + L2 cache + its System Agent equivalent into an area slightly smaller than a single Skylake core with just the core L2 cache.

    Goldmont Plus is probably much larger, but I bet you still can fit nearly 3 in there.

  5. You asked why low end solutions got 4 cores while the Y series is stuck at 2. Cores that are well over twice as large and perform better each are the answer. It’s cheaper to have four tiny cores for ok performance, _that’s your answer_. I don’t get the snark now. That it feels good enough for the most basic of tasks doesn’t change anything when you actually need performance.

    Literally the exact same situation as low end phones sprinkling 8 A53s in, each one may perform half as well as a big core but takes well under half the die space for it.

    [quote<]At any rate, having looked at the architecture a little more, it seems the real answer to my initial question is just "because this is skylake-y+++" so of course they wouldn't change the core configuration on it.[/quote<] Except that's true of the entire 8th gen that has been changing up core counts on the same architecture, so clearly not that reason.

  6. Okay I guess? Yay for numbers?

    But I’ve used both solutions, side by side, and they’re both fine.

    After a certain point the objective figures just become academic – it’s like comparing a card pushing 240fps to one doing 220fps. Sure it’s “10% faster” but at that point… who cares?

    At any rate, having looked at the architecture a little more, it seems the real answer to my initial question is just “because this is skylake-y+++” so of course they wouldn’t change the core configuration on it.

  7. L1TF is Foreshadow

    Extreme measures like disabling SMT is required when the OS can’t be trusted and customers have to share cores.

  8. Does it fix Foreshadow, which has potentially more serious performance deficit when patched? Red Hat takes it so seriously they recommend disabling HyperThreading altogether alongside Foreshadow firmware patch.

  9. And the objective figures disagree with that. Goldmont is slower per core, but the reduction in size over the larger cores is more than the drop in performance, hence more cores being frequent in the cheaper systems.

  10. Performance is moving down the stack since the high-end has started stagnating. Pretty soon we’ll start seeing 6C/12T and 8C/16T in mobile, and then it will be pretty much over except for exotics like Threadripper.

    Hopefully either single-thread starts getting more than generationally faster or software starts to be able to use all these threads soon.

  11. I would have to disagree.

    My NUC7PJYH (Pentium Silver J5005) with 8GB of RAM doesn’t feel any slower than the m5-6Y54 laptop with 8GB of RAM. They’re both fine for everyday use, browsing, video playback, etc. Granted the J5005 has a higher TDP, but clock speeds aren’t much different on the N5000.

    I wouldn’t try games on them, or really anything that required significant horsepower, but that’s not the point of systems based around these chips anyways.

  12. I definitely preferred the older naming convention. It was easier to differentiate single/dual/quad socket chips.

  13. Well they’re [b<]trying[/b<] to deceive customers. But like everything else these days, they're failing at it.

  14. So Intel moves one of their product lines from an almost totally opaque naming convention to something consistent with the rest of the product stack and that’s somehow them trying to deceive customers?

  15. Intel’s just hoping to distract users, investors, and the media with purposefully confusing product lineups. Remember the E3, E5, and E7 era of Xeons? Now you have a Xeon catalogue with so many SKUs and so few differentiating features, IT purchasers and integrators have developed Intel fatigue. They can’t produce 10nm and they hear the footsteps of AMD’s 7nm coming down the hall. Sad.

  16. Those are Gemini Lake/Goldmont Plus…Wouldn’t be surprised if one core on the Y chip was larger than two on Goldmont. Hence cheaper solutions often having higher core counts, because it’s cheaper to sprinkle in lots of tiny cores and get to reasonable aggregate performance, but the user experience is still significantly worse for the single threaded performance.

  17. Intel doesn’t support LPDDR4 on anything afaik, which is annoying but nothing new. So ultrabooks are forced to either use LPDDR3 or give up battery life with DDR4.

    DDR and LPDDR are not interchangeable and yes, LPDDR3 is still very much a thing and in production.

  18. “Intel says that Amber Lake parts will only support LPDDR3-1866 RAM”

    DDR3 is still a thing?

    Is that going to be an issue with these chips, or are they slow enough they wouldn’t benefit from/ need DDR4?

  19. They already did that. Remember when they started selling multiplier-locked CPUs and charging more for them? I believe that was called the Pentium 2.

    edit — I remember another one — remember when they removed the ability to use inexpensive PC100 RAM, forcing people to use RDRAM? That was called Pentium 4.

    Oh wait, another — remember when they dropped support for x86 and charged more? That was called Itanium.

  20. Intel needs to start removing features from its highest end products.
    And charge more for the versions that lack the features.


  21. Amber Lake doesn’t seem like much of an update at all. I’ve got a laptop with a Skylake-Y part in it and I really like it (excellent battery life, normal performance for everyday use) and it looks like these newer parts are still just a slightly higher clock speed?

