Results 1 to 9 of 9
  1. #1
    New Member
    Join Date
    Mar 2005

    Why does Pentium-M Desktop&AMD 64 socket 939 use so much watts throttled,gamepc test?

    Basically, I want to know why there is always 40-50 watt less energy being used in the Pentium-M system despite the fact that that difference does not seem to be accountable by the energy utilzed by the AMD 64 socket 939 processors as explained below.

    I am trying to understand why those systems still ~100 watts in Pentium M
    and 140watts for socket 939 throttled idle tests at gamepc?

    What is the the ~22 watt rating for the Amd 64 running at 1000 Mhz thottled if the system shows 140 watts. Does the motherboard and harddrive and memory still gobble energy. Does the 22 watt rating the maximum thermally wasted energy and really the processor gobbles a lot more energy than that? Where does AMD list the actual energy gobbled, rather than the thermal energy wasted?

    Tomshardware wrote that the AMD socket 939 3500+ processor did not use more than 32 watts at full speed in their tests. So I am so confused as to where the other 130 watts is going to (the gamepc test shows about 160 watts with just the processor running without the video card so 160 - 32 = 128 watts going somewhere???????) Where?

    Finally, why does the Mobile Pentium in the gamepc test always use 40 - 50 watts less than the AMD socket 939 processors in the gamepc tests? If tomshardware 32 watt max energy usage for the AMD64 3500+ 939 is correct, then where is this extra energy going to in th system when it has all the same components except motherboard. Is it to the nforce4 chip that uses that much energy? Or something else on the motherboard. Or does the Pentium M motherboard manage the components like harddrive and memory diffently and thus it uses somehow less energy? There is somethin fundamental that I do not understand here.

    Please explain if you have some insight into where the difference in the system is?

  2. #2
    Join Date
    Jan 2004
    Sorry, but there really is no technical answer to offer here. Simply put, different processors have different thermal ratings. This can be due to a number of things including: die shrink, number of transistors, cache size, etc.

    Also, keep in mind that the P-M is a MOBILE processor, not a desktop solution. So it's very much an error to compare the wattage of a desktop s939 chip to a processor that will never be used in the average desktop Intel system.


  3. #3
    Join Date
    Jan 2005
    possible power draws:

    - power supply efficiency is probably 70% - a 120 watt idle reading drops then to 84 watts actually being used - poor efficiency always blows out figures.
    - on the pentium-m the memory is running half as fast, thus using half the power. possibly less than half - the voltage could be lower and timings looser, these settings aren't specified. i've not seen an actual analysis, but i remember reading that an older 128mb ddr dimm consumed around 10 watts, let's say 40 watts for 1gb nowadays. half the speed of that, and 20 watts goes away.
    - the northbridges consume different amounts of power, regardless of load. the pentium-m's chipset in the test was designed for mobile applications (855), thus consuming far less than the a64's desktop-oriented chipset. many other of the board components fit this description also.

    a large part of system power consumption lies with the motherboard manufacturer, and this review is no exception. from all testing i've seen, the winchesters draw between 30-35 watts at load, for the 3000+ to 3500+ speed grades. these figures are from multimeter tests, and also derived from ac-wattage tests with decreasing vcore/mhz and fitting the power draw figures to wattage calculations.

  4. #4
    Member Quailane's Avatar
    Join Date
    Oct 2003
    Who stole my white stars?
    Those pages you linked to don't work.

  5. #5
    Join Date
    Jan 2005
    i believe they're supposed to point to here and here.

    i was thinking about holding out for the 478-479 socket adapter, but given how limited it sounds like it'll be and the lack of good 478 boards coming out, i think i'm going with 90nm a64's for my low-power-consumption fix.

  6. #6
    Member dropadrop's Avatar
    10 Year Badge
    Join Date
    Mar 2001
    Tomshardware was talking about the processors powerusage, the other one is talking about the whole system. One can also wonder how Tom has measured the powerusage of the processor, though I won't claim it's wrong.

  7. #7
    Member speed bump's Avatar
    Join Date
    May 2004
    Montana tech Butte mt.
    they are testing how much power comes out of a wall outlet by each computer and then they have a formula that they use to somehow take CPU power consumption out of that. By the way those numbers are useless as they don't specify the time period that your computer chews up 200 watts so it could be one second or 1 year.
    Big changes coming: Intel e4300, DFI Blood iron, Crucial DDR2-8500, ATI x859xt, MCW-6002>Ehiem 1250>Bip3 with 3 nidec beta Vs, Sunbeam Zorro, Corsair 650 watter
    Stuff I should be using but don't: Finally got around to running that stuff.

    Ever run a benchmark? support our forums by posting it in the 3dmark team and system benchmarking team sections.
    My heatware

    Silence may be golden but performance is platinum

  8. #8
    Join Date
    Jan 2005
    well they're not useless - they tell you how much power the computer draws when it's idling and when it's loaded (figures drawn from sustained periods of course). if your computer is usually idle, you can use that figure to calculate how many kilowatts you use in a month. if it's usually folding, you can use that number. or you can spend $25 on a kill-a-watt wattmeter and know for sure my pc draws 62 watts idling and 78 watts at load.

    i don't know how tom's figured out the cpu draw, there are a few ways but none of them are really fail-safe. the best is to use controlled frequency throttling to find the linear decrease in power draw with frequency, and voltage decreases to find the exponential change, then match the graphs. though that doesn't take into account the efficiency curve of the motherboard's power regulation, that takes a bit more work.
    Last edited by yeha; 03-11-05 at 12:02 PM.

  9. #9
    Member Lifthanger's Avatar
    Join Date
    Jun 2004
    Wöschbach LMAO
    I have to agree with yeha on this and with his explanation for the initial problem above.
    Problem with most PSUs is, that their efficience drops bad when they only put out <35% of their power rating.
    And that's exactly where systems are when idling with a ~500W PSU.

    Gentoo AMD64 - Fluxbox

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts