I'll apologize in advance for the extremely long drawn out response, but it pertains and is important.
Yes, overclocking does decrease the life of your product. Only if your altering your voltage. If you purchase a card; an example would be a GTX 580 with this criteria:
Clock-Rate: 772 mHz
Shader Clock: 1544 mHz
Memory-Clock: 4005 mHz
Voltage: 1.08 Volts
If you overclock the card to these settings:
Clock-Rate: 860 mHz
Shader Clock: 1720 mHz
Memory-Clock: 4450 mHz
Voltage: 1.08 Volts
Your cards life will not be affected; it's still performing with no adjustments to it's voltage. If you adjust the voltage it will impact the life of your card. The reason is simple, by increasing voltage your also increasing your current (Amperage).
The higher your current and resistance, the higher your voltage. This follows Ohm's law.
Symbol:
Current (Amperage) = I;
Voltage = V;
Resistance = R;
Electrical Field = E;
So these electrical items are all directly related with each other.
V = I * R
Another example would be: Current Density (J), Electrical Field (E), and Conductivity (Ơ).
J = Ơ*E
As you can clearly see altering one of these items will directly impact all of them. The three primary formulas used to calculate current, resistance, and voltage are:
Your next question about Power-Consumption not really being mentioned, isn't true. It's actually a very relevant point. It's measured in almost each and every graphics card. That is measured through Wattage.
This will tie into our previous discussion. The formula that is used for wattage; is in the below method.
Symbol:
Wattage = W;
Volt = V;
Ampere = A;
W = V * A.
So voltage times amperage will attain your wattage.
To follow our previous example above, the GTX 580:
Clock-Rate: 772 mHz
Shader Clock: 1544 mHz
Memory-Clock: 4005 mHz
Voltage: 1.08 Volts
Watts: 244 Watts (idle), 478 (Peak-Load Wattage).
So we have a generalized reference now. But the above formula isn't truly relevant for our desired goal; which is performance per watt.
We would use:
The Ώ represents an Ohm. You'll be able to calculate and you'll notice that your graphics card is idling at 244 Watts, 12 Volts, and 20 Ampere. Now you'll notice that the peak-load is almost double the idle wattage; as you've seen these all directly correlate with each other. That means your card is actually doing: 478 Watts, 12 Volts, and 40 Ampere.
You can see the importance this will play in your machine. Not just in performance per watt; but can your Power-Supply adjust and power the card properly. If you aren't able to deliver the proper values, the card or Power-Supply may become damaged.
By using the formula, you'll be able to compare the statistics of a 680 and a 580. You'll be able to notice that the 680 is three times higher in most areas (clock, memory, stream processors) but draws less power. The 680 in idle draws around 194 watts. Plus the architecture of Kepler with the items being thread based; allows for more items to congruently function without increasing heat or wattage too drastic.
This is what makes these new cards so incredible. They maintain greater efficiency at higher speeds with less power consumption, even when overclocked. To help you visualize the difference here is an article that Tom's Hardware did comparing several cards.
http://www.tomshardware.com/reviews/geforce-gtx-680-rev...
You'll notice that the 680 is around 144% better in performance. Essentially this measurement is concluded by the computational value per watt. How fast can the card compute a task 100,000,000 million times for an example.
Which it will do 32 bit's at a time (Red, Green, Blue, Alpha). Those four items make up the 32 bit's.
So modern day cards are measured by how much wattage is used. What is the constraint of the possible power drawn. The peak performance is limited by the amount of power it may draw and how quickly the heat can dissipate. So the watt; is a direct translation to it's peak power.
Your final remark about overclocking your 670. Each card is made differently, though customer A bought card C, and customer B bought card C, from the same vendor. The silicon may be different. What I mean, is it someones card possess slightly more then it'll allow for more heat dissipation. As your still bound by the same amount of transistors, but there is either a greater thickness or surface area for the heat to spread to.
This slight difference, can be the difference between 1.1 ghz or 1.3 ghz.
Additionally that is only one slice of this significantly larger puzzle. Overclocking can be hindered due to room temperature, card temperature, cooling, heat dissipation, voltage, power-supply, and several other variables.
When you attempt to attain higher performance from your card you have to eliminate several variables. Work your way up to the “Overclocking Barrier.” Essentially move each slider until your card fails; once you've found the peak level on each section you then would begin undervolting or overvolting. This will allow you to either create less heat to push it further; or produce more current to push it further. The environmental variables will start to play a greater role.
Your particular card is a terrific card. It uses the same board as a 680, but at less wattage. Which will allow you to overclock past a stock 680, and five to ten percent less performance compared to the 680. This performance isn't computational, but merely slight clock differences and slightly higher power-consumption.
I'd like to apologize for this long explanation. But some of the fundamentals are important. As it will help you understand slightly better. If I've messed up any aspect, I'm hoping the community will correct me. It's been a long day for me, my brain is fried.
Best of luck overclocking. If you need assistance, the community is here for you.