Revisiting the power consumption and efficiency of Intel’s Core i7-8700K

Last week, I held off reporting estimated power efficiency numbers for AMD's Ryzen 7 family against those of Intel's Core i7-8700K in my Coffee Lake review. The power consumption numbers I observed for the Ryzen 7 CPUs during my tests were unusually (and improbably) high even though performance remained in line with expectations. I didn't track down the precise cause of the issue, but a reinstall of Windows on our AM4 test rig today fixed the problem. With everything in order again, I finally have load power numbers and estimated task energy calculations to share.

Blender is useful for estimating the amount of energy our CPUs consume over time because it's an unusually steady-state workload—power consumption varies little over the course of a render. We use the "bmw27" test file and monitor system power consumption using our trusty Watts Up power meter. Since Blender reliably consumes about the same amount of power over time, we can estimate task energy in kilojoules by relying on the fact that a joule equals one watt expended over one second.

First off, we have instantaneous load power numbers for our test systems. Among our ostensibly 95W CPUs, the Core i7-8700K sips the least system power of the bunch by a small margin. Overclocking the Ryzen 7 1700 and the Core i7-8700K results in a small advantage for the Intel chip, but the emphasis is on small. For all intents and purposes, a system with a Ryzen 7 1700 at 4 GHz draws the same amount of power as a Core i7-8700K overclocked to 4.8 GHz for AVX workloads (like Blender) does.

Instantaneous system power draw doesn't tell us anything about efficiency, though—low power consumption isn't worth much if it takes a CPU eons to render the bmw27 test image. The Core i7-8700K delivers by slotting between the Ryzen 7 1700 and Ryzen 7 1700X, but the Ryzen 7 1800X is the winner among stock-clocked chips.

To visualize the efficiency of each part on the bench, we can plot each chip's estimated task energy against the time it takes for a given CPU to complete our bmw27 test render. The most efficient chips will tend toward the bottom left of the chart, where time-to-completion and task energy are lowest.

As you can tell by the proximity of the dots on our scatter, the Core i7-8700K consumes an estimated 2% less energy than the Ryzen 7 1800X, and about 4% less energy than the Ryzen 7 1700X, but it's anywhere from 2% slower against the 1700X to 7% slower against the 1800X. That's a wash from where I'm sitting. If you're going for the least power consumed for rendering at the expense of a slightly longer task-to-completion, the Ryzen 7 1700 emerges victorious. It's also the only 65W CPU on our test bench, so keep that in mind.

Overclocking the Ryzen 7 1700 and i7-8700K does let them finish the bmw27 render faster, but the extra power consumed this way outstrips the time saved by hastening the pace of number-crunching on both chips. The Ryzen 7 1700 at 4 GHz finishes about 4.4% faster than the i7-8700K at 5 GHz, and it consumes 2% less power. Given that we're estimating these numbers and not measuring the exact amount of power consumed, I'd call that another wash. AMD might have a slight performance-per-watt edge, but we'd need more precise data to say for sure.

Overall, these new results don't change any of the conclusions in my initial review. Both AMD and Intel offer outstanding power efficiency in this generation of CPUs, and our tests back that up. Intel won this battle on other fronts. Only one thing is for sure: the ongoing war for CPU supremacy continues.

Comments closed
    • HERETIC
    • 2 years ago

    Little off-topic,but relevant.
    Was surprised when reading this-
    [url<]https://www.techspot.com/review/1499-intel-core-i3-8100-i3-8350K/page4.html[/url<] Shows DDR4-3200 using 10 Watts more than DDR4-2400 with a i3-8100. Another surprise in the article was-Up to 10% gains using the faster Ram. If this can be accurately reproduced I don't know.................................

    • George_PPS
    • 2 years ago

    Although Core i7 series is still superior in most gaming applications, Ryzen 7 really shows its true power in the last chart. Well done AMD. Ryzen is better is almost everything except gaming. So for those who don’t game or don’t game much, Ryzen is the superior option.

      • Jeff Kampman
      • 2 years ago

      Actually, no, it’s very slightly superior in one test.

      Anybody reading comments like this one really ought to go read my full review for the full picture: [url<]https://techreport.com/review/32642/intel-core-i7-8700k-cpu-reviewed[/url<]

      • IGTrading
      • 2 years ago

      I would say it Ryzen is mostly a match in productivity apps, wins some, loses others (single threaded or heavily Intel-optimized or just apps that use better the Intel architecture such as audio processing) and shows better thermals and power consumption.

      But Ryzen loses in FullHD gaming while matching Intel in 4K.

      On the other hand, Jeff why do you say Ryzen is only “very slightly superior in one test” ? Do you mean that by considering energy used per task completed ?

      • srg86
      • 2 years ago

      I don’t play games and use produtivity apps. I also do have uses for lots of cores (compiling).

      But The Core i7 is better for me because of the iGPU (it does everything I need) and better quality chipsets.

        • Anonymous Coward
        • 2 years ago

        Is chipset quality still an issue? Perhaps I don’t read enough, but I hadn’t heard of anything wrong with AMD’s latest effort.

          • synthtel2
          • 2 years ago

          The literal chipset (some of the southbridge stuff) seems fine, but the DF/IMC had/have some quirks. Memory compatibility and the performance marginality problem seem mostly sorted out, but the idle crash definitely still exists (it only takes a slight RAM clock or VSoC boost to fix it on my system, though).

