back to article AMD performance plummets when relying on battery power, says Intel. Let's take a closer look at those stats

Intel has taken fresh aim at arch-rival AMD, this month briefing tech journos on benchmarks showing Ryzen-powered laptops not only slowing down when unplugged from mains power but also losing to Intel's latest Core microprocessors whether on battery supply or not. The results are seemingly damning, though are they to be …

  1. a_yank_lurker

    Lies, Damn Lies, and Statistics

    Cherry picking favorable statistics has long been a marketing ploy. These bench marks ignore one important bottleneck for many tasks; the user's fat fingers. AMD apparently decided that certain tasks were not critical in most situations were a laptop is unplugged and extending battery life might be much appreciated. The tasks Chipzilla is ranting about strike me as either a task that is not what most will be doing while mobile or if they are time is not as critical. But this is typical marketing BS.

    1. NATTtrash
      Alert

      Re: Lies, Damn Lies, and Statistics

      But this is typical marketing BS.

      Indeed. I suppose it's about time to replenish that popcorn. The timing is also not surprising, especially since now all the tech oriented reviews declare that "Intel has been overtaken officially by Team Red!" with the Ryzen 5XXX series. Seemingly even on gaming, Intels marketing dept. main USP to go to in case of doubt over the last couple of years. And I can't suppress the feeling that the technophile population (does that mean us?) isn't that much their concern. But if the lappy producers start offering AMD based machines, often cheaper than the same model with Intel Inside™, that makes some Inside Intel™ nervous about their end-of-year bonuses.

      Don't you just love the smell of competition in the morning?

      1. Anonymous Coward
        Anonymous Coward

        Re: Lies, Damn Lies, and Statistics

        But if the lappy producers start offering AMD based machines, often cheaper than the same model with Intel Inside™, that makes some Inside Intel™ nervous about their end-of-year bonuses.

        Don't you just love the smell of competition in the morning?

        I take it that everybody knows what happened last time around?

        It's notable that this time around AMD kit that has a USP of excellent price/performance is consistently priced considerably above the Intel prices on the main suppliers websites, on one site at the moment a Ryzen 3 laptop is almost double the cost of the intel competition and £50 off the cost of a Ryzen7 box.

        There is a smell, but it ain't competition and given the players I would be surprised should we see honest competition rather than sordid skulduggery and anti trust lawsuits again.

        Don't get me wrong; i'd be delighted to see it. I just don't expect to.

        1. Roland6 Silver badge

          Re: Lies, Damn Lies, and Statistics

          >It's notable that this time around AMD kit that has a USP of excellent price/performance is consistently priced considerably above the Intel prices on the main suppliers websites,

          That, I suspect, is a function of the time of year and the need to get rid of stocks of intel-based systems that AMD have effectively made obsolete...

          Circa 2 months back Lenovo were offering a Ryzen 7 Pro 4750U based Thinkpad at the same price point as their Intel i7 10510U version, namely circa £900+vat. Naturally, the AMD variant quickly sold out...

        2. Blazde Silver badge

          Re: Lies, Damn Lies, and Statistics

          This isn't an area where price of mainstream products benefits from competition. There's so much differentiation cost is always whatever the market will bear.

          What you get for your money however.. after years of stagnation we have 8-core laptops and 'gaming laptop' that is actually a laptop and not a laptop-shaped desktop is suddenly a thing.

          Oh for sure it's won't be honest competition, but the dishonest kind is a whole lot better than none at all.

    2. Zola

      Re: Lies, Damn Lies, and Statistics

      Oh the irony - didn't Intel announce a few months ago that benchmarks were not a good way of determining processor performance? Oh look, yes, yes, they did...

      (I tried to find the same article on anandtech.com but surprise surprise the Intel shill site didn't think it was worth covering, just as they don't talk about Intel vulnerabilities, or any Intel performance problems from the resulting mitigations. They're more than happy to discuss the merits of benchmarks - and their abuse - on ARM, however.)

      I wonder if Intel - by using possibly flawed and/or contrived benchmarks - are simply trying to prove their rather untenable point?

      1. Anonymous Coward
        Anonymous Coward

        Re: Lies, Damn Lies, and Statistics

        Like cherry picking tasks that are probably un-optimised legacy code?

