Travel Technology - How to shut down cores in a dual core proc?

In my never ending quest to gain more battery life from my Sony SZ....


I recently found out that Mac laptops have the ability to shut down one out of two cores on the dual core processors. Is there any way to do so on a Vista laptop? I've found ways to reduce the maximum processor usage, but I've yet to find out how to shut down a core...

I didn't think it would be possible, but I was wrong :eek:

http://www.windowsitpro.com/Article/ArticleID/20110/20110.html

There is a switch you can add to the boot.ini line called numproc where you can set the number of processor cores that Windows will use. I would recommend copying the existing line in your boot.ini and adding a new entry that includes the /numproc=1 switch so that you can relatively easily switch back and forth. It does require a reboot, but that isn't the end of the world.

Edit to add: that article seems to be based on NT4 and Windows 2000, but odds are it'll still work. There's also the /ONECPU switch that could work as well to do the same thing.

The utility on the Mac shuts down a core, but it doesn't actually save that much battery life. The system tends to just run the single core at a higher speed, eliminating any savings.

Wifi uses tons more power than the processor, so the best way to extend battery life is shutting down wifi. I find it adds 30%+ with no other changes (moderate disk/DVD activity and bright screen setting, so you should do a lot better if trying to extend battery life). Very surprising how much power goes to wifi (and bluetooth).

Wifi uses tons more power than the processor, so the best way to extend battery life is shutting down wifi. I find it adds 30%+ with no other changes (moderate disk/DVD activity and bright screen setting, so you should do a lot better if trying to extend battery life). Very surprising how much power goes to wifi (and bluetooth).

Oh no doubt. I'm well aware that WiFi is a battery hog....but even with WiFi off I can barely get 3 hours out of it...far from the 4 that Sony claims. Hell, my friend who just got a Macbook Pro gets over 6 hours WITH WiFi and while browsing the net and doing various other activities (with one core turned off and brightness on low)

You can disable this often in the BIOS. Look for hyperthreading or something similar.

If you are technically inclined, the best way is to throttle down the CPU itself. That way it will use less power and produce less heat, which in turn means less fan usage and power draw.

RMClock is the best around for this. It requires tweaking for each machine.
http://cpu.rightmark.org/products/rmclock.shtml

Their support forum is here:
http://forum.rightmark.org/?id=6

I got an extra 20 minutes on mine using it. Not to mention far less fan noise and a cooler running laptop.

One wouldn't get much, if any, power savings over disabling a core in the methods described. Unless specifically designed for it, the core is still power. The chip would have to be designed to have split power panes for it to work.

Reducing the core voltage is much more effective. Power draw is dependent on the square of the voltage, and is porportional to clock speed.

Wifi uses tons more power than the processor, so the best way to extend battery life is shutting down wifi. I find it adds 30%+ with no other changes (moderate disk/DVD activity and bright screen setting, so you should do a lot better if trying to extend battery life). Very surprising how much power goes to wifi (and bluetooth).

I did not know that. If this is true why does the CPU have heat sinks, cooling fans etc and the wi-fi stuff goes naked.

MisterNice

You can disable this often in the BIOS. Look for hyperthreading or something similar.

Hyperthreading is an old(er) technology. You won't find it on the current Core Duo and Core2duo chips.

Hyperthreading is an old(er) technology. You won't find it on the current Core Duo and Core2duo chips.

Many dual core systems have an option in the bios to disable "multi-core support" or something similar. I'm not sure that it will save any battery life - it might actually cost it, as WhoTF knows what clock or voltage the disabled core will run at, while if you leave both cores running they will be handled by the OS's power management.

Slightly OT, how do I check that both cores of a dual core are working and go to full speed?

Slightly OT, how do I check that both cores of a dual core are working and go to full speed?

The RMClock utility mentioned above might do it.

I know that CPU-Z does it; it's available at www.cpuid.org . It's the same site that hosts PC-Wizard, a nice little program that tells you just about everyone you want to know about the hardware on your system.

Jeff

One wouldn't get much, if any, power savings over disabling a core in the methods described. Unless specifically designed for it, the core is still power. The chip would have to be designed to have split power panes for it to work.

