back to article ♫ The Core i9 clock cycles go up. Who cares where they come down?

Owners of laptops fitted with Intel's Core i9 high-performance processor, including computers made by Apple and Dell, are finding that the machines slow down compared to the pace of older models. Chipzilla describes the six-core Core i9 as a "no compromises" chip aimed at gaming, VR and "next-level content creation". Intel has …

Page:

  1. Anonymous Coward
    Anonymous Coward

    Testing Standards...

    Are there no testing standards at Apple or Dell?

    1. Flocke Kroes Silver badge

      Re: Testing Standards...

      Yes: There is a guy with a micrometer. He gently measures the thickness of each new iThing design and it passes if it breaks.

      1. Anonymous Coward
        Anonymous Coward

        Re: Testing Standards...

        Yes: There is a guy with a micrometer. He gently measures the thickness of each new iThing design and it passes if it breaks.

        Don't forget the man and dog team measuring reflectance and fingerprint magnometry.

    2. J27

      Re: Testing Standards...

      Dell? For sure not. I've had access to the last 7 generations of XPS models and they all throttle under heavy load. You can complain to Dell all you want, they don't care because it doesn't seem to affect sales. It's an industry-wide issue, but I can only personally confirm it for Dells.

  2. pixl97

    Tick-Tock

    I am Clockzilla and this is Clock-gate.

    1. Anonymous Coward
      Anonymous Coward

      Re: Tick-Tock

      I don't think so. Thermal throttling is nothing new, nor is reverting to low power states. What's going on here is that the marketing idiots at Intel tried to dress this up as a remarkable new feature, and now its come back to nip them on the bum. Compared to the Puma 6, IME, and Spectre fiascos, this is nothing, although it contributes to that nagging doubt that your laptop ought to say "AMD inside".

      1. Tomato42

        Re: Tick-Tock

        Thermal throttling was part of Intel design since Pentium IV; so yes, this is nothing but markedroids "aiming for the sky"

  3. Anonymous Coward
    Anonymous Coward

    Intel promised OEMs 10nm CPUs

    Then pulled the rug out from under them a few months ago when they admitted they couldn't produce shit on that process. Supposedly they provided OEMs with some "golden samples" of 10nm CPUs earlier this year, but designing your laptop around something Intel can't deliver in anything remotely approaching production quantity won't end well. It is still on the OEMs to have backup plans.

    There have been rumors for years about Apple switching to their own ARM SoCs for their Mac line, this fiasco is just giving them all the more reason to do so. After all, the A11 is faster than the i9 running at its default clock (the speed at which it apparently still throttles) They're still left with the problem that their ARM SoCs running x86 code will be far slower, but they've done 'fat binaries' and ISA transitions a couple times before so Apple knows how to handle that part.

    Dell might need to finally have some serious talks with AMD, since it looks like they will have 7nm CPUs from GF well before Intel can hope to produce any of their 10nm stuff in quantity (foundry 7nm is roughly equal to Intel 10nm, so if they were both out they'd be fairly comparable from a process standpoint)

    1. Anonymous Coward
      Anonymous Coward

      Re: Tempted to downvote...

      I am tempted to downvote for the "A11 is faster than the i9 running at its default clock", but will not if you define "faster".

      Because higher clocks? Nope.

      GFlops? (AFAIK depending on workload, but majority) Nope.

      So not sure what on earth you are smoking. XD

      1. Anonymous Coward
        Anonymous Coward

        Re: Tempted to downvote...

        Because the A11 is faster at equal clocks as Intel's CPUs. OK not in stuff that's heavily dependent on memory bandwidth like Linpack, so the GFLOPS won't remotely compare, but who runs Linpack on their phone? For the type of integer code dominated tasks that are common on both phones and PCs, Apple beats Intel clock for clock.

        They aren't faster of course because Intel CPUs can hit double the A11's clock, as it should given its significantly higher power budget which allows the use of transistors optimized for high frequency rather than low power.

        1. Anonymous Coward
          Happy

          Re: Wait... huh?

          [disengage the cloaking device, it is not hostile captain]

          I'm still confused on the definition of "faster". You are comparing different workloads and different platforms. Comparing mobile to laptop. Which is fine, a compute is a compute. Except you then say we cannot take a general compute, because on mobile we don't often do that (we do web page and mobile apps, if I'm on the right page). But then compare it to a laptop made for "pro" and not "pro webpage surfing", but what people expect to be productivity. Thus general compute.