        "converting a PowerPoint presentation to PDF", "...or performing an Outlook mail-merge"

        That code probably dates back to 1995.

        1. Mike 137 Silver badge

          Re: Lies, Damn Lies, and Statistics

          "un-optimised legacy code"

          Is the currently ubiquitous bloat of most applications "optimised" then?

          I suspect not in the strict sense of the word. It's amazing how we used to be able to do word processing and spreadsheet calculations on a x486 at 20MHz with 640kB of memory, but on a modern gigahertz machine I still wait perceptibly for MS Office menus to open.

          Optimised? I work on PIC microcontrollers (64kB, 40MHz) so I've done a bit of optimising in my time and I think I'd recognise it.

    3. Roland6 Silver badge

      Re: Lies, Damn Lies, and Statistics

      looking again at the graphs, I suspect intel are trying to distract from the appalling power management capabilities of their cpu's. As from looking at the graphs we see very little difference between mains and battery performance of the Intel cpu's. Whilst at one level this would be expected, given performance benchmarking tests are being run, at another it brings into question, which takes priority: power management or available cpu clock cycles.

      Thus I wonder whether AMD have provided better power management controls to the OEM firmware/Windows OS which are then being used and these are taking precedence over the cPU demands of the benchamrking software; I note Windows reports my Rzyen 7 (6 physical cores, 12 vCores) as having cores 'parked' and whilst on battery being reluctant to fire up/unpark physical cores, something I've not seen on my multicore Intel cpu systems.

  2. IGotOut Silver badge

    Intel.. I've a dirty barrel...

    .. could you scrape the bottom of it for me,, you seem pretty good at the job.

  3. Neil Barnes Silver badge

    I want to know how the processor knows it's running on battery volts

    Do they taste different, or something?

    Seriously, how is this in *any* way a processor comparison? Surely this is down to how the system design - be it software or hardware - that tells the processor how much to throttle depending on power source and battery life. I can't help feeling one could make either chip behave as the other.

    1. Joe W Silver badge

      Re: I want to know how the processor knows it's running on battery volts

      Under Linux this is (or used to be... I have stopped tinkering with the innards a few versions ago) determined by the selected power governor. Which is decided upon by the system depending on acpi events.

    2. Boothy

      Re: I want to know how the processor knows it's running on battery volts

      I'd be curious about the power plan in use in these test.

      I've got a new AMD laptop here, although an older 4 core, 8 thread, Zen+ part from last year, not the newer Zen 2 parts.

      As far as I can see, the balanced power plan, set as default (actually the only one available on this company machine), changes the integrated AMD GPU between max performance, and optimise battery, and the CPU between active or passive cooled.

      As such, I would assume the only slow down for the CPU will be due to thermal throttling, due to passive cooling when on battery. So things like background room temperature are going to pay a part as well.

      Other than those, no other main CPU/GPU related performance settings seem to exist.

      Obviously the Zen 2 parts might use a different plan. Would be interesting to compare.

      1. Anonymous Coward
        Anonymous Coward

        Re: I want to know how the processor knows it's running on battery volts

        Yeah, the logic board and BIOS will have an impact on this as well, and as the Author points out, most of these machines are optimized to max battery live. For the on-mains testing a pure benchmark shootout makes sense. On battery, it's complicated, as performance per watt is is a bigger factor.

        Testing that is a PITA though, as due to the differences in architectures, its more apples - pears then apples-apples. Even if you tuned them close to the same baseline and ran their batteries flat repeating the same workload over and over, you'd still get skewed results as each platform has difference trade offs and "sweet spots". And the logic board an battery would have more effect then the cpu in all likelyhood.

        I guess you could try to test the number of runs a laptop could complete on a full battery, and how fast it completed them, which would at least be somewhat useful in comparing the complete laptop platforms, but still wouldn't eliminate possible bias on a pure cpu comparison.

      2. MrReynolds2U

        Re: I want to know how the processor knows it's running on battery volts

        Are you speculating that Intel forgot to put 50p in the AirCon meter when they ran the AMD tests?

        Shirley not :)

    3. phuzz Silver badge
      Flame

      Re: I want to know how the processor knows it's running on battery volts

      "I want to know how the processor knows it's running on battery volts"

      As the other commentators have said, the OS tells the CPU when to use lower power modes (search for C-States and P-States), but what those modes are, and how much power they use, depend on the CPU design.

