Rumor: Intel Whiskey Lake-U parts could offer big boost clock increases

Intel has a slate of new 14-nm parts coming this year, and one of the next stops in the land of a thousand lakes is Whiskey Lake. Intel hasn't announced much about these chips beyond an internal code name, but a new HP spec sheet could tell us more about what to expect from those parts. The sheet potentially reveals the existence of three chips: the Core i7-8565U, the Core i5-8265U, and the Core i3-8145U.

  Base clock

(GHz)

Boost clock

(GHz)

L3 cache TDP
Core i7-8565U 1.8 GHz 4.6 GHz 8 MB 15 W?
Core i5-8265U 1.6 GHz 4.1 GHz 6 MB
Core i3-8145U 2.1 GHz 3.9 GHz 4 MB

The first thing of note about these potential Whiskey Lake parts comes in the peak boost clock department. The i7-8565U's 4.6-GHz peak clock speed represents a whopping 400-MHz advantage over the top-end Kaby Lake-R part, the Core i7-8650U. Likewise, the Core i5-8265U has a 500-MHz advantage on the Core i5-8350U. Those kinds of boosts in single-threaded performance will likely result in noticeable improvements in responsiveness from ultrabook-like systems.

Meanwhile, the Core i3-8145U may enter territory uncharted by Kaby Lake-R parts. Core i3 CPUs are notably absent from that model lineup, so we can't say just how the i3-8145U might improve affordable systems. Going by the base clock and cache configuration, however, this could be a dual-core part with Hyper-Threading. Four-core Core i3s tend to have 6 MB of L3 cache on board.

Graphics options from HP include an Intel UHD Graphics 620 IGP, so the Gen9.5 processor we've known since Kaby Lake seems poised to soldier on for another refresh cycle in Whiskey Lake. Given that we have no official word from Intel on these parts just yet, all of the above information could change, but if it holds, ultrabook shoppers could be in for a nice performance boost. Cheers to @momomo_us on Twitter for the heads-up.

Comments closed
    • ronch
    • 1 year ago

    With these chips, you can have your whisky and drink it too.

    • adisor19
    • 1 year ago

    Perhaps we’ll finally see that MacBook Air refresh we’ve all been hoping for..

    Adi

      • VincentHanna
      • 1 year ago

      Has it been 10 years already?

        • tipoo
        • 1 year ago

        Hahahahah -*starts crying* – haha

    • ronch
    • 1 year ago

    Whisky Lake is the best CPU codename ever.

    • jensend
    • 1 year ago

    Expanded 14nm parts timeline/roadmap:
    2015: Skylake
    2016: Kaby Lake
    2017: Coffee Lake
    2018: Whiskey Lake
    2019: Meth Lake
    2020: Heroin Lake

      • bhtooefr
      • 1 year ago

      [url=https://www.urbandictionary.com/define.php?term=ice<]Meth Lake[/url<]'s on the 10 nm roadmap already.

      • ronch
      • 1 year ago

      I don’t believe you. You must be on meth.

      • Concupiscence
      • 1 year ago

      2021: Freebase Lake
      2022: Crack Lake
      2023: Wood Glue Lake
      2024: Krokodil Lake
      2025: Intel declares bankruptcy

      • Growler
      • 1 year ago

      I’m waiting for Bath Salts Lake, myself. I’m ready to chew someone’s face off already.

      • anotherengineer
      • 1 year ago

      2021: Chuckulake

        • chuckula
        • 1 year ago

        You don’t want that one to launch!

        • Redocbew
        • 1 year ago

        So, the upvotes here would be called up-chucks?

    • DancinJack
    • 1 year ago

    Is there any chance this is a “new” revision of 14nm? I thought there was one more on the roadmap we hadn’t seen products from yet. Or are they just binning these to holy hell?

      • ikjadoon
      • 1 year ago

      All the good bets are that Whiskey Lake is 14nm++ (what desktops have been using for a year, but laptops have been stuck on 14nm+). 14nm++, according to Intel, is ~15% more efficient than 14nm+ for the same performance:

      [url<]https://techreport.com/review/31660/intel-defends-its-process-technology-leadership-at-14nm-and-10nm[/url<] It's not sustained all-core boost of 4.6GHz, but a single core peaking at 4.6GHz for a "short' amount of time. It's more than achievable with process improvements that large. They can't be binning, in any realistic sense, on the bread & butter CPUs in every Intel Ultrabook.

      • Action.de.Parsnip
      • 1 year ago

      Could be! Interesting to see how much you can tweak and squeeze a process if you have to stay with it for years on end. Just rinse and rinse and rinse and still find little bits to use up. Makes you wonder what could have been done with the processes of yesteryear; huge dies/perf. watt

    • Neutronbeam
    • 1 year ago

    Meh. The Whiskey Bottle chips have more kick–and a smoother finish.

