Power consumption of a Athlon 64 X2 3800+

[Le_Gritche]

The best you can do is getting some friends that overclock CPU's. They love high TDP's, while you want the low ones.

I thought a lot about that, now I can understand why overclockers love high TDP's.
High TDP means the CPU let more current flow through itself, showing it is of better quality, and so more prone to overclocking.
On the other hand, a low TDP core means the core resists more to the current flow (proof it's of lesser quality).
I know it's a very crude simulation but anyway : as P=U*I=U*U/R, if the core voltage is fixed, with higher internal resistance the core will dissipate less heat. So silencers might want low TDP cores.
But the core voltage has not to be fixed, because we can undervolt it.
The quality of the core (so the TDP too) will influence how well it can underclock. EDIT: removed erroneous part spotted by Mats

If you want to undervolt, the high TDP core will run at a lower voltage because it's of better quality.
I still can't figure out if the better undervoltage will compensate for the lower internal resistance though. We have P=U*U/R but we lack the relation between the internal resistance (acting as a simulation of the core quality) and the undervolting potential.

Obviously if you don't plan to undervolt (or overclock), you are better of with a low TDP core from the start.

[Mats]

In fact according to the .pdf , high TDPs have lower stock voltage than low TDP ones.

Where in the PDF? I haven't seen it, unless you're talking about single and dual cores, which are different core types and not comparable.

[TomZ]

High TDP means the CPU let more current flow through itself, showing it is of better quality, and so more prone to overclocking.

All I can say is, "no." Your analysis and conclusions are not quite right.

Higher TDP has nothing with quality. It only has to do with the thermal design of the package - its capacity.

Power consumption of CMOS process is related to Frequency * Voltage^2. Therefore, higher frequency requires linearly more power, while voltage has a squared relationship, so increasing or decreasing voltage has a stronger effect.

For overclockers, what is important is the available margin (difference) between power consumption at "stock" conditions and at overclock conditions. If a given processor is already at its thermal maximum at stock, there's not much you can do. But if there is a good margin, then you can OC away. Reducing the voltage helps a lot because of the above equation.

[Le_Gritche]

In fact according to the .pdf , high TDPs have lower stock voltage than low TDP ones.

Where in the PDF? I haven't seen it, unless you're talking about single and dual cores, which are different core types and not comparable.

You are right, I must have mixed everything in my head ; different revision, different number of cores...I will edit my post

@Smilingcrow
I have edited my previous post about Turion/Athlon64 cores btw...


@Tomz
I *think* you are wrong, but given the number of bullshits I have written on this thread so far, I will just shut up and let more knowleageable people speak. But can you explain that :

If a given processor is already at its thermal maximum at stock, there's not much you can do.

I was under the impression that the individual TDP was the individual max power dissipated at stock, and you seem to imply that all processors of the same type, whatever their individual TDP, dissipate as much at stock, but the higher TDP ones have a bigger margin to dissipate more. Is that right ?

Regarding power consumption in CMOS you are right , though I can only provide that link given by Jaganath on another thread and quote two parts :

A simplified equation for power consumption in a CMOS gate is P = CL*V^2*f + Iq*V, where CL is the load capacitance, V is the supply voltage, Iq is the leakage current, and f is the switching frequency.

In a typical CMOS fabrication process, the leakage current Iq tends to be negligible and thus has traditionally been ignored by designers. [...]Unfortunately, leakage current cannot be ignored for much longer. As CMOS fabrication technology improves by shrinking distances between chip features, leakage current increases dramatically because electrons can literally leak into layers of material that are supposed to be insulating. [...]

But I was not asking whether frequency or voltage effect was more important (I know the answer) but wether higher individual TDP or the related* better undervoltage potential in the same chip was more important.

*The 'related' comes from my shady reasonning, so it needs a confirmation or a rebuttal first anyway.