Undervolting a Laptop

Delph1 writes "Laptops often comes with two Achilles heels, heat and limited battery time. There are, if not cures, at least remedies to make them less obvious. By lowering the voltage to the processor you can not only drastically lower the heat dissipation, but also increase the battery time significantly. NordicHardware gives a nice walk through on the process and was able to boast 18% lower temperature and a 20% reduced power consumption."

Comments

Underclocking

[selfabuse]

ATI Tray Tools (or a similar program) will let you underclock your video card too. Good for when you have a hulking gaming laptop, but aren't playing games, and don't want to use it as a space heater for your living room.

Re:Underclocking

[mjh49746]

ClockGen can also be used to undervolt/underclock supported motherboards in desktop systems, too. I routinely use it to save power and lower my temperatures when I'm not doing anything CPU intensive, like Folding@Home.

http://www.cpuid.com/clockgen.php

"They are not in Baghdad. They are not in control of any airport. I tell you this. It is all a lie. They lie. It is a hollywood movie. You do not believe them."

I'm glad to see this

[eclectro]

It should also help to save the "special purpose".

Counter productive maybe?

[squoozer]

Surely if you drop the voltage your are going to have to under-clock the processor (reasoning that to over-clock you need to increase the voltage). Most processors for laptops already throttle the processor down when under light load now-a-days which must be a great energy saving. Would under volting it really then save more or would you just end up with a laptop that is dog slow? I'm sure if it was this easy one of the big laptop producers would already be doing it as a 20% increase for basically nothing would give them a fantastic advantage.

Re:Counter productive maybe?

[Anonymous Coward]

There are definite gains to be made, even at the top speed; you just need to experiment. It works quite well because you can specify a voltage for each clock speed (or use the default tables). As long as you don't crash the OS, it won't go any slower than it would have otherwise.

Lots of laptop manufacturers are bundling appropriate software to do this for you these days, too.

Re:Counter productive maybe?

[NitsujTPU]

You end up with a laptop that is dog slow. You're right, most modern laptops throttle themselves effectively in order to reduce power consumption.

What the guy is doing, however, is trying to lower the voltage consumption to the line where the processor starts to behave a little flaky, and then pumping it up just a bit over that. Processors are made in big batches, some of them just work better than others. If yours happens to be one of the good ones in the batch, you can reduce the voltage while maintaining performance (not needing to bump down the clock speed).

If you really obsess over it, you go into the research that my roommate does, where he spends endless hours, days, and weeks tweaking processor floor plans and running them through simulators. You might hope to build a more efficient processor through all of this.

I wouldn't recommend doing this if you're not partial to your laptop randomly hanging while you're working on it, but everyone needs a hobby.

Re:Counter productive maybe?

[bigtrike]

Would under volting it really then save more or would you just end up with a laptop that is dog slow?

You might be removing the ability of the system to manage its own power. This was the case with my desktop. Dropping the CPU frequency on my P4 based desktop actually made it consume more electricity. At its factory speeds, the system uses abotu 90W when not doing a whole lot, and about 215W when under heavy load. Dropping the CPU frequency to 300MHz caused it to idle at about 110W usage. I did not experiment with dropping the voltage however, which may have produced a net savings.

Re:Counter productive maybe?

[Anonymous Coward]

"Surely if you drop the voltage your are going to have to under-clock the processor (reasoning that to over-clock you need to increase the voltage)."

It's possible. But since the power consumed is proportional to the square of the voltage, you'll see a net decrease in power required per unit computation.

"Would under volting it really then save more...?"

Yes. Power consumption is linear with clock speed. Besides, throttling the processor helps when the processor isn't doing anything. Lowering the voltage helps all the time, so the two could be combined for added effect.

"I'm sure if it was this easy one of the big laptop producers would already be doing it..."

Actually, I think if you look at the core voltages in processors over time, they've decreased, so in a way you're right.

At some point, the transistors won't switch anymore due to the low voltages, but if there's wiggle room to exploit, it's worth looking into at least.

Re:Counter productive maybe?

[Anonymous Coward]

Not really. Sure, you WILL come to that point as you lower voltage, but it should be common to get significant undervolts without having to reduce clock speed.

Re:Counter productive maybe?

[yppiz]

I underclock and undervolt my laptop using the RightMark CPU utility.

Speedstep can only throttle my processor down to 600MHz (from a max of 1.2GHz) but underclocking reduces it to an effective 300MHz.