      Depending on your OS you can probably force your CPU into full-power-and-damn-the-engines mode even on battery. Diagram of that to the right.

      Of course, this swiftly gets complicated. For example, is it more efficient for a particular workload to run a CPU at higher power for a short period of time, or to use a low power mode which will necessarily take longer?

      (I assume that things like limiting the screen brightness when you're on battery, is carried out by the hardware though).

    4. Ken Hagan Gold badge

      Re: I want to know how the processor knows it's running on battery volts

      "I can't help feeling one could make either chip behave as the other."

      Well acording to Intel, you are wrong. It is apparently not possible to tell an Intel processor to go easy on the power consumption because you are on the move. It'll drain the battery anyway. A nice own goal there by Intel, but a pity that none of their PR wonks seem to have noticed.

  4. heyrick Silver badge

    Statistics mean what the author wants them to mean

    Nothing more, nothing less.

    1. Dave 126 Silver badge

      Re: Statistics mean what the author wants them to mean

      Tim Hartford made a strong case for scepticism over cynicism. Whilst people may put out dodgy statistics, that dodgyness will succumb to our sceptical scrutiny. However, there are people who will benefit from instilling cynicism in us, which means we don't bother checking claims and just assume it's all bollocks.

      So, cynicism is not the defense against other people's bullshit that it pretends to be. Only scepticism is.

  5. don't you hate it when you lose your account

    Pinch of salt

    When looking at these sort of stats the amount of salt increases with the size of the company

  6. msknight

    "El Reg has asked AMD for comment."

    "El Reg has asked AMD how long they laughed their socks off for, before they were able to talk straight." ... fixed that for you.

  7. James 51

    I have a thinkpad A285. It's a great little machine but Lenovo have throttled the CPU. The TDP is as low as it can go and the intel chip in the X280 is allowed to consume more power. Can't find the reference but remember reading there was only one heat pipe on the AMD chip v two on the Intel. Other laptops have feature like single channel RAM which hits AMD CPUs harder. It would not be difficult to find a laptop that sips power for AMD and gives Intel every advantage in the performance stakes so I'd take Intel's stats with an ocean's worthof salt.

  8. AMBxx Silver badge
    Coat

    Bug or a feature?

    Is slowing down PowerPoint a bug or a feature?

    1. Big_Boomer

      Re: Bug or a feature?

      ANYTHING that slows down or reduces our exposure to death by Powerpoint is fine by me. Just the word "meeting" makes me twitch, but if you add Powerpoint I just want to crawl into a nice dark cupboard and hibernate until it goes away.

    2. AndrueC Silver badge
  9. Mage Silver badge
    Alert

    RUGs

    Strange choice

    "converting a PowerPoint presentation to PDF, where AMD's chips proved 29 per cent slower when running off battery power, or performing an Outlook mail-merge"

    You couldn't pay me to do those. Also, unlike a 3D render, is even a 50% slow down significant on any of their RUGs?

    1. Mark #255

      Re: RUGs

      Also, the Mail Merge is RUG 1220.

      Where are the results for RUGs 0001 to 1219?

      It's this sort of cockwomblery that gives honest marketers (both of them) a bad name.

    2. Roland6 Silver badge

      Re: RUGs

      "where AMD's chips proved 29 per cent slower "

      They might have been, however Intel's own graphs show that they still outperformed the Intel CPU's...

    3. doublelayer Silver badge

      Re: RUGs

      I don't do either of those very often, but converting to a PDF is usually a few-seconds activity in my experience. Extending a 2.5-second task to 4 seconds or a 11-second task to 15 probably doesn't change the user experience much, especially as selecting the options probably took several seconds too. The performance differences I care about fall into three categories:

      1. Things that take a long time to run. Adding three minutes to a process can be an important detail. However, I don't do many of those things on battery power.

      2. Tasks that get run in batches.

      3. Tasks that I do very frequently, such as loading new pages while browsing.

      When I'm mobile, I care most about the third category. As long as the slowdown isn't bad enough that I see it while reading or writing things, I doubt I'll have a problem. Most intensive work gets down while the laptop is connected to power. This is why the CPU is rarely the most important detail in a computer purchase in my opinion. Amount of memory, memory speed, storage speed, and repairibility are my metrics of greatest concern.