          So either we need to say the A11 is not the same as the i9, and we cannot compare clocks or workloads. As my casio watch could overclock to 10GHz and be faster than both! XD

          Or we say which type of compute, memory etc, and say it's "faster". I'm happy for that, to define it faster in prime, pi, content delivery, encoding or even just latency of a webpage lookup.

          But just saying "the A11 [at 2.5GHz] is faster than an i9 [at 2.5GHz]" fails totally to pass logical syntax. As you say though, it is faster for some "type of integer code". That is always a "sometimes", never "it *is* faster". Thanks for the info though!

          1. Anonymous Coward
            Anonymous Coward

            Re: Wait... huh?

            The A11 is faster on pretty much ALL types of integer code. Not because it is somehow better designed than Intel's CPUs or due to an advantage of ARM over x86. Its because it is designed for a clock rate of around 2.5 GHz or so, while Intel's CPUs are designed for a clock rate of around 5 GHz (because they use the same cores across their product line and want to be able to reach 5 GHz at least in turbo)

            That implies caches that have more wait states, more pipeline stages etc. which leads to lower IPC. That's the compromise of targeting higher frequencies for max speed, if it runs at half that speed it has unnecessary delays in loads from cache, pipeline stages that don't get as much work done as they could, etc. If Intel designed a CPU targeted with a ceiling of 2.5 GHz instead of 5 GHz it would likely match or exceed the A11's performance at equal clocks.

            The reason I said it isn't going to be faster on floating point is because a lot of floating point code is limited by memory bandwidth. Though an A11 used in a laptop would match x86 in FP, given the same memory speed.

            1. Anonymous Coward
              Anonymous Coward

              Re: Thanks!

              No idea if you get this reply (does Reg do notifications?), but still, thanks so much for that extra info.

              So, if I've understood, it's down to the chip silicone budget? The A11 is putting all the work at 2.5GHz to use. Where as, only the i9s running at 5GGHz would outperform it? Due to the i9's extra budget/code/transistors/cache (and wait times of all these systems).

              I never knew that the A11 was actually a quicker compute too because of this. I can see where/why Intel is panicking even with the likes of AMD creeping up on "lots of slow" in their multicore systems. They don't need to be the fastest, if power and scale can still overcome the lead Intel has on pure silicone speed... which they are possibly about to also lose!

              I knew some of the Apple and other chip makers had the low power chips outperforming Intel on everything (power use and compute power), but it's interesting to see they are now losing out in mid range, and possibly high end too!

              PS, and all of which is partially rendered mute, if Apple (and by extension Intel) had the cooling to even run this at the expected 3.5GHz+. I also remember the stage 5GHZ they "showcased"! LOL!

              1. Anonymous Coward
                Anonymous Coward

                Re: Thanks!

                I'm sure the A11 could be pushed a little if you gave it more voltage in a form factor that allowed for more heat dissipation, but maybe not quite as far as 3.5 GHz since it would haven't been designed with such relatively high clocks in mind. If Apple plans to use their A* SoCs for Macs they might they might design a separate core for the Mac line or they might share it between Macs & iDevices and take a small hit on iDevice performance. If we see a new A* SoC that looks no faster or even a bit slower clock for clock, that will be a big hint that Apple is closing to putting it (probably not that exact SoC, but the big core from it) in a laptop/desktop Mac.

                Of course Apple has the budget to design tweak their design for higher clocks so as not to compromise iDevice performance. Especially if they decide to use that same core in the mountains of servers they are filling their datacenters with. We might not see a server version of iOS running on ARM64 as a product, but that doesn't mean Apple won't use such a thing internally. That would help justify the cost of doing a separate core designed for higher clocks for the Mac.

  4. Flakk
    Trollface

    The Thermal Paste Was a Great Idea, Intel!

    It's so much cheaper than indium solder, reputational damage notwithstanding.

    1. Anonymous Coward
      Anonymous Coward

      Re: The Thermal Paste Was a Great Idea, Intel!

      Mobile CPU... are these not die exposed mobile CPU? So it's all custom cooling as soon as it hits the laptop. No solder, paste or anything except what Dell/Apple put on it.

  5. Camilla Smythe

    Yah... But...

    What's the point when you are trying to frag a Boss if both you and the Boss get really quick for a couple of minutes and then the game goes into Geocities Banner Mode?

  6. The humble print monkey

    Custom coolers

    Silly idea, having spent the day in our air conditioned server closet (no room to sit, but hey, it is cold).

    MBP (whatever model) has a known physical form. Ok, the thermal energy paste, and internal cooling are what has come out of copertino design shrink-farm.