I do not notice the performance hit, and I do a lot of photo editing on this machine.

--Pat

Re:Counter productive maybe?

[Marxist Hacker 42]

Photo editing, for the most part, is more memory intensive than CPU intensive.

Re:Counter productive maybe?

[VlartBlart]

My two home pc's are both laptops (it's a space thing - even though I have 16 USB devices stuck in one so I may as well not bother) - I can honestly say that they never run "dog slow". You can run Avid on them whilst listening to mp3s, surfing the web and making toast and they don't give a dogs doo-doo.

I think laptops have a bad press as under-powered machines (or dogs) - but they're as good as a decent desktop.

What I'm trying to say (under the influence) is that the power management of a laptop chip does not effect its performance. They can be sprightly little mothers!

Thanks to CPM (Chip Power Management (I made that one up)) - you can drag the uptime up from 24 hours to - oooooh, maybe 36 hrs before it melts down :)

Re:Counter productive maybe?

[evalf]

There are plenty of software that allow to stress-test the processor in order to ensure that the CPU is stable at the voltages that are set, such as prime95, that is mentioned in the article. It does not take "endless hours" to do that either: you just set the voltage you want to use, launch the stress-test utility, go to bed, and check if there are any errors in the morning... Then you can effectively determine the minimum voltage that is required to keep a stable system.

Re:Counter productive maybe?

[Anonymous Coward]

I wouldn't recommend doing this if you're not partial to your laptop randomly hanging while you're working on it, but everyone needs a hobby.

I undervolted my laptop a month ago. I haven't had a single crash.

Re:Counter productive maybe?

[frdmfghtr]

Actually, after having read the article, you do get the savings without a hit in performance.

Here's how I understood what was written:

When the processor is running at a particular clock rate, it is supplied a certain voltage. Reduce this voltage, and the processor clock likewise slows down. This feature is not changed.

What IS changed are the voltage thresholds when this speed shift happens. For example, when the processor was running at the reduced clock speed, the voltage (VID) was 1.000 V. However, the author was able to reduced this voltage down to 0.925 V. Hence, when the processor was set to run at the lower clock rate, the VID was only 0.925 V instead of 1.000 V. He then adjusted the settings so that the clock runs at it's original reduced speed with the new lower voltage.

For the faster clock rate, the VID was 1.450 V. However, he was able to get the processor to run at full speed at 1.175V. Again, the clock speed is the same, but the VID itself is lower. Thus, for each speed state of the processor, he was able to run it at a lower voltage.

The best analogy I can think of is the final drive ratio on a car; you have two gears, low and high, and an engine that normally runs at two speeds, say 1000 and 2000 RPM. You only drive at two speeds, 25MPH (1000 RPM) and 50 MPH (2000 RPM.) You tweak the gear ratio in the transmission and engine speed such that, in the end, the car still drives down the road at 25 or 50 MPH but now the engine turns over at only 850 and 1900 RPM. Low and high road speeds are unchanged, but the engine speeds are lower.

Why don't laptop manufacturers do this? They would have to tune these voltages for each individual processor. I'm no expert in overclocking, but if I understand it right, same-model processors can be overclocked at different rates: If you and I have the exact same model processor, you may be able to overclock it more than I can overclock mine, due to manufacturing tolerances. The same principle seems to apply to undervolting; it has to be done in a controlled fashion on a machine-by-machine basis, over a period of several hours.

Re:Counter productive maybe?

[NitsujTPU]

Congratulations?

I'm not sure if you're offering moral support to folks who want to do this, or pointing out a flaw in my rationale.

Ok, I overclocked my old tower and didn't have problems for years.

You can do all sorts of stuff if you feel like it.

I feel like posting to Slashdot, then eating my dinner and returning to my research.

Re:Counter productive maybe?

[Anonymous Coward]

How did you determine your wattage usage? I can only guess on mine, but I know it's a big difference in power consumption when I set my Athlon64 down to 1GHz and the voltage down to 1.1V. On the Asus Probe, I can see the 12V rail jump up at least 0.3V after tuning the CPU down with ClockGen. Obviously I'm drawing a lot of power running it balls to the wall. Then again, it's a desktop. The only laptop I've got is a quaint little Toshiba T4600 i486SL laptop from like 1992 or something. Really the only reason why I use ClockGen instead of leaving it to the Cool 'n Quiet feature is that sometimes I'll get unwanted horizontal lines in my video files when I vidcap in Divx format. If I leave it off, then I don't get the lines.

