Does voltage or frequency cause heat?

[SaltBoy]

Really dumb question - suppose that I run my i7-2600k at 4.2 Ghz with a set, non-variable voltage of 1.35v.

Then, suppose I bring the frequency down to 4.0 Ghz but keep the exact same, non-variable voltage (1.35v).

In stress tests, will both overclocks (in theory) produce the same amount of heat because both use the same voltage? Or will the 4.2 Ghz stress test produce more heat because it's just a faster speed? What happens if I do a 4.2 Ghz test at 1.30v instead of 1.35v?

I know that both overclocks won't require the same voltage in order to be stable, but I was just wondering, anyway...

 

[aigomorla]

BOTH..

if you make the cpu run faster, it is doing more work, hence getting hotter.

its physics... u cant gain something without losing another.

 

[nyker96]

BOTH..

if you make the cpu run faster, it is doing more work, hence getting hotter.

its physics... u cant gain something without losing another.

second this! however just to add, voltage changes tends to be squared, so shall we say it probably contribute to more heat. Freq increase to heat is linear.

A corollary to this: increasing freq w/o increasing stock voltage tends to increase efficiency of the system as a whole.

 

[aigomorla]

A corollary to this: increasing freq w/o increasing stock voltage tends to increase efficiency of the system as a whole.

until you reach the 4.0-4.2ghz territory... well if you can pull 4.2 on stock you have one hell of a cherry cpu.

 

[toyota]

and you should not need 1.30 never mind 1.35 for just 4.2. around 1.23 should get the job done for just 4.2.

 

[toyota]

until you reach the 4.0-4.2ghz territory... well if you can pull 4.2 on stock you have one hell of a cherry cpu.

what is "stock"? there is no official stock voltage and most bios will have auto or manual setting. if you leave it on auto it will likely use way more than needed when you raise the multi. if you do it manually then you will have to just figure it out based on what others are using.

 

[aigomorla]

what is "stock"? there is no official stock voltage and most bios will have auto or manual setting. if you leave it on auto it will likely use way more than needed when you raise the multi. if you do it manually then you will have to just figure it out based on what others are using.

typically stock voltage is defined at the base voltage for which the cpu is required to run.

Its hard set inside the cpu.. its not a motherboard feature..

Now Stock values arent all the same.. some have really low stock.. some have really high stock... this goes back to the ever so popular steppings.

But typically favored are the low stock voltage cpu's as they give you the best overclock within a usable range, followed by the high stock cpu's which are leaky but at the same time said to give you the highest overclock on LN2.

 

[PlasmaBomb]

P = cFV^2

P = Power
c = capacitance
F = Frequency
V = Voltage
(IIRC)

Increasing either the frequency or voltage will result in more heat being dissipated.

 

[Kenmitch]

what is "stock"? there is no official stock voltage and most bios will have auto or manual setting. if you leave it on auto it will likely use way more than needed when you raise the multi. if you do it manually then you will have to just figure it out based on what others are using.

About the only way to possibly see what stock voltage is would be running the chip with all energy stuff enabled at stock speed voltages on auto. Most of the time it'll be close. Voltage will very by load even tho 100% loads. Something like IBT with AVX support would be best. Some MB's have normal for vcore setting but not sure if any of the SB. MB's have it.

 

[SaltBoy]

P = cFV^2

P = Power
c = capacitance
F = Frequency
V = Voltage
(IIRC)

Increasing either the frequency or voltage will result in more heat being dissipated.

this.

Thanks, that's EXACTLY what I needed.

 

[Idontcare]

For data and the physics see this thread.

The bottom line is that the power-consumed (which results in heat generation) is dependent on both voltage and operating temperature.

If you keep your CPU voltage the same, keep the clockspeed the same, and simply allow its operating temperature to increase (have less efficient cooling, dust clogged fins on your HSF, etc) then the power consumption will sunstantially increase.

Alternatively, keeping everything the same while changing clockspeed is going to change power-consumption as others have noted (the special case of the generalized equation noted above).

 

[SaltBoy]

For data and the physics see this thread.

The bottom line is that the power-consumed (which results in heat generation) is dependent on both voltage and operating temperature.

...image removed...

If you keep your CPU voltage the same, keep the clockspeed the same, and simply allow its operating temperature to increase (have less efficient cooling, dust clogged fins on your HSF, etc) then the power consumption will sunstantially increase.

...image removed...

Alternatively, keeping everything the same while changing clockspeed is going to change power-consumption as others have noted (the special case of the generalized equation noted above).

Dang, you've done your homework - very impressive! I don't know if I should reply in this thread or that one, so I guess I'll do it here -

In your other thread, you bring up a "dynamic" vs. "static" power graph:

I bring it up because it reminded me of another question that I've never had an answer for yet.

I've always wondered if using variable (aka auto) voltage when using CPU throttling makes sense or if I should keep the voltage "static" at all times even when the CPU runs at its throttled speed (i.e. should I keep voltage at 1.32v when it runs at both 1.6 ghz AND 4.2 ghz?).

Right now I have my voltage set at variable, but from your data, it doesn't seem like voltage in and of itself affects power/temperature as much as frequency and/or cooling options. The voltage is needed just for a stable overclock at a particular frequency and everything else depends on cooling for low power consumption/heat.

Do I have that right? Or does excess voltage at lower speeds, all other things equal, generate more power/heat than is necessary?

 

[Idontcare]

I think you are using the term "throttle" when you mean to use the term "idle".

Throttling means the CPU has hit TJmax, 98°C with these SB's, and the CPU forces the clockspeed to reduce as much as is necessary so that the temperature doesn't exceed TJmax (98C).

Idle means the core utilization is low, so low that the CPU goes ahead and starts enabling/entering lower power "C" states (C1E and Speedstep) in which the multiplier is reduced below that of stock such that the core clockspeeds decrease.

Since the chip is idling in the first place, the power consumption is rather low to begin with regardless whether the idle voltage is set to a low value (say 1V) or if the idle voltage remains at the OC voltage.

With SB and power-gating, the idle voltage becomes even less relevant to power-consumption. You are looking at a difference of say 3-5W.

Here's my rule of thumb. If you are over-volting, then use an offset voltage method in OC'ing. It raises both the load and idle voltages but your CPU will still auto-reduce voltage at idle. Its free to do so why not do it?

But if you are undervolting, for example my 2600K doesn't need the stock-clocks Vcc to run stable at 4.5GHz. When I clock to 4.5GHz I actually decrease (negative offset) my voltage despite the fact I am overclocking.

This creates a problem at idle, the negative offset actually drops my idle voltage so low that the system will randomly crash when idle. Setting my BIOS to manual (fixed) voltage remedies this idle-voltage issue.

Although really its an artificial issue, I'm splitting hairs in terms of power-consumption by undervolting at 4.5GHz and so on. Its just 5-10W here or there, we are talking about maybe $0.25 difference on my monthly power bill.

 

[denev2004]

P = cFV^2

P = Power
c = capacitance
F = Frequency
V = Voltage
(IIRC)

Increasing either the frequency or voltage will result in more heat being dissipated.

Well..According to the Quantium Version 4, the "Capacitance" is going to change as long as you change the V or the f, but they still say it is going to be right for you to think of "Increasing either the frequency or voltage will result in more heat being dissipated"

 

[tweakboy]

Thanks for the lesson, Idontcare!