Sign in with
Sign up | Sign in

Idle/Peak Power, Performance Per Watt

Atom, Athlon, or Nano? Energy-Savers Compared

This is very interesting—despite completely different processor architectures and platforms, all three systems showed exactly the same idle power of 28 W. It’s amazing to see that a power-optimized Athlon 64 processor can reach the same idle power as Intel’s low-power Atom, which suffers from the unnecessary power consumption caused by the mainstream chipset. VIA’s new processor may be faster than the C7, but it cannot decrease idle power requirements any further.

We looked at peak power consumption in all of our benchmarks, but decided to take the SYSmark 2004 peak power results, as these are easy to track and not much different from what we saw using Prime95 or other tools. VIA’s low-power Nano system obviously can be the most power hungry solution as well, requiring up to 50 W of power. The AMD and Intel systems are rather close to each other—AMD did not exceed a 39 W power requirement, while Intel stays below 36 W at all times, despite its ordinary desktop chipset.

SYSmark 2004 Performance Per Watt

This is what many of you probably wanted to see most: We correlated the SYSmark 2004 scores with the Watt-hours required to complete the benchmark run. And clearly, Intel’s Atom platform is the most efficient. It offers the best SYSmark 2004 performance-–probably due to Intel’s Hyper-Threading technology and Atom’s better memory performance—and has the lowest power requirements. Please note that this result may be different in other benchmarks, depending on where the benchmark focus lies. We decided to run SYSmark, as it reflects desktop performance.

SYSmark 2004 Power Diagram

Lastly, you should have a look at our power diagram, which shows the system power consumption for every point of the duration of the SYSmark run. It shows that the VIA Nano L2100 system has the highest peak power and that the difference between idle and peak power for the Atom system is typically little, which reflects the small processor TDP of only 4 W. And the diagram also shows that the AMD system took the most time to complete this benchmark, followed by VIA’s device, while Intel’s Atom had the fastest time.

More Results: Compare with Core 2 Duo

If you want to see how the three system stack up against a Core 2 Duo E8500, please see the image gallery with additional test results

Ask a Category Expert

Create a new thread in the Reviews comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 46 comments.
This thread is closed for comments
  • 0 Hide
    jawshoeaw , October 3, 2008 7:19 AM
    I'd like to see how much electricity you would save in a year by having an efficient machine for basic home use - the one you could leave on 24/7 guilt free
  • 6 Hide
    alexander , October 3, 2008 7:53 AM
    There are some things with this test review that pussles me. Why did you use 3,5" drives? They draw about 10 watt instead of 2 watt for 2,5" drives. Also, I think you could have used a much more energy efficent power supply. That is probably why they all had the same idle watt; the psu was the bottleneck.

    I use a setup with the following:
    Jetway VIA C7 1.2 GHz
    picoPSU 60 watt power supply
    1 GB Kingston DDR2 667 Mhz RAM
    250 GB Samsung 2,5" drive

    This setup only draws about 20 watt when working and even less when idle (measured with a wall socket device, so I know it's accurate and total).

  • 4 Hide
    alexander , October 3, 2008 8:00 AM
    By the way, It would have been interesting also to see you review the dual core Atom.

    And maybe also compared to a more modest "normal" computer instead of a gaming rig, to see how low you can get with a normal PC.

    Otherwise an interesting article, as they most often are.

  • 0 Hide
    alexander , October 3, 2008 8:04 AM
    My last entry for today... ;)

    (And no, I don't work for the company...)

  • 0 Hide
    faithful , October 3, 2008 8:07 AM
    Here is a very nice review including the dual core Atom 330. I also has many more benchmarks.
  • 0 Hide
    faithful , October 3, 2008 8:09 AM
    This review could be seen by some as using very selective benchmarks.
  • 0 Hide
    Anonymous , October 3, 2008 8:20 AM
    my underclocked ADO5400IAA5DO consumes ~5W more than athlon in ths reaview, but I have 2x1000Mhz :D  as a bonus I can always relax minimum power requirement and take performance route a step or two :D 

    I sugest to try "AMD NPT Family 0Fh Desktop Processor Power and Thermal Data Sheet" document on the - interesting read

    by the way, my geode lx800 (500MHz) board on the full load fits into 6W :D 
  • 0 Hide
    Anonymous , October 3, 2008 8:21 AM
    It would be nice to build Core2 Duo (or even Solo) and under-clock it to similar power envelope (not very much unlike AMD system)... I wonder how It would compare with the rest of the bunch.
  • 0 Hide
    alexander , October 3, 2008 8:26 AM
    n/a, what power supply do you use?
  • 2 Hide
    randomizer , October 3, 2008 8:27 AM
    The WinRAR graph is wrong, or the comment about it is wrong. There's a typo in the Winzip comment.