  10. 9Rune5

    Battery beats AC?

    I'd be more concerned with those tests where DC beats out AC, albeit with a small margin.

    In any case, it sounds like they tested the default power plan currently selected.

    1. bill 27

      Re: Battery beats AC?

      Yep, I have a mix of Intel and AMD machines. The AMD's are newer simply because...well they're newer. The CPU models are:

      Intel B940, the sole laptop, and ancient.

      AMD A12-9800

      AMD Ryzen Threadripper 2950x

      AMD 9 3900x

      AMD A8-6410

      Intel G3220, qualifies in second place for age.

      Needless to say I don't care about how well things work without wall power.

  11. cornetman Silver badge

    Just more of the same from Intel, where they seem more interested in talking about AMDs products than their own.

    Interesting/hilarious video by Steve at Gamer's Nexus:

    https://www.youtube.com/watch?v=aFHBgb9SY1Y

  12. luis river

    Great AMD

    Intel had lost the World battle advance CPU´s, Apple flag chip A14,63,3 had humilliated Intel 1184 G7,55,3 CPU flag chip, Intel was go powerful uniquely by pacts with mogul Bill Gates and Mr. "Linux is a cancer" Steve Ballmer, Intel regard will be punished by the market. Don´t is possible rival AMD on year´s 2011 until 2016 was on almost bankruptcy company and today is better Intel.

    1. IGotOut Silver badge

      Re: Great AMD

      You know you can upgrade to GPT-2?

  13. Binraider Silver badge

    The times I use a laptop without a plug nearby can be measured in fractions of a percent; typically to move from one plug to another. If I need heavy number crunch 1) I plug in, and 2) I use a proper computer. A very occassional site visit might warrant using a fondleslab; though a camera is ultimately handier and less of a pain to try and use.

    It's an irrelevant benchmark, and one suspects AMD's performance dropping on battery is an intentional choice to preserve battery.

    Far more interested in the raw performance per TDP as far I'm concerned.

    1. Boothy

      Yup, this.

      Pre current situation, only time I generally ran on batteries was in face-to-face meetings, but I'd usually have a power brick close by if needed.

      But in those meetings, all I'd be doing usually is making notes, referencing items, or worse case doing a presentation. I might be running a VM or something for the presentation, but that typically would have been started and confirmed working before I even left my desk.

    2. John Robson Silver badge

      > Far more interested in the raw performance per TDP as far I'm concerned.

      Then you're going to love the M1.

      1. zanto
        Linux

        Depends

        Will the M1 run Linux?

        1. Morten Bjoernsvik

          Re: Depends

          And will we get it on anything other than apple iron?

          1. John Robson Silver badge

            Re: Depends

            No - but if your criteria is performance/watt then you pay for the option which gets the best performance/watt.

            The Air looks like a very strong contender for anyone who doesn't need to do lots of 10 minute+ compiles or renders, and the mini just looks downright compelling.

            Given the per watt advantage they have... even a VM on there is going to beat out your raw x86 per watt performance.

        2. John Robson Silver badge

          Re: Depends

          Almost certainly yes...

          Will we have open source drivers for everything on it - probably not quickly.

          But for the performance/watt, you could run Linux in a full blown VM and still be blown away by the performance/watt.

      2. Binraider Silver badge

        As a general use machine, yes, the M1 is pretty hot. Will the company buy one? Nope. You get a [insert generic laptop of choice] here. What do I do on generic machine? Outlook, Excel, Word, Powerpoint, occassionally access and a bunch of in-browser apps. Would I pick this setup even for general use if I had my way - nope! Do I want a laptop for "home"? Nope. Got a nice desktop and usual fondleslabs for ordering off [takeaway vendor of choice]

        To actually do my job though I occassionally use COMSOL multiphysics, a use case about as far removed from any laptop that can be imagined. 2-D models work fine with "moderate" amounts of RAM - 16 to 64GB. 3D workloads I've filled a 2TB swap file without even blinking. Workstation or server setup preferred for obvious reasons. Irritatingly the bus speed is an extremely important factor in running very large models; so I can have a trade off between very fast "gaming" RAM in moderate quantities, or bonkers quantities of rather more plain RAM. But as soon as a model has to swap to disk you can basically write off using the machine for a week; if it doesn't crash. I haven't had the opportunity to try a really big NVME for swap yet, which might be something of a happy medium.

