calculating power consumption (a64)

[flexy]

I'd like to know what my real world benefit is from running a tool like rmclock (which underclocks/undervolts my A64 when idle)

I found a formula on thw web where they say

>>>
OC Wattage = TDP * ( OC MHz / Stock MHz) * ( OC Vcore / Stock Vcore )^2
The text "^2" means "squared".
>>>

The TDP for my A64 3500+ winch is 20W TDP at min pstate (1000mhz/1.1V) and 67W TDP at max Pstate (2200mhz/1.4V)

(see AMD thermal/power whitepapers for TDP values)

So.....i can use a formula like

67W * (2607 mhz/2200 mhz) * (1.53 Vcore / 1.4V)^2

OR

20W (948 mhz/ 1000 mhz * (0.87 Vcore / 1.1 V)^2

(my max OC is 2607mhz@1.53 VCore and min is 948mhz@0.87VCore)

With above formulas i get 94W/100W for max. and 11W/25W for the min p-state, using either the 20W TDP value or the 67 W Tdp value....

AFAIK 67 W TDP assumes full load, like running prime95 with al transistors switched on (which is no real world situation)


Any thoughts on this ?

Would this be a somewhat realistic idea of power-saving...NOT knowing whether TDP is actually "heat dissipation" or real world CPU power consumption ?

I know of course that for a real life SYSTEM there is MUCH MORE to put into consideration, all the devices and effectivenes of PSU and gfx-card etc...
But in this case i just wanted to know the benefit of running rmclock or Cool'n'Quiet regarding the CPU.

 

[Parasitic]

The formlua isn't really correct because you're assuming that the thermal output is linear with respect to clockspeed and squre with the voltage difference. Case in point, if you use minimum p-state data as your "stock" values and your stock clocks (max p-state) as the "overclocked" values then the calculation would yield something like 71W instead of the 67W AMD quotes.