AMD FX-9000 specifications

**Faithh** · 2013-06-02, 11:18 PM

Originally Posted by Skelly

Half right. Of course the number is an average. But saying "it's the average power draw that would occur in normal usage" makes it sounds like if my CPU has a TDP of 100W, I should be using an average of 100W right now, which is obviously false.

What people mean when they say "maximum power draw" is the average power that the component would use, under maximum stress given "real world applications" (furmark, prime95 etc).

http://www.xbitlabs.com/articles/gra...n_2.html#sect0
The site above even goes as far as labelling TDP as "Peak Power consumption" because most of the time they're the same damn thing.

http://www.anandtech.com/show/6774/n...nce-unveiled/2
Above is an anandtech article on the Nvidia Titan. Now, the Titan could theoretically pull 300W from the PSU because it needs 8+6 pin connectors. The card itself has a specification of "250W TDP". This means that at full load in something like Furmark, it will pull an average of 250W from the PSU. The Titan also has a "TDP limit" of 265W. This limit has nothing to do with how much heat the coolers can dispel. It's a function of the firmware limiting how much power the card can draw from the PSU.

Uhms in the bios it's described as board power limit or target, not tdp. Because people are using the term TDP wrong doesn't mean it's like that.

Here copied from an Intel spreadsheet:

"TDP is the recommended design point for the
thermal solution’s power dissipation and is based on running worst-case, real-world
applications and benchmarks at maximum component temperature. The thermal
solution must ensure that the processor junction temperature limit is never exceeded
under TDP conditions.

Official intel spreadsheet starting from page 9

It's clearly saying that the cooling must be reasonable under such loads and doesn't have anything to do with peak consumption.

**BicycleMafioso** · 2013-06-02, 11:42 PM

Originally Posted by Majesticii

Least we can argue the power draw of the fx9000 is beyond rediculous.
This is just a fx8350 on steroids.

Indeed. Which is why the power draw is really low and obviously a part of a cherry-picking/binning process.

**Cyanotical** · 2013-06-03, 01:34 AM

Originally Posted by yurano

I've been using Faithh's definition where TDP is "the maximum thermal power that it would draw when running 'real applications'". As I showed with some graphs earlier, this is a reasonable assumption. 77W TDP Intel CPU pushing 120W or so at the wall. 43W consumed to PSU efficiency, motherboard consumption and GPU idling sounds fair right.

which is really the same thing i am saying, the "real applications" part means average use

TDP is used for buying a heatsink, ie, you don't buy a 75W heatsink for a 100W TDP CPU (not that anyone here ever comes short on the heatsink)

Originally Posted by yurano

This is the part you're getting wrong.

First, define output.

by output i'm putting a rather absolute view on it, i brought up the car engine earlier because there seems to be a confusion about where energy goes, in a car engine most is kinetic, in a CPU most is thermal, but not all of it, if a CPU took 100% of the electricity and converted it to heat, then you could use TDP to judge the power draw on an average day,

but the problem is conservation of energy, if all of your electricity is being converted to heat in the process of "work" then there is nothing coming out the other end of the pipe as all of it was lost as heat, however, since we get electrical signal at the other end of the pipe we know for a fact (a very very hard fact) that not all is lost

this plays into TDP significantly, since not all of the electrical power that runs through a CPU is converted to heat (we have electrical output going to the rest of the board to confirm) then that means that TDP will always be less than the power draw, even if it's only by 1w, it will always be less

Originally Posted by yurano

Not to be disrespectful or anything, but this really falls into the category of engineering (specifically electrical) and not straight physics.

There's also the factor of how you presented the problem to them. What the CPU does isn't actually work in any sense, which is why I've been using "work" (with quotations) which refers to the ordering of electrons into something meaningful.

yeah, but somebody had to bring up a "i haz degree, i is better" so why not present the question "is all electrical energy that flows through a computer processor converted to thermal energy" to people i know that have the degree that was brought up and the result was a very fast "no" from both of them

the "basically" i got from my sisters fiance (BS in physics and engineering) was that the resistance in the materials used and gate design is that generates the heat, thermal energy generation is not the goal, but a byproduct, he said that the confusion might be coming is as the the type of conversion, 100% of converted energy is thermal, but not 100% of electricity is converted to thermal energy, not all energy is converted, but when conversion happens it's only to thermal

**yurano** · 2013-06-03, 02:32 AM