Sign in with
Sign up | Sign in

Tom's Energy Efficiency Guide

Tom's Energy Efficiency Guide
By

energy efficient components

Power consumption and energy efficiency have become the most discussed items in IT hardware this year. While power consumption refers to the total input of power required for a device to operate at various given states, energy efficiency is even more relevant, as it relates performance to power consumption. The common approach is to calculate the ratio of performance per watt, whereas performance can be assessed depending on the desired unit of measurement. The industry has developed an awareness of the market sensitivity for this topic, and has been releasing products that specifically cater to an energy-conscious audience. Some of these products make sense, but many of them don’t, and as it happens, common sense is the most powerful tool to understanding how to optimize energy efficiency.

Power consumption readings for a device or an entire PC are very important to calculating the minimum and maximum power requirements when devices or PCs are idle or working at high loads, but they tend to be theoretical values. Minimum power requirements often do equal the actual requirements when a device or PC is idle, but maximum power draw only applies for components that are kept steadily busy. In most cases, maximum power requirement is reached only during a limited time, for example as long as a processor is busy compressing files into a zip archive.

As a result of application-dependent and hence severely varying power consumption, it makes the most sense to discuss energy efficiency from an application standpoint, like we did it in the articles AMD CPU Efficiency Compared and Intel Power Consumption Then and Now. The essential questions to ask are:

1. How much power does my PC require when it is running idle?

Extreme power-saving PCs offer little power consumption, but poor performance.

2. How much power does my PC require to complete a given task, and how quick can the task be finished?

Performance comes into the equation here: a slow solution might take forever to run, and thus consume more energy than a faster solution that goes back into an idle state much quicker after completing the task.

3. What’s the real energy efficiency (performance per watt over time) at a given workload level?

As you can imagine, the last question can only be answered by relating performance to power consumption for a given workload, which automatically includes a certain period of time that is required to complete the workload. The unit of measurement hence is not only watts, but watt-hours required to complete the job. While a device’s idle power requirements define the minimum power consumption you can expect, the workload type and duration of the workload define the peak power requirements and sustained peak power consumption during processing. Obviously, performance is quite an important factor as well, as a fast system will return into an idle system state much quicker than a slower energy-saving one, which might hence consume more power overall to complete the same workload.

We selected some system components to consider when assembling an energy efficient computer. Although there are low-power components available in almost all component categories, many of these don’t make sense, either because they are too expensive, or because the savings compared to standard components aren’t substantial enough to justify spending more. Let’s take a look.

Display all 42 comments.
  • 2 Hide
    Reynod , May 19, 2008 9:56 AM
    Achim and Roos does it again ...

    Another pointless article with no substance and just a heap of prattle and waffle.

    There is absolutely no point in reading this.

    You wasted 10 minutes of time that I could have better spent picking weeds in the garden.

    Start looking for another job boys !!!
  • 0 Hide
    Anonymous , May 19, 2008 11:29 AM
    Give 'em a break, reynod - recycling is everywhere these days. It's far more energy-efficient to re-resummarize their articles on energy efficiency than to come up with something new...

    Seriously, though. While it's nice to have a single comprehensive article to link someone to, as a starting point for "how do I 'go green'?" or whatever... this article does seem to be largely a repeat. I was a little disappointed to see that you didn't include performance and power figures for the example systems.

    As far as specific advice goes, probably the most helpful is the reminder to be sure to hibernate/switch off. It's surprising how many people still don't do that on a regular basis. Perhaps Tom's H. could offer an exploration of hibernate/sleep in XP & Vista? with an investigation of why "automatic" hibernate hardly ever seems to be an option...
  • 4 Hide
    jeffunit , May 19, 2008 1:02 PM
    I am disappointed in the quality of articles on tomshardware.
    Clearly there is no technical review before publishing.

    The article says:
    "because power consumption increases as the square of clock speed. In addition, voltage boosts further reduce energy efficiency."

    This is quite novel, as power consumption for CMOS is generally known
    to be a function of CFV^2, where C is capacitance, F is frequency,
    and V is voltage.