    • watzupken
    • 2 years ago

    Results are expected to some extent. I feel AMD’s approach to go wide with more cores will likely be more efficient as oppose to Intel pushing hard on clockspeed. In fact, I believe the i7 8700K does not have much overclocking headroom unless you decide to go with extreme cooling since it will require a significant Vcore to up the clockspeed further, especially if one is gunning for 5Ghz and above.

    • Zizy
    • 2 years ago

    How would i7 8700 (non-k) perform here? It should be more efficient due to lower clocks. The k part seems clocked as high as possible to get those gaming wins, even if this comes at the cost of efficiency.

    Where do R5 1600 and 1600X sit? 1700 is obviously better in both kJ and time, but how much worse are the 6C ones?

    • TheEmrys
    • 2 years ago

    Feels like the old Athlon 64 4000+ X2 days. I have very fond memories of that CPU with my Crucial 9800pro.

      • juzz86
      • 2 years ago

      A Crucial graphics SKU. I had an X1900XTX from them, too.

      Walk down memory lane there mate!

    • ronch
    • 2 years ago

    Such a far cry from Bulldozer. And remember, Intel tunes their process node more specifically for their CPUs, while AMD has to make do with an off-the-shelf 14nm process that’s probably not as 14nm as Intel’s 14nm. Now AMD just really needs to tweak Zen like there’s no tomorrow.

      • anubis44
      • 2 years ago

      12nm Ryzen incoming in February 2018…

    • exilon
    • 2 years ago

    High power usage might have come from some high performance power plan that lingered around.

    It must’ve been some sleep state that was disabled on the test platform, since the performance remained completely unchanged. Maybe the GPU?

      • MOSFET
      • 2 years ago

      Nvidia GPU with a greater than 120Hz monitor? My 1060 idles fine with the monitor at 120 but won’t idle at 144Hz.

        • Chrispy_
        • 2 years ago

        One of these days Nvidia will sort their drivers out.

        I *know* Nvidia is the better choice for gamers but I actually justified Vega for my PC because the number of trivial but irritating driver quirks with my Geforce drove me over the edge. Some of the bugs have been around since 2006, ffs.

    • chuckula
    • 2 years ago

    For all the people wondering at the efficiency of the 1700, I suggest underclocking a Coffee Lake part until it gets similar performance levels and then being amazed at the non-linear relationship between CPU clocks and efficiency.

      • Voldenuit
      • 2 years ago

      You don’t even have to underclock. Undervolt it until you get to the lowest stable voltage, and it might be surprisingly better than stock.

      I had an Opteron 165 that was undervolted *and* overclocked. The damn thing was so energy efficient I once forgot to plug the fan header on my CPU HSF, and the computer ran fanless for a whole day before I noticed.

      • abiprithtr
      • 2 years ago

      Why would anyone purchase the pricier 8700k and undervolt/underclock it to the cheaper 1700’s performance levels?

        • chuckula
        • 2 years ago

        Nobody would.

        But thanks for pointing out how buying a cheaper and lower-clocked Coffee Lake gives you the same benefits.

          • iBend
          • 2 years ago

          He said “cheaper 1700”

    • Mikael33
    • 2 years ago

    What an Age to live in, where AMD CPUs are more power effcient that Intel cpus, in that workload anyway.A far cry from their piss poor efficiency with the FX series, especially the silly FX 9590(and brethren), the whole let’s sell an overclocked chip *way* past it’s optimal voltage curve just for (frequency) bragging rights and sell it for a stupid amount.
    Now if RTG could only get the memo about power efficiency that’s not totally embarrassing being a good thing, not only for power draw but to curb a long standing complaint about AMD gpus- the excessive fan noise they make compared to their nVidia competitors, less power draw meaning less heat ofc.

      • IGTrading
      • 2 years ago

      Actually the FX9000 series made sense for people just wanting an upgrade without changing the whole platform and it was cheaper than investing in a new motherboard, memory, processor and cooler.

      The processors also showed good performance in applications that were compiled with a “cleaned” version of Intel’s C Compiler, without the “most compatible execution path / <Intel Genuine> check” . See : [url<]http://www.agner.org/optimize/blog/[/url<] For AMD, not having anything else, it gave them something to talk about in marketing slides and keep the brand alive in the eyes of the public. For AMD buyers, it gave them an upgrade path and bragging right about CPU frequency πŸ™‚ It was impressive that it was possible and respectable on AMD behalf for putting the effort into it. It was way better than Intel telling people that just bought new motherboards 8 months ago : "if you want an upgrade, buy another motherboard suckers" πŸ™‚

        • chuckula
        • 2 years ago

        [quote<]Actually the FX9000 series made sense [/quote<] Here's the airplane glue. Tell me another one!

          • IGTrading
          • 2 years ago

          Sure πŸ™‚

          Every time there’s any criticism of an Intel chip or technology from us and we get a reply from you, the reply comes at the same time with -7 votes πŸ™‚

          Not -5, not -12, but exactly -7 when you reply.

          We are honestly touched you care so much.

          It is our honest belief that offering your clients an extra upgrade option and also cover it with by the book business warranty (overclocking is not covered by warranty) is a positive move for your clients and the market.

          It was just another product in AMD’s stack.

            • chuckula
            • 2 years ago

            If you want to be an AMD fanboy that’s cute (if massively unoriginal).

            It’s when you post fact-free insults of Coffee Lake that contradict TR’s hard numbers (and I sure as hell didn’t see you doing your own scientific benchmarks of these parts) and then [b<]IN THE SAME THREAD[/b<] pretend that the FX9000 series parts were really pretty good because they were an alleged "upgrade" from... what I don't know... that you start to lose it. It's not that you're a clearly biased fanboy who will automatically get upthumbs for supporting AMD. It's that your wall-o-text drivel is so out of it that nobody even thinks you believe what you are copy n pasting.