Now I'll call it a night.

"They are not in Baghdad. They are not in control of any airport. I tell you this. It is all a lie. They lie. It is a hollywood movie. You do not believe them."

Re:Counter productive maybe?

[NitsujTPU]

Nifty :-D

Re:Counter productive maybe?

[marcosdumay]

That is not what the article is proposing. But even if you underpower your processor so much that you need to underclock it, you still save power. Power consuption increases with the square of the voltage, while speed increases linearly.

Re:Counter productive maybe?

[Achromatic1978]

I'm not an expert, though I do know what I'm doing fairly well. I do a lot of transcoding movies (at my job! seriously.) I have a P4 2.8 HT, with 1.5GB RAM and 2 Radeon 9200s (I doubt this makes much difference, if any) desktop, and a Pentium M Dothan(?) 1.86 (2mb cache) and 2GB RAM, and a Radeon Mobility X700 laptop. The laptop is around 15-20% faster than the desktop at transcoding to MPEG 2.

Re:Counter productive maybe?

[Y0tsuya]

I think he meant that processor designers (and chip designers in general) design their chips to work over a certain range of operating conditions. In order to accomplish this they build in margins of safety and simulate their designs to death using expensive workstations or clusters. After production they get speed binned with elaborate tests using multi-million dollar machines.

Of course many users think they can do better with their over/under voltage/clock regimes, all the while claiming that it's a conspiracy by the processor vendors to deliberately throttle down their CPU. But all they're doing is wiping away the safety margins built in by the engineers. It's kind of like ricers cutting springs in their hondas. Never mind that Honda engineers spend thousands of man hours and million of dollars testing, simulating, and tuning their suspension on their expensive workstations and test tracks. I see their cars bobbing up and down like low riders over small imperfections on the highway and shake just my head.

In short, people who think they're smarter than engineers, usually aren't.

Re:Counter productive maybe?

[NitsujTPU]

Yes, that probably was inflammatory. I feel lousy about it too. Oh well.

Re:Counter productive maybe?

[NitsujTPU]

Congrats to GP BTW. Perhaps I'll underclock my notebook tonight.

Re:Counter productive maybe?

[NitsujTPU]

That was pretty much my sentiment, but I guess that I did sound kind of condemnatory or snooty in the observation thereof (perhaps).

Re:Counter productive maybe?

[DonGar]

However, this only shows that the CPU is now stable enough to run 8 hours without a noticable error. You might still end up with noticable CPU errors over longer periods of time. I'm accustomed to uptimes measured in months, which I won't get with even slight increases in the CPU error rate.

That doesn't mean you shouldn't try this, or that you shouldn't overclock (which is subject to similar problems), but you should be aware of the effects and be willing to accept them.

Manufactorurs don't push chips this way, in part, because they want to come as close to 100% stability as possible.

Re:Counter productive maybe?

[rbarreira]

I'm accustomed to uptimes measured in months

On laptops?

Re:Counter productive maybe?

[DonGar]

I run OS X on an Apple PowerBook. My uptime is (mostly) measured by how often they release major OS patches. But in truth, I'm exagerating, since I've never reached two months of uptime.

I have gone about a year without a system software crash.

Re:Counter productive maybe?

[rbarreira]

So you're telling me you have your laptop on all day? Or you use standby? The fact that the uptime is big doesn't mean anything in terms of the CPU heating if you're using standby lots of times...

Re:Counter productive maybe?

[DonGar]

It's on all day while I'm at work. It's often on all evening while I play Wow. It's put into standby at least twice a weekday while I carry it to/from work. If I never need a computer after I get home, it may spend the night in standby, but that's the exception, not the rule. It certainly never spends the weekend in standby.

I have multiple computers in multiple locations, but I always use the laptop as the interface to the others through a combination of SSH, VNC, and Synergy.

Re:Counter productive maybe?

[DonGar]

However, I should point out that it almost never does any heavy computation other than gaming. Most of the time it's running Camino (web browser), Mail, Aim, iTunes and Synergy.

Re:Counter productive maybe?

[yppiz]

In my case, it's both. I photograph in RAW mode and then convert those files to JPEG in large batches. So I've got disk IO, RAM IO, and CPU. With or without undervolting, this pegs my CPU.