Reducing the core voltage is much more effective. Power draw is dependent on the square of the voltage, and is proportional to clock speed.

This is incorrect. The processor is built on a CMOS process. Power is consumed when the logic changes state, ie froma 1 ->0 or 0-> 1 transition. You turnoff the clock to the CPU and it will draw negligible power.

This is incorrect. The processor is built on a CMOS process. Power is consumed when the logic changes state, ie froma 1 ->0 or 0-> 1 transition. You turnoff the clock to the CPU and it will draw negligible power.

Processors are built on processes that are long-ago derived from CMOS, but it's not that simple. Look up "leakage current." Which process generation makes a difference - some of the older "recent" models (the Prescott Pentium 4s) were particularly bad about this.

In my never ending quest to gain more battery life from my Sony SZ....


I recently found out that Mac laptops have the ability to shut down one out of two cores on the dual core processors. Is there any way to do so on a Vista laptop? I've found ways to reduce the maximum processor usage, but I've yet to find out how to shut down a core...

In Vista, go into Control Panel, power settings. Pick the power setting you use (hopefully some kind of energy saver, although you can make any of the presets custom) and click on the "change settings" link. On that next window, click "change advanced power settings". In that box, you get all kinds of options. Go down to "Processor power management" and look at the "minimum processor state" and "maximum processor state". Typically, in a power save mode, the max state is 50%. This does pretty much the same thing as shutting down a core---computer runs slower and cooler with much less drain from the CPU. Make sure your laptop has a setting when plugged in that allows your CPU to run at 100% in this field--mine did not (HP laptop). It seemed REALLY slow until I found HP sets this to 50% by default (must be something about being "green" or something). The laptop battery does last longer when the CPU runs slower. Remember the P4 chips with "powerstep"? That did something like this automatically. Vista has that, too, but can also cap your speed in this setting. FWIW, I have found Vista vastly better than XP for this kind of thing, although you have to find the settings.

From my two years experience with the SZ:

[1] Both Sony and Intel have throttling and general power-save utilities, both came pre-installed on your laptop. You don’t need to run at 2 GHz for example to surf the web or watch a DVD.

[2] I have found that I save quite a bit of juice by turning off features that I rarely or only occasionally use or just when stretching battery power is important.

Using the device manager I usually keep the following turned off. If I need something it is simply a matter of clicking 'enable' to turn it back on.

Firewire port
PCMCIA slot
Express34 slot
Memory Stick reader
RJ-11 modem jack
RJ-45 LAN jack
Fingerprint reader

Individually these devices draw very little power, but all together I find that I get about 1/2 hour more out of my battery - oh, also make sure to use the integrated Intel graphics rather than the discrete nVidia card as that alone is good for another 15 minutes.

Using these methods I find I get just about 4 hours out of the slim battery and 6-7 out of the extended battery.

This is incorrect. The processor is built on a CMOS process. Power is consumed when the logic changes state, ie froma 1 ->0 or 0-> 1 transition. You turnoff the clock to the CPU and it will draw negligible power.

nkedel:
Processors are built on processes that are long-ago derived from CMOS, but it's not that simple. Look up "leakage current." Which process generation makes a difference - some of the older "recent" models (the Prescott Pentium 4s) were particularly bad about this.


very interesting information about increasing battery usage...

glob99 is correct. I know semicoductor processing and although nkedel talks like he knows what he is talking about, he is mistaken.

Processors are built on CMOS processes, where the oxide is used as the dielectric that creates the electric field that creates the conduit for conduction. There is no "newly created" process, that has replaced CMOS - bona-fide CMOS processes are still used. Of course, now we have multilevel copper metalization, and other interconnect improvements, but the CMOS process is the same as it was years ago, only scaled down.

Leakage current is the current that flows through the gate and reduces the dielectric effect (makes it more a resistor than a capacitor), which results in power being dissipated even when the clock is off. Processes now deposit a few atoms of a metal under the oxide to reduce leakage current, but leakage current has ALWAYS existed. So even though the leakage current is negligible (we are talking fA here, versus uA flowing when the clock is running), the power is always on for each transistor - and this is the problem with leakage current. When there are millions of transistors, this 'neglible' current can add up. nkedel is only talking about one product that had an issue with leakage current that has since been corrected.