    Would it be beyond the whit of someone to design a form hugging water cooled dock/base that would drop the base ambient?

    There are people who do have to use Apple applications (well, have to, are kind of tied to) fcpx etc.

    I’m only suggesting the Apple water cooled dock, as it would be likely to out sell a custom dell/HP until, and there is a base level of aluminium forming part of the thermal envelope...

    1. Anonymous Coward
      Facepalm

      Good intentions, but you are just like me...

      ... prone to putting mouth before brain when it comes to ideas.

      Think about it. Below ambient temperature = condensation.

      Current form factor = limited thermal dissipation to even couple your customer option too (external air pumped in or external coupling of a heatsink)

      There are already at least one water cooled docking station laptop on the market. But it requires the watercooling loop inside the case to begin with (then is docked to the res+pump+fans+radiator).

      PS, I know I've probably also spoken before thinking long enough on this reply too! ;)

  7. AZump

    The solution is simple.

    Download and run Throttle Stop. I have an ASUS Slate (EP-121) and it would throttle down to 667MHz before throttling up the fans. ...trying to be like Apple in the "noise" department. Well, this made the machine pretty much useless. Couldn't even watch Netflix in HD. Found Throttle Stop, locked multiplier in "Turbo" and ran at a constant 1.78GHz. Never saw temps above 80c and that was during extreme CPU loads playing KSP. Normally, working in my DAW, temps hang in the 50s, but, without Throttle Stop, it won't even get that hot before throttling. Machine is damn near useless without locking the multiplier because 1 minute after boot, it's down to .6GHz.

    Screw throttling! We buy high end CPU's for their power, having them throttle down to 1999's clocks is beyond ridiculous. Especially when the damn things CAN run turbo clocks all day every day. My i7-k tower, 3rd gen, runs at 4.8GHz ALL DAY LONG. At that speed, it idles at 36c in a 76F room. I could push it to 5.2GHz, but, it won't encode video at that speed. Does everything else tho.

    An old P4 runs averages higher clocks at the end of the day than modern CPUs. Hanging at 1.6GHz except when loading an app is pure BS. It's not like I have several thousand of them so what's the big f*ing deal if it eats a few more watts running full bore?

    There is no difference, but it sure makes the huggie-feelie tree hugging brigade sleep better at night.

    Manufactures need to wake up. A powerful CPU needs more power, more power means more heat, more heat means more cooling, more cooling means more heat, and all that adds up to LARGER laptops. Sure, everyone wants small, but those who buy the high end FAST stuff know and accept the thicker machines.

    My last "desktop replacement" laptop weighed over 12 pounds. I took it everywhere. My ASUS eSlate weighs 1.1kg. Know what? Panti-waist reviewers complained about the weight. 2 pounds is too heavy!

    ...and that's the future folks. Processors that can run at 100c+ but throttling down before reaching half that temp, why? Because to do it right, the machine would: consume more electricity which pisses off the environmentalists, be thicker than Apples crap*, and everyone wants to be Apple, also a machine too heavy for our panti-waist youth to carry across the room.

    It is what it is and its our fault for letting things get this far.

    *You could have defended Apple in the past and I'd have agreed to their superiority in certain areas. ...right up until the day they went Intel. Now, it's just a regular PC with a custom "OS". Your i7 with whatever OS version is NOT superior in any way to my i7 running Slackware Linux. Actually, in many ways, it's inferior because frankly, I have 100% control over mine. ;)

    1. Anonymous Coward
      Anonymous Coward

      Re: The solution is simple.

      Sounds good. But kind of defeats the real problem. It's not the throttling down when not being used. This extends battery life, or lowers power use when actually idle. This also generally works fine on older models, unless... the software/hardware is buggy! So it is the over throttling up, of the failure to detect use that messes things up. :(

      Glad you got KSP going though!

  8. razorfishsl

    Do these ass clowns actually think those big blocks of copper used in desktops are just for decoration?

    How the fuck does a portable prototype get all thru development without it being detected

    or did they just use "thermal modling"?

  9. Fading
    Holmes

    Undervolt

    If you are suffering the thermal throttling issue and are on a windows based machine then a bit of tweaking with Intel's XTU can go a long way in helping. If you undervolt the chip step by step you can get the same performance but with lower temps - hence avoiding thermal throttling. YMMV and note: only certain chips are supported.

  10. Anonymous Coward
    Anonymous Coward

    Maybe this is actually Apple's solution to the butterfly keyboard problem? Incinerate any specs of dust before they break the keyboard.

Page:

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like