In my earlier post, I forgot to mention the advantages of undervolting and underclocking. I now get about 20% more time on battery, and my machine runs cooler. The fan on my laptop is very loud, and so I go to great lenghts to keep it from spinning up.

--Pat

Re:Counter productive maybe?

[DarkHand]

Mod this up please, this is exactly right.

Re:Counter productive maybe?

[bigtrike]

I used a P3 Kill-A-Watt:
http://www.the-gadgeteer.com/review/kill_a_watt_el ectric_usage_monitor_review

Re:Counter productive maybe?

[Jeff Molby]

Why don't laptop manufacturers do this?

Probably because it takes time, which is money (even if automated somehow), makes it harder to price (just think of how many model numbers you'd need), and would eliminate the tolerances that allow them to offer decent warranties.

Re:Counter productive maybe?

[ClamIAm]

Still, benchmarking utilities only go so far. You may run into a real-life situation where under most circumstances your voltage level is fine, and the disk spins up or something that causes a sag in the power level to the CPU and hangs the system.

OK, this is probably a bad example. What I'm trying to say is that most benchmarking/stress-testing utilities don't usually push all the buttons in the same way that normal usage does.

Re:Counter productive maybe?

[VirionNW]

I lowered my voltage from 1.45 to 1.2 a few months ago and have yet to have issues, for a while I used it at 1.25 but eventually did the slow test down to 1.2 and haven't crashed from it yet, though going below gave me a one-way ticket to the power button. This is the same voltage as the minimum clock speed, so I find it kind of wierd, though with full throttle down and low voltage I never risk the fan whirring up like the full speed and the lower voltage setting, which causes both annoyance (audibly, the fan is aging poorly) and increased power consumption.

Re:Counter productive maybe?

[Frodo420024]

Hi

Surely if you drop the voltage your are going to have to under-clock the processor (reasoning that to over-clock you need to increase the voltage). Most processors for laptops already throttle the processor down when under light load now-a-days which must be a great energy saving. Would under volting it really then save more or would you just end up with a laptop that is dog slow?

What he does is to maintain the frequency (and the multiplier), and testing how low he can go on voltage. The minimum voltage reduced 7,5% is nice, the maximum reduced 20% is great.

I have just reproduced his results (my laptop is very similar), and can confirm that what he does causes no harm to performance - CPU hugs run at the same speed before. I'm sure if it was this easy one of the big laptop producers would already be doing it as a 20% increase for basically nothing would give them a fantastic advantage.

They have stability concerns to consider. Undervolting at the expense of stability is not at all worth it. We have the time to tweak our individual machines and get the most of it.

Underclocking is cool :)

Re:Counter productive maybe?

[kevin.fowler]

Why don't laptop manufacturers do this?

So Joe Schmo buys a horrifically overpriced backup battery for his laptop?

Re:Counter productive maybe?

[glazed]

I've got a P4 Northwood 2.8 Ghz running at 3.25 Ghz with good cooling, at 1.525 volts. Similar results with a couple of other processors I have at home. I just decided I was sick of the heat but not the speed and just started spending a bit of time seeing how low was possible. Took me for quite a surprise how often things remain stable at the lowest vCore supported by the motherboard. And I do go through the proper memtest, prime95...for hours and hours.

Re:Counter productive maybe?

[default luser]

Why don't laptop manufacturers do this?

Because Intel and AMD already do it for them, as best they can.

AMD offers the Turion 35w max TDP (1.45v load) and Turion 25w max TDP (1.2v load), and both chips scale back to VERY LOW voltages and speeds when idle.

Intel offers the Pentium M LV and ULV (Low Voltage and Ultra-Low Voltage) parts.

These parts are binned within a few voltage grades. This simplifies testing, because instead of dozens, perhaps hundreds of voltages to test, you only have a few.

With dynamic clock and voltage scaling, these processors automatically run at their lowest settings when you're doing something casual, like reading a webpage or document, or are watching a standard-defintion movie.

Sure, you can get a tiny bit more out of undervolting, but manufacturers leave that for enthusiasts because it is not worth it in their bottom line. As I mentioened before, testing parts with such fine granularity would only make all processors more expensive, and on the retail end, selling dozens of parts with different voltages would only make the marketplace more confusing.