    EDIT: took out formatting tags

    • Chrispy_
    • 1 year ago

    Surely this is mostly just a marketing stunt?

    Higher boost has always been an option, just at the cost of shorter boost duration. I would take a higher base clock over these “stunt” boost clocks any day of the week.

    I never used to notice it in the Ivy/Haswell days but modern CPUs with their aggressive cTDP there’s a big difference; 25W on boost is great, but when it drops down to 7.5W after a minute or two, the performance drop is immediately obvious.

    It’s a tradeoff I’m willing to accept because the cTDP down state of 7.5W is great for battery life but I just wish Intel would offer a more realistic cTDP up state that could handle more than a couple of minutes of load.

      • Redocbew
      • 1 year ago

      Since this is specifically for U-series chips I’d have to agree. Presumably you’re not going to have the space and/or the airflow for super efficient cooling or you wouldn’t be using a U series chip at all. That’s probably going to make reaching the boost clock and staying there for an extended duration difficult.

      • TheRazorsEdge
      • 1 year ago

      [quote<]when it drops down to 7.5W after a minute or two, the performance drop is immediately obvious[/quote<] You are probably not the target market for U-series CPUs. If you want the ultralight ultrabooks, you have to accept that the tradeoff is lower sustained performance. If the lack of sustained performance is a no-go for you, then you shouldn't be buying laptops with these CPUs. Intel does have parts with higher performance and TDPs.

        • Chrispy_
        • 1 year ago

        I’ve accepted that these are the constraints of the U-series; I already have a 7200U and genuinely need it for the portability.

        My gripe with the boost clock and cTDP is not that they throttle back after a while, but how quickly they throttle back. 4.6GHz sounds amazing but if the choice was between 4.6GHz for 30 seconds, or 3.2GHz for five minutes, I’d choose the slower option every time.

        The higher the clock, the higher the voltage. Power consumtion is proportional to the square of the voltage so even the tiniest increase in voltage to reach those “marketing stunt” clockspeeds has an [i<]enormous[/i<] cost to the TDP and thus the duration of the boost clock.

          • Duct Tape Dude
          • 1 year ago

          [quote<]if the choice was between 4.6GHz for 30 seconds, or 3.2GHz for five minutes[/quote<]You know that both are exactly what happens, right? You get full boost until the thermals give out, then you get as much as the manufacturer's cTDP can allow. The point is you [i<]always get[/i<] the slower option every time. The faster option is a bonus that lets you burst through short workloads. Besides, by undervolting my 8550U I get another 10-15% sustained clockspeed. As an added bonus, if you disable cores you can go even faster.

            • Shobai
            • 1 year ago

            Perhaps I’ve misread his post, but I think you’re assuming his 3.2 number as a base clock; if the base clock is the same for both, and your task takes enough time to revert to base clock for both systems, then his contention is that he’ll get more done in 5 mins at 3.2 gigglehertz than 30 seconds at 4.6, and then the same amount done per unit time by each system at base clock.

            • Duct Tape Dude
            • 1 year ago

            What you and Chrispy describe is not how TurboBoost, base clocks, and power throttling work though. Of course you’re going to get more done at (300s * 3.2GHz = 960GHzs) vs (30s * 4.6 GHz + 270s * 1.8GHz = 624GHzs). But what actually happens is the CPU runs as fast as it can to run at ~30W for a few seconds, then it cuts the max Wattage to the OEM’s cTDP rating, which can range from 15W to 25W, and runs as fast as it can given that power constraint.

            Intel specs the 8550U for 1.8GHz base at 15W, but realistically I see clocks of 2.2-2.7GHz sustained on 8 threads in various workloads, all while limited to 15W. If I use fewer cores, the clocks remain solidly in the 3.0GHz+ range.

            What Chrispy should be complaining about instead is the fact that he purchased a machine from an OEM that made his U-series cTDP 7.5W instead of the standard 15W or superior 25W.

            • Spunjji
            • 1 year ago

            I think it’s a legit complaint to make that Intel have allowed for OEMs to sell a nominally faster and definitely more-expensive CPU in a way that won’t actually perform better than a cheaper one. You can blame the OEM all you want for the configuration, but Intel allow them to use the same name for different performance.

        • Eversor
        • 1 year ago

        There is a massive difference between a 8W U series and a 15W one. It is the same for Ryzen U and it is very bad that when you’re buying something you don’t know if you’re getting full or half performance devices.

        • Spunjji
        • 1 year ago

        Devices with these CPUs are frequently advertised as being great for photo editing on the go. Apple have built an entire brand of laptops around that concept.