    There is no F^2 term there.
  • 0 Hide
    impar , May 19, 2008 1:05 PM
    From the article:
    "Any energy efficient processor, such as an Intel Core 2 Duo E8000, will no longer be efficient at all once overclocked, because power consumption increases as the square of clock speed. In addition, voltage boosts further reduce energy efficiency."

    Thats wrong.

    Voltage increases cause power consumption to square.
    Processor speed causes a linear increase in power consumption.

    Article is wrong in regard to processor speed causing
  • -1 Hide
    Anonymous , May 19, 2008 4:24 PM
    I agree with reynod- worthless article. Toms set their own article worthiness benchmark for their article "GPU vs. CPU Upgrade: Extensive Tests". Live up to that standard on all articles please.
  • 0 Hide
    njalterio , May 19, 2008 4:53 PM
    @ reynod: It's Patrick Schmid and Achim Roos. At least get the names right if you are going to call someone out!

    I do agree this is a worthless article. The misinformation in this article aside, I could careless about energy savings. Just another example of people being penny wise and pound foolish. So what if I could save $200 a year by cutting electricity costs. That would be the price to pay if you do not want to live in the stone age.
  • 0 Hide
    Anonymous , May 19, 2008 4:55 PM
    to jeffunit:


    CFV^2 is also the same as C^2 x F^2 x V^2.

    It is harder to create new but interesting articles.
  • 0 Hide
    1971Rhino , May 19, 2008 5:28 PM
    y'all totally ruined my warm fuzzy feeling I had for actually having a couple of parts the mentioned here......

    lol
  • 3 Hide
    jeffunit , May 19, 2008 5:48 PM
    To Zodiacfml

    Perhaps your basic math skills need some work.

    In general, ABC^2 is considered to be (A) x (B) x (C^2), as the ^2 binds to the nearest element. Though it is true that (ABC)^2 is the same
    as A^2 x B^2 x C^2, that isn't the formula for CMOS power consumption.

    To make it very clear:
    CMOS power consumption is a function of Capacitance * Frequency * Voltage^2

    There are other errors in the article, such as poor English. For example,
    "Make sure you find a motherboard that is either based on a small number of voltage regulator phases in order to save power, one that dynamically switches phases on and off as required."

    This should say ...power, *OR* one that...

    Perhaps the authors are not native English speakers. That's ok, but the article should be proofread for correct English, if it is being published in English.

    In general, the article is a fluff piece. I have been reading tomshardware for at least 10 years, and it has been going downhill.

    I didn't even notice a mention of using solid state hard drives for reduced power consumption. On my personal web pages, when I say that switching to a notebook hard drive saves me power, I have a measurement of how much power I save. It isn't very precise, but it is better than nothing. See
    http://www.weasel.com/comp_server.html#bit
    for my article about a 'low power bittorrent server'.
  • 0 Hide
    awolfe63 , May 19, 2008 5:56 PM
    I agree - we've been getting meaningless data-free articles with technical errors about one or twice a week now for 3-4 months. Someone needs to be let go.
  • 0 Hide
    psouza4 , May 19, 2008 6:15 PM
    FYI: it's "Achim Roos", not "Achim and Roos". One person there. And agreed: the quality at Tom's is horrible. I've posted several times already that you can't rely on hardly anything you read here anymore.
  • 0 Hide
    Jakt , May 19, 2008 7:12 PM
    Ha! I thought I was alone in thinking that the quality of these articles are getting out of control! While I have been lurking here for years, I have only recently begun posting, and only then has it been to correct some perceived deficiencies in the articles.

    It may just be that my computer expertise has risen, but I seem to recall that Tom's Hardware used to be highly technical and difficult for average reader to understand. They seem to have made it a bit "user-friendlier" over the years, but lately it seems to have been dumbed-down to the point of almost useless.

    I'm not giving up yet, as there are still loads of nuggets to unearth at this site, but I resent wasting time sifting through garbage to find them.
  • -1 Hide
    Anonymous , May 19, 2008 9:31 PM
    Well, I liked the article and never read the previous article you guys mention. Who cares? I glance at Tom's home page a couple times a day and sometimes I read and sometimes I don't. I don't give a rats ass if they recycle stories here and there. Plenty of new articles are written...

