Sign in with
Sign up | Sign in

Efficiency: Performance Per Watt Shootout

Does Intel's Dual-Core Atom Improve Efficiency?
By

SYSmark 2004 SE Efficiency

Here is the deal: the numbers you see reflect the actual power in watt-hours used to complete the SYSmark 2004 SE run. The Core 2 Duo clearly required the least power, followed by the Atom 230. The Atom 330 dual core finished last, utilizing the most power for this workload.

We then divided the SYSmark 2004 SE score by the number of watt-hours used. The result seems disappointing for the Atom 330, and attests the great efficiency of the Core 2 system—again.

PCMark05 Efficiency


We did the same efficiency testing by tracking power and benchmark results using PCMark05. In this case, the differences are less significant, but the Core 2 Duo E7200 system was still the solution that required the least power to complete this benchmark.

In this benchmark the Atom 330 dual-core looks much better, as it provides clearly more performance in PCMark05, which helped to complete the benchmark run earlier and conserve some power this way. The Atom 330 is chasing the Core 2 machine, but it still doesn’t stand a chance.

Display all 31 comments.
  • 1 Hide
    Anonymous , February 5, 2009 7:55 AM
    Doesn't seem like much of an improvement from its single core predecessor compared to the difference between the core solo and duo; however, you can't argue with the performance per watt statistics. I wonder when we will see these in web oriented laptops like the Asus Eee PC?
  • 6 Hide
    dangerous_23 , February 5, 2009 8:30 AM
    for f*** sakes watt-hours is energy not power
  • 3 Hide
    apache_lives , February 5, 2009 8:54 AM
    cost is still the higher priority over performance and power etc - its still cheaper then a celeron etc, and is another performance increase in terms of usability NOT benchmarks - same deal as the Pentium 4 era with HT - the HT made systems seem more responsive reguardless of benchmarks - any P4 HT owner will aggree with me, but still yes it has weak (but sufficent) performance.
  • 2 Hide
    nihility , February 5, 2009 9:53 AM
    Which one of these tests checked multitasking performance?
  • 7 Hide
    3lvis , February 5, 2009 9:53 AM
    Some how it doesnt seem fair to compare atom to desktop CPUs. A better comparison would have been mobile CPU's to atom.
  • 4 Hide
    salgado18 , February 5, 2009 9:55 AM
    I still think this article, although great, misses one point: 95% of the time we use a pc, be it notebook, desktop, cellphone, whatever, the processor is idle. Nobody would buy a netbook to compress large files all day, or render complex scenes. They buy it to surf the web or type stuff. So it would be interesting to see a small one-day marathon: give three editors the pcs above, and measure the power used over one day. I bet the Atom beats Core 2 easily.
  • 7 Hide
    nihility , February 5, 2009 10:19 AM
    salgado18I still think this article, although great, misses one point: 95% of the time we use a pc, be it notebook, desktop, cellphone, whatever, the processor is idle. Nobody would buy a netbook to compress large files all day, or render complex scenes. They buy it to surf the web or type stuff. So it would be interesting to see a small one-day marathon: give three editors the pcs above, and measure the power used over one day. I bet the Atom beats Core 2 easily.


    I bet the editors will want to smash the atom PCs with a hammer by the end of the day.
  • -1 Hide
    chjade84 , February 5, 2009 11:29 AM
    I still think that the best place for these is in netbooks. My friend has an Asus Eee PC with the Atom in it and he loves it for college classes. It's great for taking notes and browsing the web and can go something like 6-7 hours on a charge. When used like this it really shows how much power efficiency can help.

    It even plays Starcraft! :) 
  • -1 Hide
    salgado18 , February 5, 2009 11:30 AM
    chjade84It even plays Starcraft!


    Now that's a surprise! It should be benchmarked! :D 
  • 1 Hide
    Anonymous , February 5, 2009 12:32 PM
    I would like to know efficiency of the power supply used in this review with 20 W, 40 W and 60 W loads.
  • 1 Hide
    cknobman , February 5, 2009 12:43 PM
    nihilityI bet the editors will want to smash the atom PCs with a hammer by the end of the day.