From my two years experience with the SZ:

[1] Both Sony and Intel have throttling and general power-save utilities, both came pre-installed on your laptop. You don’t need to run at 2 GHz for example to surf the web or watch a DVD.

[2] I have found that I save quite a bit of juice by turning off features that I rarely or only occasionally use or just when stretching battery power is important.

Using the device manager I usually keep the following turned off. If I need something it is simply a matter of clicking 'enable' to turn it back on.

Firewire port
PCMCIA slot
Express34 slot
Memory Stick reader
RJ-11 modem jack
RJ-45 LAN jack
Fingerprint reader

Individually these devices draw very little power, but all together I find that I get about 1/2 hour more out of my battery - oh, also make sure to use the integrated Intel graphics rather than the discrete nVidia card as that alone is good for another 15 minutes.

Using these methods I find I get just about 4 hours out of the slim battery and 6-7 out of the extended battery.

Wow...didn't even occur to me to disable some devices....

I don't even own anything that uses a firewire...but I'm not quite sure what it's called under device manager. It's definitely not under "firewire." Any advice on this one?

Also, I'm always running the graphics in the "stamina" mode

In Vista, go into Control Panel, power settings. Pick the power setting you use (hopefully some kind of energy saver, although you can make any of the presets custom) and click on the "change settings" link. On that next window, click "change advanced power settings". In that box, you get all kinds of options. Go down to "Processor power management" and look at the "minimum processor state" and "maximum processor state". Typically, in a power save mode, the max state is 50%. This does pretty much the same thing as shutting down a core---computer runs slower and cooler with much less drain from the CPU. Make sure your laptop has a setting when plugged in that allows your CPU to run at 100% in this field--mine did not (HP laptop). It seemed REALLY slow until I found HP sets this to 50% by default (must be something about being "green" or something). The laptop battery does last longer when the CPU runs slower. Remember the P4 chips with "powerstep"? That did something like this automatically. Vista has that, too, but can also cap your speed in this setting. FWIW, I have found Vista vastly better than XP for this kind of thing, although you have to find the settings.

Yeah, I just discovered this a few days ago. I've yet to notice if it increases battery life, but I'll probably test it out next week in class. In my machine there's two settings for min/max processor state...one for plugged in and one for on battery. I set the max state on battery to 50%....we'll see how it works out.

Wow...didn't even occur to me to disable some devices....

I don't even own anything that uses a firewire...but I'm not quite sure what it's called under device manager. It's definitely not under "firewire." Any advice on this one?

IEEE 1394

Processors are built on processes that are long-ago derived from CMOS, but it's not that simple. Look up "leakage current." Which process generation makes a difference - some of the older "recent" models (the Prescott Pentium 4s) were particularly bad about this.

I know all about leakage current. The power dissipation due to leakage current is at least two orders of magnitude less than the dynamic power. I have designed integrated circuits and have a PhD in Condensed Matter Physics. If you would like a dissertation on CMOS leakage current I can oblige.:D

Hyperthreading is an old(er) technology. You won't find it on the current Core Duo and Core2duo chips.You will, however, find it making a comeback in the Silverthorne processor to be introduced later in the first half of this year for ultra-mobile devices and smartphones. (Silverthorne will use 45nm technology and have full C2D instruction set support, so it's hardly a throwback to when dinosaurs roamed the earth.) Hyperthreading has less analysis overhead than out-of-order execution architectures, so it's better suited to that application.

I know all about leakage current. The power dissipation due to leakage current is at least two orders of magnitude less than the dynamic power. I have designed integrated circuits and have a PhD in Condensed Matter Physics. If you would like a dissertation on CMOS leakage current I can oblige.:D
I defer to your PhD, but most chips out there are designed to run all cores at the same clocks or voltages. There are newer chips being designed to change that. I don't know how letting one core sit idle at operating frequency would be better than letting it consume a tiny bit of additional power but reducing the length of time both cores stay in the higher power states.