Computer Performance

[ranton]

How does reducing the Voltage in this way effect performance? If performance drops, then you could have just bought a computer with less processing power that also had lower power needs in the first place.

If there are no performance problems, then why dont all laptop manufacturers already do this?

--

Re:Computer Performance

[drinkypoo]

Generally speaking, limiting processor power limits maximum clock rate. If you undervolt you generally underclock. Most mobile processors already have a power-saving scheme that allows you to select the highest speed that will be used while the system is on battery. Even older systems (like my stinkpad A21p with Mobile P3) have multiple speeds and they will run at a slower one automatically when on battery. So there's not much of a difference unless you're reducing voltage to something lower than the system does automatically.

Re:Computer Performance

[Anonymous Coward]

it dosent thats the point, cpus are different some can handel undervolting and be stabel some cant.
You se the same syndrome when overcloking some does fine without overvolting others need it to be stabel at higer clockrates.

Re:Computer Performance

[Anonymous Coward]

wow SWING and a miss on that one...

more advanced components make better use of the power, in particular this is true with CPU's

so current chips use less power and havebetter batteries, now add the lower power usage by limitingthe voltage and you have even less power usage.

Re:Computer Performance

[Ruff_ilb]

"How does reducing the Voltage in this way effect performance? If performance drops, then you could have just bought a computer with less processing power that also had lower power needs in the first place.

If there are no performance problems, then why dont all laptop manufacturers already do this?"

As was stated above: Two reasons. That is:

a) Mass production
All processors are different, and some inherently work better than others. The default voltage setting for a laptop should probably be a little over the highest requirement they see for a particularly ineffecient laptop. Therefore, if you have a more effecient laptop, then you can safely scale down your voltages.

b) Stability
Most laptops focus on being 100% stable, so they can overshoot the real MINIMUM required voltage in the name of stability. I'd guess the difference in voltage between a 100% stable laptop (Just kidding - a 99.9% stable laptop) and a 95% stable laptop can be quite large. If you're willing to sacrifice that ~5%, you can get a longer battery life.

As for performance losses - well, if you don't underclock it, it's possible to lose stability. But I think the point is you don't even have to underclock it, and you don't necessarily have to lose stability either.

Re:Computer Performance

[buraianto]

While I know you said "generally", I'd like to point out that I undervolted my desktop and did not underclock it. It's just the same as overclocking without overvolting.

Re:Computer Performance

[ranton]

Sorry, you are write and I should have been more clear. I didnt mean use older/slower chips, I meant using new/slower chips that are slower mostly so that they save power. Such as Via chips or the low voltage versions of AMD chips.

Re:Computer Performance

[Jozer99]

Voltage does not effect performance. CPUs need a minimum voltage to operate the transistors. Thisd minimum voltager varies from batch to batch, so the company chooses a "safe" voltage that will always work. If you are lucky, your processor will run at a lower voltage than the OEM voltages. However, it may not, and will crash. The fact that this guy got 19% voltage decrease is pure luck. If you bought an identical laptop, you may only be able to get 5% voltage decrease, or even at 35% decrease.

No Con's?

[randomErr]

I just scanned through the article and saw they never listed any con's. How much of a performance hit are you taking? Is there any long term damage on the processor or memory? Are you voiding your warranty?

Re:No Con's?

[Anonymous Coward]

Well if you look at your machine cross-eyed, it voids the warranty.

So what you do is, add water to the display, turn it on briefly, and claim that it came out of the box that way.

Theres lots of other fun ways to get around warranty requirements

Re:No Con's?

[obious]

No, no cons. Most processors, especially mobile variants, can operate above their "standard" specification. So even if the spec calls for 2GHz @ 1.5v the processor might be able to operate correctly at 2.3GHz @ 1.5v. Similarly, a processor with a spec of 2GHz @ 1.5v could operate at 2GHz @ 1.3v. Thus, no performance hit. This is because after the processors are manufactured they are tested and separated based on the highest performance they can reach under a set of standards set forth by the manufacture. This means that two processors coming off the same wafer could actually become an AMD 64 2800 and an AMD 64 3000 for example. Now, one of those 2800's could have been a 2950 but since AMD doesn't have a 2950 it was put in the 2800 bin. Think of it as a lowest common denominator that ensures even the shittiest processors run fine. ...and it only voids your warranty if you tell them.

Re:No Con's?