    When even a $180 NUC used DDR4 and gets a real 4 cores with the Pentium J5005, and the 6W Pentium Silver N5000 is already available, why leave the “i7” Y chip as a 2C4T solution?

    Also still annoyed they didn’t keep these as “m3, m5 and m7” but considering there was no real difference between them, I guess they needed to change the names to make people opt for the more expensive parts.

  22. This. I’m an enthusiast and even I’ve lost track of which dies are in which Lakes. It’s obvious what they’re doing, but it’s infuriating. One die family should be one Lake.

  23. Intel has done this for years as part of their market segmentation strategy. Two easy examples — hyperthreading and ECC support.

    Apple removes features that they think are no longer necessary in order to make the product better. One can certainly debate whether they are right about that. But they don’t remove/disable features for market segmentation reasons. For example, with the audio jack in the iPhone 7 they didn’t just disable the jack in some models so that they could sell models with the jack for a higher price — that’s the kind of move that requires Intel-level courage.

  24. Intel needs to show some courage and start removing features from its chips so that they are insanely great.

  25. I think that’ll rather be this budget 13″ Macbook that’s been making the rumors. Drop to a non-Plus part and maybe drop thunderbolt and you have this “mystery” Macbook out of a lower end 13″ MBP.

  26. That i7-8500Y is nuts. I wonder how much of an impact on real usability that boost clock really has. It’s impressive that they’re able to do it at all, but I can’t imagine its sustainable even over very burst-y workloads in that 5W TDP.

    Also, kudos to Intel for ditching the old Y series naming.

  27. If these 4 core CPUs will make it in the new MacBook Air with an optional 16GB RAM and all for a decent price, Apple will have a new winner on their hands !

    I can easily see us stop buying the Pro and switching over to the Air for all of sales staff. Heck even I would switch to a new maxed out Air if it still supports Thunderbolt and can drive 2 external monitors !


  28. Base MacBook Pro without Iris Plus is a great idea. Prices for the Pros are high and the touchbar is silly and adds to the cost.

  29. Sorry, that comment was not where it was supposed to be. I’ve updated the article for clarity.

  30. Yep, well ARK confirms the table has the correct core and thread counts – they’re both hyperthreaded quads, so surely the i7 is the “best performing 15W quad – core” ?

  31. I want to know the origin story of the notation.

    I saw a mechanical engineer (ME) do this internally on one of our projects after being steamrolled by management to use this notation. I could barely contain myself during a presentation when he deadpanned on a slide already containing +, -, ++, +-, +++, +-+, ++-, +–, “okay so you want a ++++ and +++-?”

    Of course, that notation never made it to a customer presentation as the ME had been using config numbers and version numbers all along. I guess Intel’s engineers weren’t able to disabuse management/marketing.

  32. Whether or not there’s a Lake on the end of it, trying to have english names for everything in this fine-grained a manner is a recipe for confusion. Having Broadwell-E and Broadwell-U (etc) made sense; you know what you’re looking at when you look at those names. (Skylake-S versus Skylake-X is a bad example since those two actually do differ substantially.) Coffee, Whiskey, and Amber tell nothing about the product and would be much better as Coffee-S, Coffee-U, and Coffee-Y.

  33. Skylake, Kaby Lake, Kaby Lake refresh, Coffee Lake, Cannonlake, Whiskey Lake, Amber Lake, Cascade Lake, Cooper Lake, Ice Lake, Tiger Lake… I’m probably forgetting at least one, aren’t I? This is getting ridiculous.

  34. [quote<]Intel launches Whiskey and Amber Lake[/quote<] [quote<]The CPU parts of these new processors are the same Kaby Lake Refresh parts as Intel launched a year ago—just with slightly tweaked clock speeds.[/quote<] Go home, intel, you're drunk.

  35. 15W parts still seem to lack Iris Plus, so maybe still not for the nontouchbar Macbook Pro…This alleged new budget Macbook, maybe, and the 5W parts definitely for the 12

  36. [quote<]Moving up the stack, the Core i5-8265U looks poised to supersede the i5-8350U as the best-performing 15-W quad-core in Intel's arsenal[/quote<] It's not clear, from the table, why you're limiting the comparison to the i5 chips. [Edit] I know it's easy to just click the down thumb and move on, but perhaps you could clue me in to what I've missed. Is not the i7 8565U also a 15 W quad with faster clocks and more cache than the i5-8265U?

  37. [b<]Intel stops development of its 10-nm node[/b<] [quote<]The frighteningly expensive race for the last shrinks of silicon processes has claimed another contestant. Intel announced today that it is suspending development of its 10-nm process indefinitely in order to shift its resources to specialized, continued development of its 14+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-nm and 14++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-nm FinFET nodes. This abrupt strategy change comes just months after the company touted its confidence and the scale of its investments in leading-edge fabrication technology at its Fab Whatever facility in Chandler, Arizona. [/quote<]

  38. Can you throw the Kaby Lake i3-8130U into that chart for comparison?
    It’s a pretty common CPU to see in the lower priced laptops these days.