        There's not a laptop in the land that can get close those needs right now, short of calling up a server to run them for you. I'd actually recommend that model; but the software licensing is rather more pricey and with only one or two of us in the whole company using it, it doesn't stack up.

        1. John Robson Silver badge

          So your complaint is that an entry level laptop can't do the work of a server with multiple TB of memory.?

  14. Dave 126 Silver badge

    Software idea

    An application that sits in your computer and logs the workloads that you personally use in a month, with attention on heavy loads that might slow your workflow, time between charges, etc. It then (locally, of course) compiles a profile of tasks that have slowed you down and consumed battery.

    This application then cross references your personal profile against multiple benchmarks of new laptops, so as to allow you choose a laptop that is a best match for your workloads / charging habits etc.

    1. Sgt_Oddball

      Re: Software idea

      Lovely idea with just a few caviats..

      1. Not every application behaves the same for each user (some users just do basic maths in excel, others have Lovecraftian monstrosities with VBA/database work grafted on the side).

      2. A standardised test would need all variables accounted for - power profiles, screen brightness background applications etc, etc.

      And that's before we start accounting for privacy (I mean browsers don't use that much juice unless there's streaming videos involved. Would that then mean it would need to track urls? Would you be happy if it started leaking your high battery impact streaming video habits?).

      As much as it's a nice idea unless I'm getting paid for it, I've got better things to do.

  15. Anonymous Coward
    Boffin

    I am literally 157% certain that Intel's marketing department are once again spouting bollocks.

  16. GrumpenKraut
    Happy

    Poor intel

    So desperate.

    1. I am the liquor

      Re: Poor intel

      It's a good demonstration from Intel of how making a bad argument undermines your case worse than making no argument at all.

  17. Aoyagi Aichou

    Question

    Isn't underclocking and power management in general broadly controlled by the Laptop manufacturer as opposed the CPU manufacturer? I think they are more likely to benefit from marketing hours upon hours of battery life thanks to aggressive throttling.

    Regardless if I'm sat at a conference, and aeroplane, or at the garden, I want to have the option to be the one in control of these things. I want to be able to set all its components to be unrestricted by power-saving measures (assuming the battery can take it anyway).

  18. Anonymous Coward
    Anonymous Coward

    An extension in battery life is only useful if it exceeds the increase in time taken to do your work

    "If you're in that position, wouldn't you want your computer to extend battery life by throttling its CPU engines, if necessary?"

    If the resulting increase in battery life is greater than the increase in time taken to perform the tasks, then that could be a win (if you have a long flight...)

    But if the battery life is only increased by (say) 10 minutes while increasing the time taken to complete tasks by 25 minutes then it is primarily users' frustration that will be increased, as their work will take longer and they still won't be finished before the battery goes flat.

  19. Wolfclaw
    FAIL

    Intel Porkies

    Intel getting so desperate to do anything to dampen AMD rise, that they have to report to questionable and maybe misleading tactics.

    1. Saphilous

      Re: Intel Porkies

      It's almost sad.. It used to be a decent brand before all the price hikes and bs claims.

  20. Brad16800

    statistics can be biased, 90% of people know that

    1. IGotOut Silver badge

      I thought it was 103% of people knew.

  21. Anonymous Coward
    Joke

    The performance of my Intel based machine drops precipitously when I unplug it. In fact, it runs so slowly, even my monitor switches off.

  22. Anonymous Coward
    Anonymous Coward

    "Outlook Mail Merge"

    Can you think of anything that summons up the dismal nature of Windows/Intel more?

    On a side note, good to see Intel's dirty tricks division has woken up. Being #3 in desktop CPU performance out of 3 must really hurt.

  23. hammarbtyp

    It does tie in with our own experience that AMD CPU power management is far more aggressive than Intels. Whether is that through design, or because the architecture is inherently more power hungry, i don't know

    Of course in the "real" world, you may not see any difference. There are so many competing demands on a computing system, that you may not realise any slow down is due to CPU speed. Generally most processors out there are "good enough", unless you have some CPU intensive task (such as colour grading an 8K movie).