        It’s bullshit, because they overheat and throttle and the application stutters. That peak clock speed gets kicked off by moderate load and immediately saturates the cooling solution, dropping your clock back down, providing more thermal headroom for the rebound and boom, back you go again.

        It’s not actually the sustained loads where you regress to conservative turbo clocks that are most problematic. It’s the ones that cause you to yoyo back and forth between an obviously unsustainable boost and the insufficient base clock.

      • dodozoid
      • 1 year ago

      If you require full throttle for more than a second from this CPU, you are using a wrong machine…

        • bitcat70
        • 1 year ago

        Well, then what’s the point of full throttle here? Sounds like a marketing gimmick.

          • ikjadoon
          • 1 year ago

          Well, snappiness. Single-threaded performance largely affects how snappy a system is in day-to-day usage. The time to load a website, which is what these 15W systems do for about 80% of the day, is largely dictated by 1T CPU performance on the hardware side (and ad quantity, probably, on the software side).

          4.6GHz snaps go by 10% faster than 4.2GHz snaps.

            • DancinJack
            • 1 year ago

            I’m not sure the website load times use case is the proper example here, but your point is still mostly accurate.

            • dodozoid
            • 1 year ago

            Indeed, system responsiveness is greatly affected by maximum attainable single thread performance and by the time it takes the core to reach said performance or frequency.
            As much as I root for AMD, intel is probably still miles ahead in this regard in mobile processors.
            It would be nice to have a benchmarking tool that would show how fast the CPU reacts to burst load – are there any such tools? And if so, how does RR compare to the likes of recent rehashlake?

      • ikjadoon
      • 1 year ago

      It’s almost assuredly due to 14nm++, though. Whiskey Lake should be the first Ultrabook 15W CPUs built on 14nm++ (double plus—good?), where 4.5GHz boost on Kaby Lake was limited to 28W-class CPUs.

      14nm++ actually has a wider gate pitch than 14nm+. From Anandtech, “This [wider gate pitch] allows for higher leakage transistors, meaning higher peak power and higher frequency at the expense of die area and idle power.”

      If there are any mitigations translating 14++ nm to mobile, we’ll see. But Intel’s benchmarks of 14nm++ show another shift in power/effiency similar to 14nm+ from 14nm. It’s not just ramming more amps in a shorter amount of time.

      [url<]https://en.wikichip.org/wiki/intel/microarchitectures/coffee_lake[/url<]

        • derFunkenstein
        • 1 year ago

        Keep adding + modifiers and eventually you go up to 15nm.

          • Mr Bill
          • 1 year ago

          Skylake Big & Tall edition confirmed!

      • DPete27
      • 1 year ago

      I have some recent first hand experience with an i5-8250U that might be useful:
      Asus K570UD – i5-8250U + GTX1050 2GB

      Prime95 only:
      Maintains all-core boost frequency of for between 30-60s
      Settles in to 2.2GHz for steady-state all-core frequency. Ran test for at least 5 minutes.

      Prime95 and Furmark:
      Maintains all-core boost frequency for 30-45s
      Settles in to 2.2GHz for steady-state all-core frequency. Ran test for at least 5 minutes.

      I should add that the shipping BIOS had steady-state frequency at 1.9GHz, but the latest BIOS has apparently upped that to “Improved System Performance”

        • Chrispy_
        • 1 year ago

        Sounds about right. I’m impressed Prime ran for more than 30s at boost. Mine doesn’t even manage that but perhaps I was dealt a poor hand in the silicon lottery.

        Under 1 minute definitely seems like the typical result though. Rules out most kinds of IGP gaming, rendering, encoding, transcoding, and flash/java content. I noticed a loss of scrolling smoothness halfway through reading a PDF the other day (admittedly a vector-heavy, 50-page PDF) and that’s something I’ve not experienced in [i<]a while[/i<].

      • Duct Tape Dude
      • 1 year ago

      OEMs set the cTDP, not Intel.

        • Spunjji
        • 1 year ago

        Intel specifies the cTDP they can use. It also fails to specify nomenclature for revealing that cTDP to the customer.

        They could just have, you know, a specially named series of chips for low TDP uses. But nobody liked the Core m chips, so now they sell you them by stealth.

    • tipoo
    • 1 year ago

    Not sure how this will go. I did the math on some Coffee Lake parts after the macbook pro throttling thing, and it seemed their peak power use in the highest boost state had gone up 24% in order to get an 11% clock speed increase. To me this shows a desperation to push things along on a stagnant architecture in lieu of IPC gains or fab node gains, and an abandonment of Intels old rule that a new feature had to increase performance by twice as much as it increased power draw.

    Indeed, that ratio is flipped.

