How to Calculate Overclocked Power Consumption

[G L]

The purpose of this thread is to show you how to calculate the wattage of your processor, and how to tell how hot that should make your processor run under load. This is just a compilation of other sources, so thanks to all the websites I've linked to.

To begin with, note the formula near the middle of the following linked page. The author lists it as a "rule of thumb", by the way, so takes its results as an approximation only:

http://www.amdmb.com/article-display...D=105&PageID=5

OC Wattage = TDP * ( OC MHz / Stock MHz) * ( OC Vcore / Stock Vcore )^2

The text "^2" means "squared".

This formula is fairly self-evident. The stock/OC MHz is just how fast the processor is meant to run, and how fast you are in fact running it. Same with the vcore, which to any new people is the voltage that is run through the processor. Higher voltage means more stability, but also can lead to much higher power consumption, and if excessive, and early demise.

The only term that needs explanation is the TDP, or Thermal Design Power. The TDP is meant to be the wattage of the processor at load. I say "wattage" because it is unclear if this is meant to corrospond most immediately to how much power is consumed in watts, or how much heat is produced in watts, but as near as I can tell the TDP is pretty much meant to indicate both.

To get your processor's TDP, you must depend on AMD or Intel's own figures. In the case of AMD's XP line both maximum and typical are listed, and typical is what you want. Maximum is apparently if every single transistor were powered at the same time, which will never happen. Typical is meant to be the realistic maximum wattage.

http://www.amd.com/us-en/Processors/...9_7203,00.html
http://www.amd.com/us-en/Processors/...9_3748,00.html
http://www.intel.com/design/Pentium4...cp4f+tech_doc&

By picking the appropriate processor and scrolling through the pages, you should be able to find the TDP, and also the stock core voltage if you do not know it already.

For this example I will use the Athlon XP 2500+ desktop processor. Its TDP and core voltage information are found on page 29 on the second link on the Athlon XP page.

Athlon XP Model 10 512k cache 2500+
1833 MHz
1.65v
53.7 watts TDP

For this example I will use a relatively high overclock with a high core voltage, 2.5 GHz with 1.95v.

OC Wattage = TDP * ( OC MHz / Stock MHz) * ( OC Vcore / Stock Vcore )^2

OC Wattage = 53.7 * (2500/1833) * (1.95/1.65)^2

OC Wattage = 53.7 * (1.36) * (1.18)^2

OC Wattage = 53.7 * 1.36 * 1.39

OC Wattage = 101

So this 36% overclock and 18% core voltage increase has resulted in a 88% wattage increase.

If the overclock were instead 2.4 GHz with 1.95v, this would result in 97.8 watts instead. If, on the other hand, we did the same 2.5 GHz but with only 1.85v, this would result in 92.8 watts. If it were actually possible to get that 2.5 GHz at the stock 1.65v, this would result in only 73 watts drawn. So the moral of this story is, don't raise your core voltage unless you absolutely need to, and a weak power supply may hold back your overclocking.

Next: how wattage translates into temperature

[G L]

The measure a heatsink's ability to remove and dissipate heat is called its thermal resistence, which is expressed in C/W, or degrees per watt.

This figure it often stated on the manufacturer's website, and many reviewers also do so as well.

Here are some examples:

http://www.dansdata.com/coolercomp_p9.htm
http://www.dansdata.com/coolercomp.htm
http://www.overclockers.com/articles373/
http://www.frostytech.com

An example is the Zalman 7000a-cu heatsink/fan combo.

http://www.zalman.co.kr/eng/product/...79&code=005009

The manufacturer lists its C/W as .27 in silent mode and .20 in normal mode, the difference being the fan speed. We will stick with the latter figure. The .20 C/W figure means that the heatsink (and thus processor) will rise .2 degrees celsius for every watt of heat it draws off.

Therefore, 100 watts of power would result in a temperature rise of 100 * .2 = 20 degrees. Zalman's .20 C/W figure seems a bit low, so I will rely on frosytech's instead:

http://www.frostytech.com/articlevie...id=1585&page=5

They found a 24 degree rise above ambient at 100 watts, which translates into .24 C/W. Now we're able to calculate how hot our hypothetical 2500+ at 2.5 GHz 1.95v will be while being cooled by a Zalman 7000a-cu fan/heatsink.

101 Watts * .24 C/W = 24 degrees above ambient

The last step to determining how hot the processor will run is to add in the ambient. Let's say the air temperature in the case is 29 degrees Celsius. That means that the processor would end up at 24+29=53 degrees under load, which should be an acceptable temperature.

The Zalman 7000a-cu was a convenient example because it comes with its own fan. With other heatsinks that do not, thermalright for example, the fan will come into play. Higher-powered fans will push the thermalresistence figure lower, lower-powered fans will push it higher.

[VashTheStampede]

thanx again

[JohnMaclane]

i was wondering if u can calculate the wattage needed for ur computer so ur computer wouldnt be underpowered

[G L]

This page will tell you the power cunsumption of an overclocked processor, otherwise refer to the Intel/AMD links to look up the figure. This tool will help with the rest of the system:

http://www.jscustompcs.com/power_supply/

Though I believe the DDR consumption may be a bit outdated. You may want to double that...

[JohnMaclane]

thanks G L i hope it can help me with getting the right power supply needed for my new computer

hey joe if u need to calculate the power supply u cant just use cpu wattage there are other peripherals like thye mobo case fans which add to the overall wattage needed

[G L]

Yeah, the link I posted a couple messages ago is a tool that's meant to account for those variables.