            • derFunkenstein
            • 2 years ago

            well now you’re at -8, so you’re welcome.

            • IGTrading
            • 2 years ago

            Much appreciated πŸ™‚

        • just brew it!
        • 2 years ago

        [quote<]Actually the FX9000 series made sense for people just wanting an upgrade without changing the whole platform and it was cheaper than investing in a new motherboard, memory, processor and cooler.[/quote<] Only if you had free electricity and already owned a motherboard that supported 220W CPUs. (Most of the earler AM3+ motherboards only went to 140W or so.) [quote<]The processors also showed good performance in applications that were compiled with a "cleaned" version of Intel's C Compiler, without the "most compatible execution path / <Intel Genuine> check" . See : [url<]http://www.agner.org/optimize/blog/[/url<][/quote<] This is completely irrelevant to 99.99% of users. How many people do you figure had the ability and inclination to recompile their performance-critical applications with a hacked compiler? [quote<]For AMD, not having anything else, it gave them something to talk about in marketing slides and keep the brand alive in the eyes of the public.[/quote<] It smelled of desperation. [quote<]For AMD buyers, it gave them an upgrade path and bragging right about CPU frequency :)[/quote<] It was a PR stunt. [quote<]It was impressive that it was possible and respectable on AMD behalf for putting the effort into it.[/quote<] It was nothing more than a factory overclocked part with ludicrous power consumption, an equally ridiculous price tag at launch, and a 17% bump in stock clock over the FX-8350. I suppose it was "impressive" that AMD's marketing department managed convince top management that it was a good idea... [quote<]It was way better than Intel telling people that just bought new motherboards 8 months ago : "if you want an upgrade, buy another motherboard suckers" :)[/quote<] Yeah, providing an upgrade path is nice. But as I noted above, even if you already owned an AM3+ system you would've probably needed to replace the motherboard anyway. And if you already owned an AM3+ motherboard capable of pushing 220W to the CPU, chances are you were overclocking; you would've been better off just getting an FX-8350 and OCing it instead of paying AMD hundreds of dollars to do it for you. I'm an AMD fan. I've been using AMD CPUs since the K6 days. The only Intel build for personal use that I've done in ~20 years was because I won a pile of parts in a giveaway, so I am not anti-AMD by any stretch of the imagination. I'm sorry, the FX-9000 series was just pathetic.

          • IGTrading
          • 2 years ago

          We agree with most of what you’ve said. πŸ™‚

          The upgrade WAS for the people already owning the compatible motherboards.

          Some countries have very cheap energy. Some other places welcomed even the extra heat produced.

          But it doesn’t matter just as energy consumption doesn’t matter for people overclocking the 8700K or the X299 platforms.

          It wasn’t pathetic as it wouldn’t be pathetic for Intel to offer an 8800XE 5.5 GHz CPU for gamers and enthusiasts. It would actually be impressive if they would do it and cover it by warranty. But they will not, because for the moment they can’t. (nobody was able to overclock the 8700K to 5,5 GHz on air or water cooling yet)

          It was just a product. It worked. It was covered by warranty and it came with clear specs on the box, unlike the 7980XE that claims a 165W TDP and sucks up to 190W in productivity applications.

          It did not win AMD any crowns or halo PR , nor did they claim any of that.

          I wasn’t talking about people recompiling the programs. I was referring to programs that were compiled by the developers with a “cleaned/patched” Intel C compiler. (like CineBench for example, which is compiled with the patched Intel C compiler, not the default version) .

          If there is anything else you think we don’t agree on, let’s just agree to disagree. πŸ™‚

            • Redocbew
            • 2 years ago

            Who is “we”?

            • chuckula
            • 2 years ago

            [url=https://www.youtube.com/watch?v=E09LU6XVyxs<]The Royal We[/url<]

            • Redocbew
            • 2 years ago

            The Dude abides.

            • IGTrading
            • 2 years ago

            Me and my colleagues, although I’m the technical consultant in the team.

        • abiprithtr
        • 2 years ago

        The only thing I can agree with in your post is the following:

        “For AMD, not having anything else it gave them something to talk about in marketing slides and keep the brand alive in the eyes of the public.”

      • watzupken
      • 2 years ago

      Actually, Intel always appear more power efficient is because they have the advantage of a more advanced fab. With this advantage slowly eroding away, it is expected that they are losing the power efficient race. Unless they radically change the design of their CPU, they are going the way of the P4 days with the clockspeed race. Adding 2 additional cores may be a good step by going wider instead of competing with clockspeed.

        • NeelyCam
        • 2 years ago

        When it comes to practical efficiency, I thought the integrated voltage regulation (FIVR) really kicked it up a notch in laptops with Haswell/Broadwell. If I understood correctly (experts please correct when I’m wrong), the integrated solution was more responsive, so the CPU can switch on/off more often, saving battery? Anecdotally, my Haswell/Broadwell chromebooks have batteries that never die… (C720/C740).

        Intel removed it in Skylake generation, but I saw some articles that said FIVR would be used again for Icelake… any confirmation on that?

        EDIT: Here:

        [url<]http://wccftech.com/intel-ice-lake-return-fvir-wip/[/url<]

    • bhtooefr
    • 2 years ago

    You know, seeing a Ryzen [i<]outright win[/i<] on energy efficiency, and by a decent amount, makes me wonder what this might mean for Raven Ridge. Raven Ridge is planned to go down to [i<]4[/i<] watt TDP, there's been documentation of Zen being around 0.6 W per CPU core at 1 GHz, and all of this with Summit Ridge near the end of the frequency/voltage curve - so mobile parts would be in a much better part of the curve?

