SLFM- battery life?

When my laptop drops down to 800mhz through SLFM, is any other voltage being affected? If not, there is no point in allowing it when on battery since I have the lowest possible undervolt at full speed.

 

A CPU uses power every time it switches.

Therefore the more switches per second the more power used.

Declocking does reduce power consumption. Of course the biggest gains come reducing both speed and voltage.

 

having certain VID and Frequency on the CPU almost does not reflect on how much power your CPU will use, and/or dissipate. That depends on what the CPU is doing as work.

besides, I bet you're not using your CPU at full speed all the time when you're on battery, so why would you want to disable the SLFM ?

And that about using more power when switching is a nonsense.

 

When a transistor switches from on to off, thats when it consumes power. There is a leakage current but that is usually pretty low.

While there are certain techniques to lower power consumption at idle, voltage certainly affects power consumption. Depending on the generation of CPU it can skip cycles when not needed so frequency has less of an impact.

The core 2s do benefit a lot on lowering the voltage even when idle as a lot of the fancy power saving tech was put into the Core i series.

So please, unless you have a grasp of the fundamentals at work, don't jump in.

 

<- electrical/electronics engineer, thank you. I would like to say that I'm waaay past the fundamentals.

if you'd try to look into how many times a modern CPU actually switches its frequency you'd probably start thinking that this would make it use much more power. Yet what happens is overall power savings, highly dependable on what the CPU actually does.

at idle modern CPUs would put their cores in lower power states that are not controllable by the user voltage-wise. Those states use much lower VID than the SLFM VID, so undervolting at idle does not do much. Besides, I dont think you can undervolt lower than what is for the SLFM anyways, therefore undervolting at idle = pointless.

also, no CPU can skip cycles. What matters is what the CPU will do when the cycle comes in.

 

Well you still seem to find it hard to distinguish between saying that when a transistor switches it uses power and the CPU switching frequency which are two different things.

The CPU stepping up or down in frequency itself will consume no greater amount of power than normal.

With the Core i series CPUs intel realised its actually better to go full burn for a short period of time and park the cores.

However the core 2 series can't do this as well as the core i series, its power consumption at complete idle is higher.

Depending if he is doing any voltage tweaking he may already be running his CPU at the SLFM voltage.

What I was saying to him is that a CPU at the same voltage will consume more power at 2ghz than it will at 800mhz.

I'd also like to point out that the FSB is dropped so the rest of the system will use a bit less too.

 

again, power consumption of the CPU is dependent to the high extent on what the CPU actually does.

I can make mine for example run at 2.8GHz locked dual core at 1.075Volts and while the CPU does minimal work it will stay way cooler and use less power than if I make it locked at 2GHz at 1.025Volts but load it at 100%.

besides,

I still cant see any sense of why you brought that topic about power usage when transistor switches, as this is so minimal and irrelevant, not worth even talking about it. Nonetheless the P-N junction leakage current.

 

Because its the very reason a CPU consumes more power if its at the same voltage but different frequency.

It's also the reason why it consumes more power at load as more of the transisitors switch.

I you know, answered his question. Why having the frequency drop actually helps lower power consumption.

 

Oo getting spicy. THanks for all the insight everyone!
Does SLFM mode affecting anything other than the CPU? System bus etc..

 

I believe the system bus declocks to 800mhz, you can check using CPU-Z or throttlestop.

 

FSB remains the same as far i remember. The issue with Core 2 Duo was the voltage at the multiplier. SLFM just makes the clock to 800MHz with normal 800MHz or 1066MHz FSB with the multiplier being dropped.

 

I don't think its a 3x multi.

I think its 4 x 200mhz.

 

so let me speak fundamentals here, as it appears I'm not the one having problem with those:

Consumed power

= Voltage [V] x Current ; (for a given moment in a DC electrical system, such as in computers).

Frequency - reflects consumed power (and dissipated heat) over period of time.