[sak28wa]

Why worry about if your psu is enough for your sys? save a bit of $ and get one that'll be able to support your needs for the next few years and possible next several systems. I picked up a Enermax EG651P-VE (~550W) psu for about $150 USD.


Can be found
HERE!

[Mrsteve4011]

150 for a power supply............dont have that kinda money

[Jflo22]

This might sounds stupid, but does anybody know if there is a formula for calculating how much Voltage is needed for certain OC'd CPU's?
For Example: 3.2E P4 @ 3.6E, Stock voltage is 1.42, what would it need to be to cover the 3.6?

[G L]

I don't think there's any specific forumla, and even if there was something the problem is that different CPUs overclock differently, even if they're technically the same model. Processors have different OC capabilities just coming off the assembly line, but then newer batches will do better as well, sometimes substantially better. There are people here who have recent 3.2c P4s at 3.85 or 3.9 GHz on air, which used to require exotic cooling closer to when they first hit the market. To know what core voltage you'll need, the thing to do is just start overclocking slowly, and bump up the voltage when an OC fails and see if it helps you continue.

[adrenacrome]

Originally posted by Mrsteve4011
150 for a power supply............dont have that kinda money

You can get these 500W beauty's for about $69.99 on Xoxide.com's X-hour...

[Atomic]

Sanda can tell you the power rating in watts under the cpu and bios information module.

wow...just ran a few numbers using that equation and appearently i was putting 159w through my CPU when the TPD is 89w when I made my highest score in sandra at 2.5ghz....sheesh....thats alot of juice.

owww... my head.... math ..... AHHHH!!!!!

[G L]

Unfortunately, this isn't as relevant as it used to be because both AMD and Intel are no longer nice enough to post specific power consumption figures for each processor. Now they just list a maximum number for the whole line, or for a group of processor speeds.

[BuMp]

Thanx GL, u provided with some handy info.

[Dr. Vodka]

I cant even understand that guide, Man I speak spanish and a lil english not chinese!!!

[perera]

its not complicated just think a bit =P

2 points
: = divide
its just OC CLOCK : STOCK CLOCK
OC VOLTAGE : STOCK CLOCK
normal wattage multiplicated by result of clock :and stock clock. then multiplicate that result for the result of oc voltage : stock clock
gets like this :

(example)

3400:3000 = 1.133

1.45:1.35 = 1.074

(stock wattage) 89 X 1.133 = 100.57

therefore -

100.57 X 1.074 = 108



what id like to know is if fsb speed also diferenciate the watts consumage

[G L]

The equation is a bit easier to handle in the original article, about 2/3 of the way down the page here:

http://www.pcper.com/article.php?aid...e=expert&pid=5

Just substitute TDP for "MaximumListedPower".

The only problem is that I no longer trust AMD and Intel's TDP numbers. It used to be that each processor speed grade had its own TDP, now you just see one for the whole line... and I really don't see how that could be possible. I thought perhaps SOI or something about the Athlon 64 might explain it originally, but then Intel started doing the same exact thing.

So for instance the 89 watts for the .13 micron Athlon 64 does allow you to do the equation, but unless you have the fastest processor for that TDP it presumbably results in too big a number. Of course, you can always knock off a few watts for each speed grade and approximate...

----
Oh yeah, its actually (1.45/1.35)^2 = (1.074)^2 = 1.15... more like 116 watts, I think.

[perera]

but what about the fsb? overclocking it also changes the wattage?


and actually its a p4, and it is in the top of its series

i did the math on that power supply test, and i got 320 watts ..and im still with a generic agps400q of vcom... lol

she really does suffer, i already got 11.3 volts on 12+ with it

[G L]

Well, guess I might as well give a newer example than the old beloved Athlon XP. Here's the Intel 6xx series datasheet:

ftp://download.intel.com/design/Pent...s/30638202.pdf

If you go to page 76, you'll see that the TDP of the 3.8 GHz P4 670 is listed as 115 watts, and the vcore is rather confusingly specified as 1.2 to 1.4 volts on page 25, though of course if you own the processor in question you just use hardware monitoring to see what vcore you're running. I'll use what is apparently the worst-case scenario, 1.4v.

So let's say we OC to 4.2 GHz, which apparently is quite attainable with sufficient power and good cooling, and let's say that requires 1.6 volts. That gives us:

OC Wattage = TDP * (OC MHz / Stock MHz) * (OC Vcore / Stock Vcore)^2

OC Wattage = 115 * (4200 / 3800) * (1.6 / 1.4)^2

OC Wattage = 115 * (1.105) * (1.142)^2

OC Wattage = 115 * 1.105 * 1.304

OC Wattage = 165.7

For 166 watts, if you want both decent temps and barable noise you're going to want a fairly monstrous heatsink or spring for water-cooling.

[G L]

Originally posted by perera
but what about the fsb? overclocking it also changes the wattage?

Most directly by changing the CPU speed... but changing the FSB will also overclock the chipset, which would presumbably also increase wattage. Though if you knew the chipset TDP and voltages, you could presumbably run the same equation for that. RAM is another one, but same deal. Though I don't think chipset and ram consume much more than 25 watts each stock. CPU and video card can both consume over 100 watts each, but after that I'd be suprised if the rest of the system combined topped 65 watts, assuming you're not running a DC pump or TEC module off your power supply, of course.

[perera]

changin a bit the subject, i know that the recommended +- voltage is 5% , but im gettin somtimes 11.3 volts at 12+..is it that bad?

[G L]

Sounds pretty normal to me... I have a $200 power supply, and I still only have 11.65v right now.