      • brucethemoose
      • 2 years ago

      With a die shrink, that’s pushing into tablet/phone territory.

      I would definitely buy a phone that could dual boot Android and full x86-64 Windows.

        • bhtooefr
        • 2 years ago

        Well, it doesn’t say anything about how good AMD’s clock gating is, how aggressive they are at ramping clocks up and down, or things like that that determine idle power.

        But, it does still look rather impressive, and makes me hopeful for tablet-class (4 watts is into tablet class, at least in x86 tablets) Raven Ridge.

          • Anonymous Coward
          • 2 years ago

          I’d be impressed enough by a competitive showing in the 15W class. If Zen can keep winning as the watts drop into single digits, that would make my head spin. If that happens, then we [i<]know[/i<] Intel took their eyes off the ball. They would make it clear they were daydreaming and profit-taking, [i<]not being paranoid[/i<]. Only the paranoid survive, you know. Raven Ridge, very interesting stuff....

          • anubis44
          • 2 years ago

          “it doesn’t say anything about how good AMD’s clock gating is, how aggressive they are at ramping clocks up and down, or things like that that determine idle power.”

          Jim Keller, who also worked on the Apple A4, A5 and A6 SoCs, is a master of power efficiency. His designs have some of the very highest performance per watt of any CPU design. I’m pretty sure Ryzen is right up there with the best of them in this regard, although I don’t have numbers on me.

            • bhtooefr
            • 2 years ago

            Excellent point, re: Jim Keller knowing this stuff.

      • ronch
      • 2 years ago

      The Zen core has the efficiency figured out. Vega, I’m not so sure, given how it doesn’t seem to be as efficient as can be. We’ll see.

        • bhtooefr
        • 2 years ago

        Although, I don’t think we’ve seen Vega when it [i<]hasn't[/i<] been pushed to the top right corner of the frequency/voltage graph. IIRC Polaris looked ugly at launch, but works well in mobile - could be the same sort of situation for Vega?

        • Zizy
        • 2 years ago

        Vega and Polaris are disastrous against Pascal chips. But they still have higher efficiency than Intel’s IGP, so it should still result in a nice RR performance.

      • Anonymous Coward
      • 2 years ago

      So whatever advantage Intel’s magical fabrication provides, its largely eaten up by sub-optimal choices elsewhere. This works out great for Intel on the desktop, less well in servers and mobile. I am surprised how well AMD has done here, I think Intel is also.

      • Klimax
      • 2 years ago

      Absence of full 256-bit AVX units and much smaller bandwidth in lower cache hierarchy. (Especially L1/core)

        • Anonymous Coward
        • 2 years ago

        I once was following my parents along when they were car shopping, and they were talking to the salesman about a Jeep Grand Cherokee. There were V8 and straight-6 engine options, the salesman said the fuel economy was the same. That was of course bullcrap.

        This might be the first time we’ve seen a similar effect in processors. Previously, we had huge disparities in architecture and manufacturing that obscured differences on this scale, or made them irrelevant in the face a huge performance gap.

          • Anonymous Coward
          • 2 years ago

          Downvotes for my theory of excessive hardware resources leading to poor efficiency. Sad!

            • Jeff Kampman
            • 2 years ago

            It’s pretty complimentary to Intel that its engineers can implement wider data paths and AVX units and still come out about even for power consumption and performance, all from fewer cores. It’s not like you’re paying (much) for the extra logic and wires compared to eight-core Ryzen CPUs with narrower data paths and vector units, at least from a system power consumption perspective.

            • Anonymous Coward
            • 2 years ago

            Intel has no doubt crafted the ultimate desktop processor core here, also the ultimate for HPC.

            But AMD has found an opening elsewhere, it seems. Yay!

          • Waco
          • 2 years ago

          You can get the same mileage in heavy vehicles from smaller versus larger motors. The bigger motors are more efficient and more powerful at lower RPM where the smaller engine struggles to move the weight around and needs to be wrung out a bit more.

            • synthtel2
            • 2 years ago

            Most engines reach peak efficiency at much higher output than we think. Turbos and/or high compression do make this less true, as then high load means richer mixtures and/or much more roll-off of ignition timings, but recent tech advances have been able to compensate for a lot of that.

            [url<]http://ecomodder.com/wiki/index.php/Brake_Specific_Fuel_Consumption_(BSFC)_Maps[/url<]

            • Waco
            • 2 years ago

            Full load (assuming non-turbo) is almost always most efficient. That said, a large engine at low RPM will generally use less fuel than a small engine at high RPM even when producing the same amount of torque.

            I should calculate the BSFC for my car (I built the ECU, so that should be pretty easy to modify/log). It is turbocharged, but I bet the map below atmospheric would be interesting.

            • synthtel2
            • 2 years ago

            Why not map above atmospheric? Does it meter differently in that region? That would be pretty interesting regardless.

            Taking that literally, large and small engines making the same amount of torque means the large engine is at much lower load/BMEP and is unlikely to be as efficient (but it will use less fuel in absolute terms, because the case you specced has the big engine making less power in proportion to the reduced RPM). At equal power instead of equal torque, it is a bit closer than BSFC map shapes alone would suggest since large engines have got natural efficiency advantages (like less heat loss to cylinder walls due to better cylinder surface area to volume ratio). Being away from peak efficiency for a given engine generally hurts a lot quicker than the gains from bigger engines help, though.