Thus, judging about power consumption with having the VID and the Frequency of a CPU is pure knowledge fail.

I haven't heard of a system bus working at 800Mhz yet. The 800MHz that you refer to could be either the Front Side Bus (i.e. CPU to northbridge), or the CPU cores working in SLFM mode.

the SLFM, or Super Low Frequency Mode, refers to cutting the bus clock speed in half, i.e. from 200MHz to 100MHz, thus allowing the CPU to cut clock speed too using it's lowest multiplier. However, as preset by Intel on many of their CPUs, when running in SLFM the CPU does not use it's lowest multiplier, but one or two steps above that. Specific example with Core 2 Duos is at SLFM mode the CPU uses x8 multiplier rather than x6, and at 100MHz bus the CPU is working at 800MHz.

with some tricks it is sometimes possible to make the CPU run at its lowest multiplier using SLFM bus.

 

A transistor only consumes power when in the middle of switching. It should consume none when fully on or off. When it does this is leakage current (which they all do ofc).

Therefore the more times it switches from on to off then the more time it spends between on and off and the more time current actually flows. Therefore at the same voltage a higher frequency transistor consumes more power than a lower frequency one.

A perfect transistor would switch instantly from on to off and never actually consume power.

Oh and of course when I am talking 800mhz I mean the quad pumped FSB frequency (real 200mhz). That was from memory however since quads dont support it. All I know is the FSB frequency drops.

As for this "system bus" clock, I have no idea what that is in relation to the core 2 arch as unless something is locked on its own clock source everything tends to be derived from the FSB.

 

<- physicist, chemist and materials engineer, thank you very much and I'd just like to say I am WAAAAAAY past the 'fundamentals'

Thats the problem everyone on this forum has



I got a p7350 to 1.5ghz MAX frequency. 200mhz fsb. I forgot to check on what the heck SLFM was doing, but I am pretty sure it was getting the CPU to 500mhz or something.

AFAIK P series lowest voltage is 1.05v.

Also I'm positive lower frequencies save battery life and reduce heat regardless of the voltage, if thats even what you're arguing about here.

 

Its 0.9v for idle IIRC.

1.05V is minimum for quads.

 

a transistor also consumes power if it needs to support the logical "1" when other circuits are connected to it for the purpose of "reading" its value. The more circuits connected, the more current (i.e. power, considering the same voltage) goes through that transistor. Put millions of transistors there and you got noticeable power consumption.

frequency doesn't always make a transistor to switch. That is dependent on many other elements in its circuit. For example: it a transistor was at logical 0 and at the rising edge of the first impulse it switched to logical 1, it doesnt mean that with the next impulse it will switch back to 0 - it may just remain at logical 1 for the next n-number of impulses (or clock cycles). What would determine if the transistor switches or not is based on what the CPU does as work .... so back to what I was saying before.

 

That is True.

Its 600MHz i believe. With my i7 quad now , its more like 900MHz when idle.

 

Hi

do you think I have a problem with my degree ?

thank you

P.S. Lowest voltage of a CPU would depend on the quality of the batch that comes off the line, within certain range for that model CPU. This is where the specs of the chip come into play. Just for info - SLFM of my old T7500 was 0.85V (lowest spec for the model chip), and of the current T9500 it is 0.925V (not the lowest spec for the model chip).

and you can not judge of battery life savings only from lowering the frequency. Does it help - maybe, depending on what work the CPU does at the same time.

 

Or in a laptops case it would depend on the minimum value set by the specifications for the socket.

Plus you keep going on about work, but we are assuming the same workload here so it's a constant.

 

Lets put it this way. If frequency had little impact on battery life intel would not bother changing it.

 

I have to agree with meaker on this point. The Frequency does have quite a effect to be frank. I'm no Electronic Engineer(Aeronautical lol) but i'm pretty sure of the fact.