    Exactly. What a complete and total failure. 80% of the benches showed worse performance for the dual core vs single core Atom. Intel's atom processor is a joke.
  • 9 Hide
    slowstuff , February 5, 2009 12:50 PM
    How do these 4+ year old benchmarks test Multi core processors? Oh that's right, they are from the era just before benchmarks started caring about multiple cores. How about you disable one of the Cores on the Core 2 & see how much it drop... probably not much if any as 98% of the work is happenind on 1 core.

    How do these tests represent real world useage? Oh that's right, they don't. who runs 4 year old benchmarks, encodes music or does rendering on a netbook / netbox.... NO ONE with a brain.

    What use is a set of test that don't represent anything realistic or even test the processor fairly? Oh that's right, none. Seriously, test it in REAL WORLD USE against each other, test it's multi taking capabilities VS the single core system (230 vs 330).

    Lets test a moped vs a motor cycle & see which one wins in a drag race & used the most gas, that is essentially what you just did... IT COMPLETELY MISSES THE POINT OF THE OBJECT BEING TESTED.
  • 3 Hide
    pug_s , February 5, 2009 12:57 PM
    I don't understand why Tomshardware or any other hardware websites try to use an AMD equivalent. AMD already have something to compete with the Atom with their AMD Geode NX, which is based on the Athlon XP processors. Intel Atom processors is not built based on the Centrino design rather than a P5 at a higher clock speed.
  • 2 Hide
    danimal_the_animal , February 5, 2009 2:11 PM
    FAIL!
  • 4 Hide
    loftie , February 5, 2009 2:22 PM
    Considering the idle power is the same, could we not undervolt/underclock the C2D and see how much more power the C2D would use at load then. I'd find that interesting, especially if it still out performs the atom, which I'd assume it would.

    +1 for a low power AMD being thrown into the mix and a possible undervolt/underclock on that too - I'm sure AMDfan girl will be able to comment on this aspect :) 
  • 1 Hide
    bounty , February 5, 2009 3:13 PM
    PCmark 05 works fine as a test for simple web browsing, email etc. File compression is not something you do all day, it's something you do maybe once a day. When I do it, I don't want my machine to poo it self. WinRar is multi optomized, and it shows. Encoding music should be possible on a desktop, and we're talking desktop platforms.

    A good test to possibly add might be PCMark05 plus either WinRAR or iTunes as a resonable multiple work load test. We need a "some bloated flash/java sites" + IM + youtube test... or lets just see how these machines handle youtube fullscreen @ 1680x1050. Does the Atom fall apart while doing all that plus watching a movie? Watching an AVI on a "desktop" while browsing web(flash/java/super complicated+email+IM) seams a reasonable standard.
  • 2 Hide
    bustapr , February 5, 2009 3:24 PM
    How could you guys compare a netbook cpu with the super efficient fast e7200 desktop cpu!Its like comparing an athlonx2 to a phenomII.
  • 2 Hide
    Area51 , February 5, 2009 3:58 PM
    Toms,
    on page 4 the slides shows Core 2 Duo with HT which is not correct. please advise.
  • 3 Hide
    jeffunit , February 5, 2009 5:01 PM
    As slowstuff points out, it is pointless to run single threaded benchmarks on multi core processors. They could have used lame-mt, but didn't (it only has 2 cores). Of course there will be minimal differences between a 1 core atom and a 2 core atom on a single threaded benchmark.

    Try to use some multi threaded benchmarks, or at least run several benchmarks at the same time. Toms hardware keeps on getting worse, I think I will stop reading it.
  • 1 Hide
    jeffunit , February 5, 2009 5:02 PM
    I meant to say lame-mt only uses two threads. Still a better test than plain lame though.
Display more comments
React To This Article