            • Waco
            • 2 years ago

            Mine is tuned for power and safety, so I made no attempt to keep fuel consumption in check into boost. It might be interesting to see just how inefficient it is, though.

            • synthtel2
            • 2 years ago

            Gotcha. It would still definitely be interesting.

            • Anonymous Coward
            • 2 years ago

            What you really want is a low-internal-friction motor, independent of size. This is why a lot of modern small motors disappoint if you compare fuel used per liter, vs old “crude” motors. Generally high horsepower means low efficiency because it seems that internal resistance is related to potential output.

            • synthtel2
            • 2 years ago

            Modern engines have impressively low internal friction compared to a couple decades ago, but it isn’t any kind of dominating factor. At low RPM / high load there’s the heat loss to cylinder walls, at high (especially above atmospheric) load there are fuelling and timing changes to avoid damage, at low load there are pumping losses (moving air through a closed throttle plate isn’t free), valve timings are going to affect thermal efficiency at different speeds (EVO probably being the simplest), and so on.

            Modern cars don’t get such good efficiency as they might, but emissions and safety have a lot to do with that. Many potential efficiency improvements involve higher cylinder pressures and temperatures, but that makes a lot of NOx and we’ve got increasingly tight controls on that.

            As far as how much internal friction an engine has, the biggest tradeoff is that making an engine low friction tends to make it less durable, especially against low-RPM high-load operation. Smaller hydrodynamic bearings slow things down less, but they need more speed to support a given load. The other one very likely to apply here is that aggressive cams (and maybe complex VVT systems) have higher losses.

            • Anonymous Coward
            • 2 years ago

            Hmm, I recall seeing idle throttle was ~30% of full throttle when I once had access to one of those boxes that interfaced with the engine control computer. I wonder if that was a car from the early 90’s. Nothing was wrong with its efficiency, but certainly running at 50% throttle was wasting a lot of gas compared to running at 80%, if it was possible to do the same work in each case.

            I also have in mind the rather unimpressive fuel economy numbers returned by fancy new engines like Ford’s ecoboost, also the surprisingly small advantage that small turbocarged Japanese cars had over comparatively crude V8 Mustangs. Or the drop in fuel economy associated with switching to a turbocharged (or otherwise higher output) version of mostly the same motor. Most of this is from 10+ years ago so maybe its no longer accurate. Maybe its coming from aggressive cams and VVT systems like you say.

            Anyway, I think my point about excessive potential output leading to reduced efficiency is valid despite the details of why.

            • bhtooefr
            • 2 years ago

            A lot of the problem with turbocharged engine efficiency is that, when it’s used for efficiency, a lot of automakers get greedy with the downsizing benefits – they’ll optimize the engine displacement for the test cycle that they’re on.

            Now, this results in a vastly undersized engine for normal acceleration demands, so that’s where turbocharging comes in. The added boost gives you the power back… but it also greatly increases cylinder temperatures and pressures, meaning that the engine needs to run richer (and therefore less efficiently) to avoid detonation.

            Done properly, a mild degree of downsizing and turbocharging can improve overall efficiency – it can allow what would otherwise be the base model engine, with acceptable naturally aspirated performance, to perform like a higher-end engine, for instance – so you get base model efficiency when you drive normally, but higher performance when you want it. Done improperly, it can produce [i<]amazing[/i<] results on the test cycles (especially the NEDC, the older of the two European test cycles), but in the real world, you're always deep into boost, and therefore enrichment. And, you can also get some similar effect from gearing up a big engine, especially if you have VVT to make it drivable at low RPM - ultimately, it's about reducing the amount of air that naturally goes through the engine, and if a 6 liter engine is doing 1000 RPM down the freeway, that's as efficient - at least as far as pumping losses are concerned - as a 3 liter engine doing 2000 RPM down the freeway. Also, hybridization can help here, too - hybrid systems do two things. The first, kinetic energy recovery, is obvious... but there's a second one as well, engine optimization. During times of high power demand, you can keep the engine in a more efficient lower RPM state, using stored energy. And, during times of light, inefficient load, you can either shut the engine down and use stored energy, or you can increase load on the engine and store energy.

            • Anonymous Coward
            • 2 years ago

            So your position is that a lot of the modern turbo motors are designed primarily to game the standardized fuel tests?

            So your example of 6L vs 3L at twice the RPM. Pumping losses match. Lets assume they have the same number of cylinders. What about the rest of the internal resistance (or shall we just call it losses)? Travelling at twice the speed should involve a lot more than twice the energy, and presumably the 6L motor has a lower piston velocity. Also less piston wall surface area being swept per time. (The components would be heavier.) So would the 6L motor win this?

            I’m getting to be a big fan of hybrid powertrains, especially if they route the electrons to the back wheels and the hydrocarbons to the front. (Or reversed for a load-carrying vehicle.)

            • synthtel2
            • 2 years ago

            If they had the same number of cylinders (I6?) and were designed to the same peak output, the 6L would probably win (if nothing else there’s more space to make tolerances good as percentages), but as far as real-world examples the 6L is probably a much higher output engine working much further below it’s ideal operating point.

            A lot of truly large engines (think big rigs, construction, or even marine) are extremely efficient, have few massive cylinders (I6 can be perfectly balanced and scales pretty much indefinitely), and have low power densities that might even look like they should be misprints.