    1. druck Silver badge

      You've obviously not seen our old ASUS Vivobooks, both the 11.3" i3 and 15.4" i5 models clock the CPU down to 50% on the default power management setting. They need putting back in to performance mode to do anything other than reading static web pages.

  24. Tom 7

    But the power savings are unbeatable

    when switched off.

  25. Anonymous Coward
    Anonymous Coward

    Pretty sure my Intel laptop slows down, as it moves to balanced power mode (and not high performance) when not plugged in. The screen dims and the NVidia graphics card is switched off (and it moves to Intel HD).

    Not sure what Intel are trying to state here - apart from the obvious - that laptops slow down when not plugged in.

  26. AlanSh

    Passparm test show it's true - but so what.

    I ran the passmark CPU tests on my HP Envy laptop which has an AMD Ryzen 7 4700U in it.

    With mains on, I got 13923. WIth just battery, I got 10252. So, yes, there is a difference.

    However, in reality, I don't notice any difference apart from, when I ran the tests, the fan wasn't running as fast on battery power (indicating the CPU wasn't getting as hot).

    Alan

    1. Boothy

      Re: Passparm test show it's true - but so what.

      Interesting, the default balanced power plan I have on a Zen+ laptop, a late 2019 HP, switches to passive cooling when on battery (active on power).

      That's seems to be the only CPU related tweak.

      But should still mean short bursts of speed, are likely just as fast as when plugged in, it will just hit thermal throttling sooner on battery with any sort of sustained load.

    2. Roland6 Silver badge

      Re: Passparm test show it's true - but so what.

      >With mains on, I got 13923. WIth just battery, I got 10252.

      Did you changed the power scheme to max. performance etc. for both mains and battery?

  27. Krassi

    look squirrel

    So no comparisons to Apple's ARM chip then, which independent experts give the impression can beat both AMD & Intel offers with eyes closed & running backwards.

  28. cdegroot

    Popcorn time

    Don't you love some good competition? 2020 is interesting: AMD overtaking Intel, Apple tossing ARM into the competitive mix, and all that will happen is that we, the users, get better chips (whether that means faster, better performance/watt, cheaper, it'll be better whatever your preferences are). Competition is good, and I think it's good Intel's hegemony got broken.

    Typing this on an XPS15 because, heck, whatever you buy is "fast enough" these days and more than the CPU core counts.

  29. Zola
    Joke

    I wonder...

    Have Intel hired Trumps legal team to help produce these stats? They do have a similar way with numbers.

  30. EnviableOne

    Zepherus G14 missing

    Could this be because its runing Ryzen 9 4900HS that knocks spots off everything they have?

    1. wegie

      Re: Zepherus G14 missing

      Probably couldn't get hold of one, luckily for them! They seem to be permanently back ordered/out of stock. I've been lusting after one to replace my finally knackered 2009 Zen ultraportable for months.

  31. The Aussie Paradox
    Terminator

    Interesting.

    I noticed that, if you do not apply power to the computer, it runs almost 100%* slower than it does with power.

    As someone once said "Companies** use statistics like a drunk uses a lamppost. For support rather than illumination"

    * With a margin of error of plus or minus 100% if the CPU is in a Terminator.

    ** Original quote MAY have said politicians, but my memory is old and feeble (386 CPU & no math co-processor).

  32. UBF

    The apparent "drop" in battery performance is irrelevant

    A given lengthy operation, e.g. a large mail merge in Outlook, requires a specific amount of electric energy to be performed. If laptop A is plugged in, full power will be used and the operation will end quickly. If laptop A is on battery, the power used will be less (e.g. 30% less than when plugged in) and the operation will obviously take 30% more time. (Please note that in this scenario there will be no battery power "saved" - the amount of energy needed for the operation is given and constant.) Now, if laptop B, having a CPU comparable to A's, opts for say, only 20% less power used when on battery (regardless of CPU brand!), that laptop will appear to be 10% faster than laptop A. Therefore, the apparent "performance drop" or "performance gain" just reflects the different power management choices used by the manufacturers, and not any inherent advantage of a specific CPU brand. Intel, please correct me if I'm wrong.

    1. This post has been deleted by its author

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