[freidog]

Since he kept the clock speeds (the multipliers) the same as the origianl settings, there would be no perfomrance hit at all. You could introduce your own P-States (faster or slower, anything that the multipliers allow for) if you found a need. For instance if you found watching DVDs would load the CPU enough to move out of the lowest power state, but not really stress the CPU in the high power state adding a mid power (say 6x multiplier) may be a happy middle ground in terms of performance and power consumption.

Lowering the voltage doesn't hasten long term damage, for the most part lower voltages and temperatures tend to increase the lifespan of a CPU.

I would assume it voids the warranty. But it's also something that could be undone by uninstalling the software and impossible to tell you did it.

Re:No Con's?

[ccool]

Actually, yes there is some Cons... By doing so, you could have stability problems, if you're unlucky... There is a reason why Intel don't do that right out of the box. Transistors need a minimum voltage to work correctly.

Re:No Con's?

[Anonymous Coward]

that's what people are talking about...

You feed in a voltage to the core and that determines the overall clock speed, but also at the same time you are running a voltage over the transistors. Now changing the voltage that determines clock speed will result in a higher/lower clockspeed. This not the voltage they are varying though. As several other people have mentioned, this type of underclocking is already done by many modern processors. Instead what this does is lowers that voltage running over the transistors. Thus the variance you see is the threshold at which various computers can still read the signal from the transistors. This threshold is determined by various machine tolerances and is a luck based thing similar to dead pixels in an lcd screen if you will :).

Now, the reason why you would run various programs like prime95 is that they stress your CPU fairly hard. Thus gives your computer many chances to fail, and thus gives you a realistic idea of how stable it would be for a much longer time period with a certain level of confidence.

hah I'm probably all wrong though as I'm really stoned and forgot what the question was...

isn't this what speed step did back with the PIII?

[ulysses38]

i'm not so sure about the heat, but it seems that it would follow.

"Mobile Intel® Pentium® III processors with Intel SpeedStep® technology let you customize high performance computing on your mobile PC. When the notebook computer is connected to the AC outlet, the new mobile PC runs the most complex business and Internet applications with speed virtually identical to a desktop system. When powered by a battery, the processor drops to a lower frequency (by changing the bus ratios) and voltage, conserving battery life while maintaining a high level of performance. Manual override lets you boost the frequency back to the high frequency when on battery, allowing you to customize performance.?

lol

[Anonymous Coward]

OR you can just buy a laptop that allows you to do this stuff natively.

I have an acer aspire 1691 laptop and i can control how fast i want the cpu to run ,how bright the panel is if wifi is on and stuff like that all through software.

Why would I undervolt it when my laptop can do it through software already.

Re:lol

[Da Zeg]

I have an acer aspire 5021 (ML-28) and the lowest voltage the amd drivers will go to is 1.0 and 1.5 respectivley with no steps between 100% and 50% cpu clock through CPUID i run at the lower speed at 0.925V and the highter at 1.325V with no hitches apart from everything runs cooler and longer. I also get the added bonus of being able to set speeds in between.

Re:lol

[Achromatic1978]

My ASUS laptop will run all the way down to 16% of clock in the "Max Battery Saving". It's impressive - gets ~4 hours, even though it's a 17" clearbright display, with 3d accelerator card.

Bad Idea

[AKAImBatman]

This sounds like a really BAD idea to me. Low Voltages can produce the exact opposite of the intended effect. Instead of lowering the power consumption, you'll get higher amperage spikes as the equipment draws more power to compensate. The result is that you could be damaging your electronics and not even know it.

I'll grant that modern manufacturing methods have greatly increased the survivability of hardware under less than ideal conditions. However, that shouldn't be taken to mean that you can't do serious hardware damage by operating outside of the device's specifications.

Re:Bad Idea

[JumboMessiah]

Not necessarily true in open loop systems. Amperage draw is heavily influenced by the voltage applied to the device. Though there are other factors involved (aka clock speed), a lower input voltage will almost always result in a lower amperage draw, which results in lower power consumption.

I've never seen a light bulb stay the same brightness when I reduced the voltage to it. It gets dim, draws less amperage, and less power.

Ohms Law regulates this, though in AC environments it isn't as cut and dry as V = I * R.

Re:Bad Idea

[serbanp]

Instead of lowering the power consumption, you'll get higher amperage spikes as the equipment draws more power to compensate.

Sorry, this is wrong in the context of a CPU power supply.