            • Anonymous Coward
            • 2 years ago

            Oh, the sad story of I6 in personal vehicles…

            • bhtooefr
            • 2 years ago

            Mercedes-Benz is bringing it back, though, and BMW never abandoned it…

            • bhtooefr
            • 2 years ago

            Yes, that is my position on a lot of modern turbo motors. Even then, they can help on real-world highway cruising, but the tiny ones in many cars nowadays generally aren’t a good idea, and I’m not a fan of them in pickup trucks that are actually used as pickup trucks (because you’re generally asking for all the power you can get, and there, the turbo engines don’t help you on efficiency).

            Assuming all else is equal, I think you’re right on the 6 liter having lower friction than the 3 liter engine in this scenario. Potential downsides, though, are related to the fact that you may well be below the 6 liter engine’s efficient running RPM (to the point that combustion stability may be affected by loading it at basically idle RPM). So, you might need to downshift, and then there goes all of your benefit. (Assuming all else is equal as far as RPM optimization, the 3 liter engine will be able to remain efficient at lower power settings than the 6 liter engine. It’s just that at the same power setting, there’s plenty of scenarios where the 6 liter engine can win.)

            This is also why some big V8 cars can be about as efficient as their V6 or 4-cylinder counterparts at the same technology level on the highway – high enough power demand that the V8 isn’t running inefficiently. City is a mix of low-power operation where the V8 will always lose, and high-power operation where the V8 will… always lose in efficiency, although not performance.

            • synthtel2
            • 2 years ago

            Throttle in that case was probably actually MAP (manifold absolute pressure), which is often measured in kPa so at sea level full throttle is 100. 30 kPa is a typical enough idle, and min load at higher RPM (as in engine braking) tends to be 15-20 on cars with mechanical throttles IME and a bit more on ones with electronic throttles. Both fuel use and power output are very non-linear against it – 50 kPa might use a third as much fuel/air as 100 kPa, and the power difference is almost certainly even bigger.

            One of the big tricks used to improve fuel economy on some modern engines is doing some power adjustment with VVT instead of the main throttle. If intake valve closing (IVC) can be delayed at-will, for instance, that lets the fuel/air charge size be reduced (some gets pushed back out to the intake manifold) without the high pumping losses of the normal throttle (and with some miller-cycle-ish benefits too). Lean burn also helps much more than it otherwise would due to this effect.

            Aggressive upshifting for better fuel economy is mostly about avoiding these pumping losses, as the peak efficiency RPM is rarely below 2000. If it sets the engine at a higher load operating point for the same power output, though, it’s likely a win.

            Ecoboost low efficiency? Maybe at peak – I don’t have any g/kWh figures handy or anything. The 3.7L in a ’13 F-150 is amazing at maintaining efficiency at low load, though, which is pretty important for practical end-user MPG numbers. Most cars have peak efficiency somewhere in the 40-55 MPH ballpark, dropping off quickly on the low side as soon as top gear can’t be maintained anymore. Said truck can get 30 MPG cruising at 30 MPH, which is frankly ridiculous. What it doesn’t like is hot weather, which makes the high-load efficiency drop-off very steep (as is typical for a turbo vehicle).

            Pre-GDI turbo engines definitely had lower peak efficiency than an equivalent N/A engine, because the compression ratio would have to be lowered substantially (as in 10:1 -> 8.5:1). GDI seems to have allowed 10:1 turbo engines to be mainstream, though I’m still not sure precisely why it helps so much with some of those limitations.

            In general, yes, having any kind of higher-output motor for a given car is going to hurt fuel economy. Aggressive (high-RPM) tuning, more displacement, and more boost all have that effect in the end, assuming the engine wasn’t terribly undersized for the car to start with.

            • Anonymous Coward
            • 2 years ago

            What’s terribly undersized these days? They have 1L motors in Europe, I have a car with a 1.6L diesel (90HP, 5 speed manual, lovely easy motor IMHO). 2L gets you into a pretty good sized van, or 2.5L will definitely go well into things that resemble plus-sized Astro vans. I suppose 3.x can come with dual wheels. I note that VW chooses to stay with 2L from 110 horses to 180 in their vans, why not upsize? I imagine they have a point with this downsizing.

            I get a mild shock every time I read about the latest monster pickup diesel in the states. 400HP? I remember when a 5.9 Cummins did 160, and that was plenty. Surely they must be suffering for all that technological overkill?!?

            Oppose of undersizing, there…

            Interesting take on the ecoboost motor. Maybe they need to be preferentially selling the 5.0 V8 in the south, and ecoboost stuff up north.

            You say peak efficiency RPM is rarely below 2000RPM. Does that statement presume a reasonable amount of load? I imagine spinning faster than needed is always a loss, separate from pumping losses.

            • synthtel2
            • 2 years ago

            Not terribly undersized as in displacement, but as in output. If an engine has to spend a lot of time near 100% power output, that’s likely an issue, but nobody builds cars like that because nobody wants to drive a car like that.

            Diesels (mostly) don’t have the low-load pumping losses that so trouble gas motors, since they (mostly) just reduce fuel injected rather than trying to cut airflow. It’s all lean burn, all the time. This means they can do overkill 400+ HP stuff with less fuel economy downside, but lean burn plus all that compression makes it pretty tough to keep the NOx emissions in check, hence DEF and all that.

            I wouldn’t be surprised if the Ecoboost vs 5.0 V8 regional thing is already happening – it makes perfect sense to me. It must be a bit of a headache for the marketing department though, having a distinction like that that matters but isn’t something they’d want to advertise.