    You complainers need to get in perspective. But I did seem to like the Toms of a few years back a lot better, but this isn't too bad either. It's mostly the site design and organization I don't care for. It's really boring and not focused properly on the intersting parts. Bring back that initial design from about 3-4 years ago with the scrolling comments right near the top along with the single icon next to the articles. Old articles cycled off. That and the news...and simply DELETE the rest of the web site. Seriously. And leave these article comments, I like reading them.
  • 0 Hide
    randomizer , May 20, 2008 1:42 AM
    @Jakt: Yes, the site did use to be more "technical". But you know what? Technical doesn't make much money if you don't get many hits to the site. If you look at a webpage and say "WTH?" there's a good chance you won't hang around and read the whole thing, or ever come back, until you understand the content. By making the content easier to understand for Joe Computeruser, they attract a much wider range of viewers, while at the same time losing a couple of so-called "enthusiasts". Tom's Hardware is owned by the Bestofmedia Company, ie. they are trying to make money, not provide free tech support.
  • 0 Hide
    randomizer , May 20, 2008 2:04 AM
    I do have to poke a this paragraph:

    "The Core 2 Duo E8000 series, which is based on Intel’s 45 nm manufacturing process, is the most energy efficient dual core processor available today, and is the most powerful dual core model as well. We recommend going for any version, as the performance differences among them are small; if you stay below 3 GHz, the power consumption will also be low even under load. All E8000 Core 2 Duo processors have a total of 6 MB shared L2 cache, and they will run in all current Socket 775 motherboards. It’s usually possible to exchange an Intel dual core for a quad core processor, should you need more performance."

    First bolded phase: The E7200 which has only 3MB cache is actually the most energy efficient. It is also nearly half the price. Perhaps the article was written before this chip was released, but even so, this should be updated for accuracy.

    Second bolded phrase: I'm not sure how you define "current". I would assume that means P35/X38 and later chipsets. There are still some motherboards with compatiblity issues even with these chipsets, but usually a BIOS update fixes those issues. Unfortunately, some motherboards (Abit ones in particular) currently only properly support these chips with beta BIOSs, so you can't find them on the abit website.
  • -1 Hide
    Sunfighterlc , May 20, 2008 2:34 AM
    I just wish the green hysteria would end soon. Global Warming fear mongering makes the terror fear mongering look legit lol.
  • 0 Hide
    jeffunit , May 20, 2008 3:22 AM
    I don't have green hysteria.
    I do have 4 computers in a relatively small room, in a hot climate.
    Minimizing heat is important for me.

    For people who have computers in server rooms, server rooms now have
    pretty dense, hot computers, and the server rooms are usually limited by
    the amount of ac power going in, and the amount of air conditioning.
    Less power per computer means more computers, which generally means
    more money.

    Not green hysteria, but the desire to maximize server room resources.
  • 1 Hide
    Pei-chen , May 20, 2008 3:25 AM
    Tom's, please stop the your-PSU-is-too-weak mongering. A 500w high effiency PSU can easily power two 3870 or 9600 if all other parts were chose well.
  • 0 Hide
    randomizer , May 20, 2008 4:20 AM
    With 36A you'd be pushing it after a few years of capacitor aging.
  • 0 Hide
    cmj , May 20, 2008 7:53 AM
    The good point about this articles is that the whole concept of power saving is changing. While it was still 1 or 2 years back, we only talk about reducing the CPU frequency during idle time as the mean to power saving. We are now talking about cutting out power to each and every possible components. On many of the new motherboads, they already have the capability to shut down part of the circuitry, during idle time, to save even more power.

    However, as a Linux user, I feell like I am still left behind in the cold. These power saving are proprietary design from each of the MB manufactorers. However, I cannot see any of them are willing to release the driver for the Linux OS. May be we are not a customer group big enough for the right support.
Display more comments
React To This Article