    So I’m imagining these parts could similarly use their boost overhead early on, while not impacting anything lasting more than seconds, while also eating their turbo time faster and so not impacting much beyond a page load. We’ll see.

      • derFunkenstein
      • 1 year ago

      Arguably if you need more than a few seconds of all-out performance, you need a desktop

      [quote<]Intels old rule that a new feature had to increase performance by twice as much as it increased power draw[/quote<] We'll have to blame that on the 10nm process not being ready yet. Intel is out of options so more cores, more speed, and more power consumption are all that's left.

        • tipoo
        • 1 year ago

        [quote<]Arguably if you need more than a few seconds of all-out performance, you need a desktop[/quote<] If you're mostly stationary, and even then we often only have so much influence on our corporate buying decisions. The problem is the higher max boost state just eats itself even faster for its disproportionate power use vs speed increase. You're right that the 14nm process has been drawn out far too long, and 10nm's early results looked even behind 14nm+++ on that one weird lenovo laptop with it.

          • bhtooefr
          • 1 year ago

          ATX and Micro-ATX luggables still exist…

            • oldog
            • 1 year ago

            I remember when my best friend brought one of these 28 lb babies home from work in 1985 or there bouts.

            [url<]http://oldcomputers.net/compaqi.html[/url<]

            • chuckula
            • 1 year ago

            Mobile Threadripper 2…. CONFIRMED!

            • JustAnEngineer
            • 1 year ago

            Can I get that with SLI’ed Titan V’s?

            • Redocbew
            • 1 year ago

            That’s the only configuration that makes sense.

            • bhtooefr
            • 1 year ago

            …I don’t think you can SLI them?

            Gotta go for Quadro GV100s (the NVLink connectors on the Titan V are blocked off at minimum), I think.

            But, otherwise… several of these could hold that much hardware: [url<]http://www.acmeportable.com/portable-systems[/url<]

      • Sahrin
      • 1 year ago

      This is what happened last time Intel was getting competitive pressure from AMD. They jacked up clock speeds and blew thermal budgets – this is where Intels “TDP but not TDP” originated. They couldn’t publish the actual thermal spec (because no one would buy a consumer processor that threw 200W) so they “smoothed out” the peaks and redefined TDP.

      They didn’t call it “Presshot” for not reason.

      • DavidC1
      • 1 year ago

      The Brian Kraznich management was shoddy and overconfident. They must have believed 10nm would execute without issues, and didn’t have a contingency plan in case that failed. Something like a better architecture based on 14nm could have been here now. But no. Based on the fact they have to rehash-Skylake chip in 2018, it could have been late as 2016 when they held onto the belief 10nm would deliver.

      10nm issues are likely due to the same faulty management too. This will take years to fix.

      • Eversor
      • 1 year ago

      Very relevant post.

      This seems to be even worse in the Braswell to Apollo Lake transition. Braswell quad can keep full turbo for minutes, rarely dropping to base, and peak voltage is around 0.8v.
      Then you put similar loads on a *dual* Apollo Lake and it goes slightly below base as soon as you put any kind of light load on the GPU. The whole SoC uses a lot more power for what is a newer generation and Turbo peak voltage is in excess of 1v – above even what is used on similarly clocked dual Core U chips.

      Something is very wrong in certain parts of newer Intel chip/process design. This reminds of the issues AMD had with Bobcat. They were using TSMC 40nm and pushing 1.35v (or more) at almost 1GHz less than what AMD achieved on K8/K10 running 1.25v or less.

        • tipoo
        • 1 year ago

        In both cases they’re burning the boost candle at both ends to appear to have a generational improvement, but making it burn out even quicker for it.

        This is no contingency plan, pants down Intel at its worst.

        • DavidC1
        • 1 year ago

        Apollo Lake is fine, because at the same process it was 30-40% faster. It’s believed done optimally the gain is square root of the die and power increase, so what Apollo Lake did was lot better than that.

        Look at Gemini Lake. Its another 30-40% faster than Apollo Lake on the same process. It even holds Turbo better.

        Early 14nm wasn’t that good, like with Broadwell.

    • DPete27
    • 1 year ago

    The current i3-8130U is 2C/4T at 2.2GHz/3.4GHz.
    You can get one in [url=https://www.bestbuy.com/site/lenovo-330-15ikb-15-6-laptop-intel-core-i3-8gb-memory-1tb-hard-drive-platinum-gray/6221790.p?skuId=6221790<]some pretty cheap lappies[/url<]

    • chuckula
    • 1 year ago

    Less than 32 cores in a notebook?

    Go home Intel, you’re drunk.

      • tay
      • 1 year ago

      Errbody is rippin’ threads out here.

Pin It on Pinterest

Share This