            On peak efficiency RPM etc, check that [url=http://ecomodder.com/wiki/index.php/Brake_Specific_Fuel_Consumption_(BSFC)_Maps<]BSFC map[/url<] link. [url=http://ecomodder.com/wiki/index.php/File:Mercedes_Benz_OM617_5cyl_turbo_diesel.jpg<]Here's[/url<] one that's a good example, because of the other data overlaid on it. Any particular speed you want to cruise is going to require a particular amount of power, hence the lines of constant power overlay - say you want 20 kW, you can look at that line and see that a gear letting it run about 1500 RPM will be ideal. The dashed line shows that the actual high gear has it turning more like 2500 RPM at a speed that takes 20 kW, burning 295-300 g/kWh where 1500 would take more like 255, despite the engine's actual peak efficiency being at ~2200. So yes, peak efficiency RPM is only strictly the peak for a particular power demand (in this case ~45 kW, which probably corresponds to a cruising speed that's usually illegal). Some engines have fairly steep high-RPM efficiency drop-off, and for others it's all about load. For instance, the BSFC map for my car's engine looks like [url=http://ecomodder.com/wiki/index.php/File:Subaru-ej22-bsfc.png<]this[/url<] (beware that isn't in g/kWh, so look at slopes, not absolute numbers). The net effect is about the same, though, so long as you're not spending a lot of time near max power output.

            • bhtooefr
            • 2 years ago

            Also, one thing with high performance gasoline engines is, when you’re running at near stoichiometric mixture, you can actually cool down the cylinder by adding more fuel. So, a lot of high performance gasoline engines will do this to reduce engine temperatures, and avoid detonation or other overheating conditions. That’s why a high boost gasoline engine is so inefficient when it’s at high load – it’s injecting more fuel to cool itself.

            I wouldn’t say that high-power diesels get their efficiency from their low pumping losses directly, though. As you say (in a different way), diesels don’t even get [i<]close[/i<] to stoichiometric average mixture - the fuel just isn't volatile enough, so you have little droplets of fuel in the cylinder. Get even close to stoichiometric, and you get lots of incomplete combustion and resulting soot. (This is also why soot is hard to manage on a diesel, and they have particulate filters - you ALWAYS get the soot, it's just a matter of how much you get.) When you're running that lean, you cool things down by reducing fueling. And, detonation isn't a concern for a diesel - fuel is withheld until it's time for combustion, so premature detonation can't happen. So, the only concerns with running hot are high NOx emissions (because nitrogen and oxygen react at high temperatures) and potential engine damage. Hence diesels maintaining efficiency even in the face of extreme boost. As far as my own engine, here's a BSFC map for an older version of it (as well as its predecessor - the green area and red line are for the Gen 2 Prius engine, the rest is for the Gen 3 Prius engine, I have a Gen 4 Prius which has a revised version of the Gen 3's engine for more thermal efficiency): [url<]https://thewolfweb.com/photos/00525525.JPG[/url<] Note the line, though - the Prius does everything in its power to stay on that line of high efficiency operation (until you need lots of power, then it closes the EGR above 3600 RPM, hence the spike in torque - but it still stays on the line). The hybrid system's continuously variable behavior ensures that the engine can stay on the line (or shut down) up to a rated 68 MPH, I've seen it shut down as high as 73 MPH indicated. Within those conditions, if you need more power, it'll use the battery to try to keep you from climbing that steep climb at 3600 (or if you do climb it, to spend less time there). If you need less power, it'll shut down the engine (if possible), or it'll charge the battery with the power you don't need (if possible).

            • synthtel2
            • 2 years ago

            High boost gas engines’ inefficiency at high load is both that and ignition timings – they’re hardly ever going to run richer than ~10.5:1, which puts a cap on the efficiency loss from extra fuelling alone, but timings can bring efficiency waaaay down if conditions are wrong.

            There still is an alternate thing to cool down diesels, and that’s more boost (to get it running leaner at the same fuel rate). It’s even better than that, because a few revolutions at high fuel rate while the turbo is still spooling are unlikely to damage anything, so the practical limits to off-boost power are just soot and NOx. (Of course some override those and roll coal for days – argh.) Pressure is also a whole lot easier to handle than thermals, and some engines can handle ludicrous amounts of boost.

            Timings are a thing for diesels, though I don’t know how much of a practical concern they are on current engines. More advance does still definitely result in both higher peak pressure and better efficiency (to a point).

            CVTs are great for all of this, yeah. They mean whoever’s designing the engine gets to optimize efficiency on a line and worry a lot less about the whole rest of the map.

            • bhtooefr
            • 2 years ago

            Timing affects NOx, and retarding timing is one NOx control strategy.

            It also affects cylinder pressure, and I’ve heard of head gaskets getting blown by too advanced timing on one vendor’s TDI tunes.

            • Anonymous Coward
            • 2 years ago

            Now, that is what I call a discussion.

        • Goty
        • 2 years ago

        Was there a point there, or are you just grasping at deficiencies of the Zen architecture that only matter in niche applications (AVX) or don’t matter at all in the face of the general performance of the architecture (cache bandwidth).

          • derFunkenstein
          • 2 years ago

          As soon as AVX (edit: AVX 512, as Voldenuit points out) makes its way into Intel’s full lineup it’ll stop being niche. Of course, that may happen on the second of Never.

            • Voldenuit
            • 2 years ago

            Do you mean AVX512? Because all of intel’s mainstream lineup has had AVX2 since Haswell afaik.

            • derFunkenstein
            • 2 years ago

            Yes, sorry. That was my intent. AVX2 is in i3 and up since 4th gen.

            Weird that Pentiums and Celerons don’t have AVX2, but they’re cheap and Intel needs something to make them less desirable.