When you lower the core voltage, several things happen at once:

1) the power dissipation due to the clock switching is lowered with the square of the voltage reduction. i.e. a reduction from 1.3V down to 1.1V will reduce this power component by 40%

2) the power dissipation due to the junction leakage and off-state punchthrough decreases by the ratio of the voltage.

3) but the switching speed of the MOSFET transistors decreases. Effects 1 and 2 are good as they mean an overall lower power dissipation. For 90nm processes and up, effect #1 dominates. For 65nm and below, the effect #2 becomes increasingly larger.

The downside is #3. Lowering the voltage means that some critical paths inside the CPU logic could become longer than the clock period, generating timing violations and system crashes. The only remedy against this is under-clocking.

In the end, the one thing you can gain by under-volting is the margin between your particular CPU and the lousiest one in the same class that will still perform OK at the same clock speed. As each CPU is tested and binned especially for power dissipation AND maximum clock speed, this margin is low and the gains minimal. And you spend a lot of time to find out what is the lowest safe voltage.

If you want less power dissipation and longer battery life, under-voltage and under-clock. This is done automatically already in the mobile CPUs, both from Intel and from AMD.

Re:Bad Idea

[Anonymous Coward]

Boy do you have it ass-backwards. Given a resistance and discharged capacitance (of course in series), a higher voltage applied actually causes more instantaneous current to flow to the capacitor. A lower voltage causes less instantaneous current to flow to the capacitor.

Since all wires on chips are basically resistances and capacitances, I'll leave the conclusion about which causes greated current "spikes" as an exercise for the reader.

Now how about the transistors? Same deal. Lower voltage means lower short circuit current. Total power depends on the switching time.

Anyway, device specifications typically do have low voltage modes. There's a nice graph that can be drawn to compare voltage versus frequency that shows what combinations are possible.

This argument is again independent of transistors designed specifically for high voltage or low voltage. Just a generalization for wires and CMOS.

Re:Bad Idea

[serbanp]

Ooops, got carried away and did not really refute the dumb claims the OP made. Yes, the CPU is a monotonic load (i.e. when the voltage decreases, the current decreases).

The OP may have had in mind some constant-power type of load, where the current consumption is (indirectly) driven so that the output power stays the same. From the I-V perspective, the CPU is a glorified non-linear resistor.

Re:Bad Idea

[Rakishi]

I'd like to add that Intel undervolts their Pentium-M cpus to lower power usage already. It also lowers the speed itself of the cpu.

So given that my cpu on batteries probably switches at least once a minute and is not dead... yeah I think you get the point.

Re:Bad Idea

[AKAImBatman]

Hmm, I think I understand what you (and others) are saying. Basically, the power draw is controlled entirely by the power supply, thus there is no concern over the chip drawing too much amperage. Changing the voltage thus changes the overall power available, resulting in a lower rate at which the chip's capacitors charge. If the capacitors charge too slowly, it could potentially result in a cycle that is longer than the clock rate. This is bad because the state of the chip won't finish propogating before the next clock switch, potentially screwing up the internal logic. Close enough?

Thanks for clearing that up.

Re:Bad Idea

[Anonymous Coward]

Yes but it is not really the capacitors charging more slowly that can create the problems, it is the fact that the type of transistors used in the CPU have slower switching speeds at lower voltages. In my EE classes, we don't build circuits that would be effected by this drop, but at the speeds modern CPUs operate at, the decreased switching speed could result in some transistors switching so slowly, they miss a clock cycle, resulting in major problems in the CPU pipeline.

Re:Bad Idea

[Savantissimo]

You have to get a certain amount of charge on the gate before it reaches the requred voltage, and the gate has a certain capacitance, so lowering the supply voltage increases the time to switch, yes? The capacitance is parasitic, but I think the GP has it exactly right.

Re:Bad Idea

[F00F]

As each CPU is tested and binned especially for power dissipation AND maximum clock speed, this margin is low and the gains minimal. And you spend a lot of time to find out what is the lowest safe voltage.

It should probably be mentioned here that the "lowest safe voltage" (if there can be said to be such a thing) is temperature-dependent (a function of the effectiveness of a fan, the density of the air, the load on the regulators, the number of components powered up vs. down...), and can even be data-dependent.