    WinRAR: "Still, VIA’s Nano still is more powerful."
    Well, it looks to me like Atom won.

    Winzip: "Hence VIA’s Atom does well again."
    Oops :kaola: 
  • 1 Hide
    Anonymous , October 3, 2008 8:56 AM
    Really stupid test setup ...
    Using slowest AMD clocked 1Ghz vs 1.6Ghz Atom and 1.8Ghz via ... You should use faster x2 losing only few more watts but gaining fastest and best platform in test.

    Atom is including old platform slow crap, but this "test" is obviously aimed to show that AMD is bad, buy intel. Choosing BEST cpu from intel and VIA and testing it against SLOWEST AMD ... what is the point???

    This AMD 1Ghz/8W will have aprox 12W on 1.5Ghz ... and then including excellent 780G chipset will be total winner of all test including price, performance per watt etc.
  • 1 Hide
    randomizer , October 3, 2008 9:00 AM
    pifChoosing BEST cpu from intel...

    If Atom is the best, Intel is screwed.
  • 1 Hide
    faithful , October 3, 2008 9:33 AM
    Does CPU manufacturers sometimes pay reviewers for reviews? I was just wondering because I have it on other websites but fortunately not here.
  • 0 Hide
    cangelini , October 3, 2008 10:16 AM
    faithfulDoes CPU manufacturers sometimes pay reviewers for reviews? I was just wondering because I have it on other websites but fortunately not here.

    Have what, Faithful?
  • 0 Hide
    cangelini , October 3, 2008 10:18 AM
    randomizerThe WinRAR graph is wrong, or the comment about it is wrong. There's a typo in the Winzip comment.WinRAR: "Still, VIA’s Nano still is more powerful."Well, it looks to me like Atom won.Winzip: "Hence VIA’s Atom does well again."Oops

    Nice catch Random, fixed.
  • 1 Hide
    eugenparaschiv , October 3, 2008 10:50 AM
    The AMD processor is clocked at 1000MHz. One ideea for the next article would be to take a real 2000+ Lima (or even an X2) and underclock it until it reaches 10-15 W (not 8). This would be a much more fair comparasion with VIA, because that particular solution needs 18W, so you could argue that the bast comparison would be a VIA at 18W and a AMD also at 18W (probably a Lima at 1600Mhz, or a X2 at 1000Mhz). Any chance at this article being done?
  • 1 Hide
    faithful , October 3, 2008 10:59 AM
    Have what, Faithful?

    ..because I have seen it.."
  • -1 Hide
    zodiacfml , October 3, 2008 11:31 AM
    the atom processor would always win in this segment.
    the price. the design and manufacturing technology for the atom will allow intel and consumers on a win-win situation. profitable for intel and low prices for consumers while offering adequate performance for net use.
    i am sure the atom can still use less power.
    its as if, intel drove the atom to maximum clockspeed for the given die space and architecture so that it can achieve that adequate performance.
  • 2 Hide
    Anonymous , October 3, 2008 12:33 PM
    CPU-Z memory speed for Athlon X2 is right. K8 processors have minimum divider 1/5 from clock speed, so at 1000 MHz it just cant go above 200 MHz physical clock or 400 MT/s (DDR2-400). You can check it with C&C on any Athlon - drop to 800 MHz, and the memory goes DDR2-320 (160 MHz physical). So i wonder why you use horrible 6-6-6 timings for the memory? At DDR2-400 it should have no problems with 3-3-3.
  • 1 Hide
    coldmast , October 3, 2008 2:07 PM
    Some PSUs are most efficient for low loads, while others are better for high loads. However, if you use an 800 W PSU and only use 28-50 W, the efficiency will certainly not be in an ideal range. This is why we used the FSP220—it guarantees that the PSU runs within an efficient load corridor.

    I'm glad this was mentioned!
    I had problems with other reviewers that would insist on using 1000W power supplies for low power consumption hardware, the P/S efficiency is only around 50-65% on low power conversion.
Display more comments