          • Jeff Kampman
          • 2 years ago

          My testing over the past few months suggests that we’re well past the point where AVX can be considered niche for productivity tasks where high performance is paramount.

            • Goty
            • 2 years ago

            Maybe I should revise my statement to say that BOTH points matter very little in the face of the overall performance of the products then. If we consider that the 8700K has twice the theoretical AVX throughput of the Ryzen processors, either the 8770K’s performance is really disappointing in the only two tests you did that utilize the AVX units in either processor or the AVX units aren’t being utilized particularly heavily in those tests (making any claims as to the usefulness of AVX in your testing somewhat spurious.)

            • IGTrading
            • 2 years ago

            I was looking at Linus’ test results in comparison with your own and I’ve noticed 2 bizarre aspects that I find interesting and weird at the same time.

            Looking at his 8700K review : [url<]https://www.youtube.com/watch?v=EJOnwF8mgXc[/url<] you can see that often the 8700K gaming AVG rates are slower than on the 7700K. I've seen that a few of your own results show the same problem, the 8700K slower than the 7700K despite the clockspeed and core advantage ... GTA, JetStream, Kraken, Optane ... What could be the issue ? The applications are clearly Intel-optimized and both processors are Intel, so what do you think might be happening. The second thing is that on his systems and using his "synthetic load" the power consumption of the 8700K was very VERY close to the power consumption of the Threadripper 1950X. (I have to mention that your tests show a more significant difference of power consumption although quite close 153W (6core) to 265W(16core ) ) Did you perhaps test power consumption in other scenarios / applications ?

    • Srsly_Bro
    • 2 years ago

    I get it. My 2700k sucks. I would like to see how it does in 1440P gaming. We all know it’s awful in 1080P compared to more recent CPUs.

      • Krogoth
      • 2 years ago

      It is held mostly back by DDR3.

        • ikjadoon
        • 2 years ago

        What relevance does RAM have at 1440p? Likely smaller than 1080p.

          • Krogoth
          • 2 years ago

          Dual-channel DDR3 isn’t enough to keep an aggressively OC Sandy Bridge happy. It does affect it at 1440p under newer titles if you trying drive any of the high-end GPUs on the market.

          I already see this my current 3570k and Vega RX 64 setup.

      • ptsant
      • 2 years ago

      It only sucks if you have bad min framerates and, most of the time, this can be corrected with very slight adjustments in game detail. Very often the difference between a low-end and high-end CPU/GPU comes down to antialising, textures medium vs ultra and geometry medium vs ultra. I rarely notice that stuff while playing. Yes, you can definitely see it if you look for it, but it doesn’t significantly alter the enjoyment of the game, at least for me.

    • tsk
    • 2 years ago

    Would you look at that, AMD topping the efficiency charts.

      • chuckula
      • 2 years ago

      AMD should be proud that its most efficient chip can almost match the [url=https://techreport.com/r.x/2017_09_27_Intel_s_Core_i9_7980XE_and_Core_i9_7960X_CPUs_reviewed/scatter-power.png<]7980XE[/url<] in efficiency. If not performance.

        • Jeff Kampman
        • 2 years ago

        Man, I need to go fix that graph now, as well.

          • chuckula
          • 2 years ago

          The fact that you never really found the source of the bug is a little weird.

          Given that the AMD parts actually sped up somewhat (at the cost of efficiency) do you think it might have been some frequency governor that was being too aggressive with turbo-boosting the AMD chips?

            • just brew it!
            • 2 years ago

            That seems plausible to me.

            • derFunkenstein
            • 2 years ago

            Since AMD distributes a custom power plan as [url=https://community.amd.com/community/gaming/blog/2017/04/06/amd-ryzen-community-update-3<]part of the chipset driver install[/url<] (see 4/26/2017 update in that post), that seems like a reasonable guess. Power plan could have gotten hosed.

      • Klimax
      • 2 years ago

      For one particular benchmark. Although in Blender difference between SSEx and AVX is not that large, Intel chips have full 256-bit units for AVX and much wider L1/core bandwidth. (For regular chips that’s 32b/cycle, while -X chips got 64B/cycle)

      Different tests would show different stacking. (And not just different programs, but even different inputs can have brutal influence – movie versus video camera output under x264…)

        • Anonymous Coward
        • 2 years ago

        I’d like to see an integer-focused efficiency test. Maybe even javascript or browser benchmarks tested for efficiency. A lot of server time is spent moving around data and running inefficient scripting languages.

      • IGTrading
      • 2 years ago

      The question now is, why even bother with Coffee Lake ?!

      Sure, like Jeff correctly points out in every review, Intel optimized apps like audio processing or the Adobe suite, clearly justify the investment of the professionals using those specific tools.

      But for everybody else ?!

      Productivity : @stock AMD Ryzen is faster and consumes less power.
      overclocked AMD is again faster and consumes less power.

      Gaming: @FullHD Intel shows significant upsides of 5% to 15%
      @4K AMD’s Ryzen is less than 1% within Coffe Lake’s performance

      Platform longevity: AMD has a guaranteed upgrade path to 12nm in spring
      With Intel you absolutely never know ….

      Reliability: The thermal resistance of even the best TIM will rise by 25% over the course of 2000 cycles of heating and cooling and the temperatures are already super HIGH.

      How will those temps look like when the TIM gets old ?!

      Therefore, for any regular PC user or 4K aspiring gamer, what would be the point ?!

      • Growler
      • 2 years ago

      Thanks, Scott!

Pin It on Pinterest

Share This