 

because they are trying to squeeze every bit of possible power savings. In the case of same continues work load - yes, the battery will last a bit longer. That however doesnt mean that more work will be completed due to the extended battery life.


but we never assumed same workload though, I just saw that in your other post. My points is that the workload is what affects battery life to the great extent, not the frequency of the CPU. The CPU can run on free clock cycles without actually doing much, thus frequency doesnt matter so much.

- let me put it this way:

if work is done by the CPU, heat is generated, so power is used. I think we can all agree with that. Consider same cpu fan speeds for the cooling purposes:

I can make my CPU run at 2.4GHz locked and idle, and it would keep ~48 deg C.

I can also make my CPU run at 1.2GHz locked and idling, and it would keep a bit lower temp, ~46 deg C.

so for the CPU to drop 2 deg at lower frequency it means power savings, or this is your point. If I load the CPU at both frequencies, the temperature difference will be close to what the difference is when both are idling, i.e. couple degrees.

So if I load my CPU while at 2.4GHz, it will exceed 80 deg C (same fan speed). My point is that from loaded 2.4GHz CPU to unloaded 2.4GHz CPU there is much more power savings than from frequency drop at same work load. And the temperatures prove that.

- if you assumed same work load then lowered frequency will cause longer battery life, yes, but this is not guarantee that the work (if any) will be finished as now it would take much longer time to complete.

 

But SLFM is only activated under low load conditions.

 

true.

you also can not set the VID any lower than what is already preset for the SLFM. Rarely you could set lower multiplier while under SLFM to achieve lower CPU frequency as well, so for pretty much every user SLFM is the lowest setting.

the OPs question was weather or not to leave SLFM available for power savings. Might as well leave it active for the tiny battery savings that may or may not occur.

there are other ways of saving tiny bits of power too, such as switching off the LAN card, the modem, etc. A bigger one would be to lower the brightness of the display.

 

Well, wlan is almost always essential (though I think I've left it maximum performance), display @ min.. backlighting used if needed.
Never thought of turning off lan or odd even, for that matter.
IIRC shouldn't be more complicated than dev. manager +disable..?

 

miro_gt is right.

I've done some tests with my Asus UX30 (Intel GS45, SU7300)

I've always thought that lowering Mhz (FOR A GIVEN FIXED VID) would correspond to less power consumption, but from my tests it seem true only with full workloads, so that the cpu stays almost always in C0 power state.

Both in idle and real world usage (browsing, office, video) cpu load values are highly variable, so fastest the work is been done, fastest the cpu goes into lower power states (C1->C2->C3->C4->C6)

So, for example monitoring power consumption with HWInfo32 while watching Youtube 480p videos, the laptop consumes 0.4/0.5 watt more if I lower multiplier to 6x (1200Mhz) from 6.5x (1300Mhz) and more surprisingly +1 watt in SLFM mode (100x8=800Mhz)

CPU vcore fixed to 0.875v (SLFM VID), with Crystalcpuid I can use it even in HFM, High Frequency Mode


In idle no difference between the three configurations.

 

I would use a wall monitor rather than software, especially for such small changes.

 

Most wall monitors are not accurate enough to measure such tiny differences in power consumption.

That is exactly what I found when I tested a T8100 Core 2 Duo. These CPUs were redesigned so running them slow at idle doesn't save you anything. They spend most of their idle time in one of the deeper sleep states where the ThrottleStop FID/VID settings make virtually no difference.

Here was the testing I did.

http://forum.notebookreview.com/6369641-post5060.html

The Core 2 Duos also use lower VID voltages internally when they enter the deeper sleep states. That's another reason why the VID voltage setting in software will make little to no difference in power consumption at idle. It's already at a much lower VID voltage so this request is ignored when in C3/C6.

Here's an example for my T8100 from the Intel specs.

VID=0.650 - 0.859 V ~ C4
VID=0.600 - 0.850 V ~ DC4
VID=0.350 - 0.700 V ~ C6