Measuring Power taken by AMD processor

CCWCCW Suffolk, UK
edited July 2003 in Hardware
I am embarking on a project to measure exactly how much power AMD processors draw to aid me in writing up my heat dissipation article. I wasnt pelased with the new one so its been re-written but id prefer to colelct primary data for myself instead of relying on secondary data (other peoples data) as i am now as this would make the article much better.

Basically, I need help in working out how to do it. Basically, my idea is to have a motherboard. Secondly have a sepoarate Zif socket where the processor being measured will be in. This zif socket will conenct to he one on the motherboard and a True RMS meter conencted to the relevant pins I need to, to measure the power.

What do you guys think?

Thanks,
Craig

Comments

  • TexTex Dallas/Ft. Worth
    edited June 2003
    Killer. Who are you writing the article for anyway? cough...cough...cough... I may know of a site interested in aserious hardware articles... cough...cough...cough....

    Tex
  • dydxdydx Cymru, UK
    edited June 2003
    Dig arround to find the documents AMD release for mobo developers to design mobos for their CPUs. That will give you some clues as to which pins youre going to be dealing with. Finding a socket 462 might be difficult. Maplin dont seem to stock them.

    Best of luck with this project either way.


    mD
  • CCWCCW Suffolk, UK
    edited June 2003
    Already have the pin guide so I know exactly what pins, also about socket i have that covered, a guy from Epox is sending me one.

    Tex, articles to be hosted on my site which we have discussed in PMs a few days ago I do believe.

    Craig
  • dydxdydx Cymru, UK
    edited June 2003
    Looks like youve got it sorted dude.

    Cant wait to read it.


    mD
  • CCWCCW Suffolk, UK
    edited June 2003
    Well, i need to know if this will work because this has been done only a few times as far as im aware, there is already an older version of the article out "there" but im not telling where "there" is because 1) the old version sucks 2) the new one is much much much better and if i can pull this off will be even better.

    Craig
  • DexterHolland911DexterHolland911 Hong Kong
    edited June 2003
    CCW - Sounds like a good idea! You would be mesuring the current and due to the handy voltage readings onboard would calculate power? To be honest I would just get the values from AMD, they are probably more accurate. But this way you could see how the power output/current of the processor changed during load or when you add extra fans to the mobo etc. Would be interested to see this pulled off.
  • TexTex Dallas/Ft. Worth
    edited June 2003
    If he measured though he could do it a increased voltages and OC'ing which I think may be hiis point. To see how much effect those introduce into the rated power requirements

    Tex
  • CCWCCW Suffolk, UK
    edited June 2003
    Tex has hit it on the head, its about Overclocking as its for my Processor Heat Dissipation Article.

    Craig
  • CCWCCW Suffolk, UK
    edited June 2003
    *bump* c'mon guys! :D
  • dydxdydx Cymru, UK
    edited June 2003
    Stop talking and get to work! :D


    mD
  • CCWCCW Suffolk, UK
    edited June 2003
    i would but my current work relies in the answer i get from this thread

    tmod seems to know a lot about BIOSes wonder if he would know about this?


    Craig
  • dydxdydx Cymru, UK
    edited June 2003
    Are you testing a variety of CPUs or just the one? I reckon the Tbred 1700b is a must for this test.


    mD
  • Cool-CanuckCool-Canuck
    edited June 2003
    Like Dexter mentioned, AMD has white papers which so the max and average power usage of their chips.

    You could also work this backwards. Seeing as how heat is directly related to watts, you could measure the temp and convert it to power.
  • DexterHolland911DexterHolland911 Hong Kong
    edited June 2003
    CCW - What else is needed? What pin are you going to use and how? You'll need to take a current reading so you'll have to cut that wire and attach an ammeter?
  • CCWCCW Suffolk, UK
    edited June 2003
    Originally posted by dydx
    Are you testing a variety of CPUs or just the one? I reckon the Tbred 1700b is a must for this test.


    mD

    I am testing a variety of processors. ALso, definetly the THB B Core processors, they rock!

    Originally posted by Cool Canuck
    Like Dexter mentioned, AMD has white papers which so the max and average power usage of their chips.

    You could also work this backwards. Seeing as how heat is directly related to watts, you could measure the temp and convert it to power.

    Ibe been doing this previously in a way using formulas, I however do not tlike the formulas because its hard to agree on a formula that will be accurate.
    Originally posted by DexterHolland911
    CCW - What else is needed? What pin are you going to use and how? You'll need to take a current reading so you'll have to cut that wire and attach an ammeter?

    This is one way Ive been looking at it except ive been in contact with some electronics engineers and they suggested a True RMS meter.

    I was having a look at the pins and there are a lot, I didnt count but the majority of them supply power to the core so it looks widely distributed so I have no idea which to measure.

    Craig
  • DexterHolland911DexterHolland911 Hong Kong
    edited June 2003
    I am well educated in Physics but to be honest I know little about CPU electronics. Can't help ya there.
  • fatcatfatcat Mizzou Icrontian
    edited June 2003
    is this sorta on the lines of what ur doing?

    http://computernerd.com/cgi-bin/thermalcalculator.cgi

    fc
  • CCWCCW Suffolk, UK
    edited June 2003
    Welcome to Icrontic!

    It is what im after, to calculate wattaages but I need first hand data. The problem with that site is that the formula used, its hard to agree on a non-load or load number that represents the CPUs work if you get what I mean and this is 0.88 on all of the sites that do this because edn-user software will never use more than 88% CPU cycles apparently. This is a worst case scenario so not the exact data Im looking for.

    Craig
  • CCWCCW Suffolk, UK
    edited June 2003
    Here is my article so far: http://ccw.develo.co.uk/articles/phd1.asp
  • dydxdydx Cymru, UK
    edited June 2003
    Looking good CCW.


    mD
  • fatcatfatcat Mizzou Icrontian
    edited June 2003
    Originally posted by CCW
    Welcome to Icrontic!

    It is what im after, to calculate wattaages but I need first hand data. The problem with that site is that the formula used, its hard to agree on a non-load or load number that represents the CPUs work if you get what I mean and this is 0.88 on all of the sites that do this because edn-user software will never use more than 88% CPU cycles apparently. This is a worst case scenario so not the exact data Im looking for.

    Craig

    ah!!...ok..and thanks for the welcome but ive been with icrontic since oct 2000 ;)... ill read ur article a little latter

    fc
  • JPPJPP Stuttgart Germany
    edited June 2003
    Just a thought.
    I think this project the way it is planed is very effording. Measurement insrumentation manufactureres are offering Current Measuremnt Clamps which can be used to measure the current draw through a single wire directly. Hence if you clamp the device to the wire at the ATX plug supplying the CPU voltage regulators would enable you to measure the change of the current draw depending on the differing V core.
    I am aware that this will not give the real current draw value of the CPU but it would make it easy to measure the effect of the V core variation.
    JPP
  • CCWCCW Suffolk, UK
    edited July 2003
    Originally posted by Tex
    Killer. Who are you writing the article for anyway? cough...cough...cough... I may know of a site interested in aserious hardware articles... cough...cough...cough....

    Tex

    PM me ;)



    Article now at: http://www.ohmshouse.org/articles/phd1.asp

Sign In or Register to comment.