In order for a CPU to work properly, an awful lot of "digital ballet" needs to be happening, billions of times per second, in perfect harmony. Failing a single setup or hold check on a clocked logic gate leads to erroneous bits entering your data and control streams. Quite a few logic blocks, implemented in quite a few different transistor-level CMOS techniques, have different timings depending on what data is flowing through them at the time -- i.e. whether certain bits are zeroes and others ones, or whether the bit transitions from a one to a zero at the same time as an adjacent wire is switching, etc.

What I'm saying, of course, is that if you "spend a lot of time to find out what is the lowest safe voltage", and then (oops) you plug in a USB peripheral that you weren't using before, or (oops) your room heats up a little, or (oops) you use a laptop cooler to get it off your lap a little (setup and hold times, as well as clock and data paths are voltage dependent y'know...) or (oops) you access an unusual data pattern in a program you haven't run in a while, or (oops) you use your laptop on an airplane, your happy little CPU (that you just had to eke out that last little bit of thermal margin from!) starts failing. Quietly, sometimes.

If you're lucky, Windows just crashes -- but manages not to trash your hard drive while doing so (remember, this is weeks or months after you ran your little three-day burn-in marathon "torture test"). But in a not-so-pleasant outcome, your data just starts quietly... rotting. Binary file formats start getting corrupted. Programs stop running. Checksums start failing. Spreadsheets start changing quietly behind your back. I grant you that on a Windows-based laptop this sort of thing could generally be considered the norm (due to virii and spyware and so on), so you might not really notice.

But fortunately for the retarded selves of the people "publishing" (I use the term loosely) this crap, there are thousands of engineers at Intel and AMD running detailed and comprehensive timing simulations of all of the tens of millions of transistors in those teeny-tiny CPUs. They can't possibly EDAC- and parity-protect every net in the design, but they do run static and dynamic timing checks, with probabalistically-developed parasitics and 3-dimensional noise parameters. They consider crosstalk and electrical noise, process variability, and electromigration. They run extensive tests on the CPUs at the wafer and package level to detect slight variations in current consumed during carefully-chosen JTAG test vector execution. They speed-bin and scrap parts that don't pass the multi-billion-dollar test regimens that they spent close to a decade perfecting, and use the detailed results to craft application notes and engineering design guidelines that they feed to their OEM partners, who in turn carefully couple the CPUs to circuit boards with traces perfectly matched to one another, as well as to the impedance characteristics of the solder they're using, and attach them to advanced power supplies and controllable oscillators.

And then you go ahead and download a shitty little ninety-nine cent program off teh unternet and procede to pick your own CPU voltage that "seems to work OK".

Yeah, good luck with that. Sounds like a plan to me. There once was this handy expression about fools and money -- I guess "data" is just the modern substitute.

Re:Bad Idea

[serbanp]

I, of course, agree with you.

However, your statement that ... OEM partners, who in turn carefully couple the CPUs to circuit boards with traces perfectly matched to one another, as well as to the impedance characteristics of the solder they're using, and attach them to advanced power supplies ... is a little off, especially in the case of a VR for notebook CPU.

You would not believe what motions the ODMs go through to shave off that extra MOSFet switch or output capacitor, all in the name of saving pennies! I can tell you from first-hand experience that most VR designs happening since a few years ago violate in some way or another the Specs Intel or AMD provide for their CPUs (larger ripple, a little too much overshoot in load transients, bad loadline etc). This is regardless of the brand name, as anyway most OEMs (some Japanese makers being the exception) delegate the design to the ODMs while in the same time placing awfully tough cost constraints on them.

The only slightly better built systems are the ones labeled "commercial" because the selling price can be higher and there's more room for decent design.

Desktops and much much better built and this is just because there's more space there, the competition is more heated and simply because there's more granular choice than with notebook products.

Been doing this for almost a year with CHC

[Anonymous Coward]

CHC = Centrino Hardware Control, now called Notebook Hardware Control.

CHC/NHC even has built-in stability testing.

It's fairly easy to run 400MHz FSB Dothan CPUs at 533MHz FSB on Sonoma (i915) or ATI Xpress200 laptops. I run a Pentium-M 715A (1.5GHz) at 2GHz with only 1.14v.

Re:Been doing this for almost a year with CHC

[evalf]

A big advantage Notebook Hardware Control (NHC) has over the RMClock utility is that it does not override the power scheme setting that are applied by the computer's manufacturer's software: for example, using NHC I can define the different voltages applied for every multiplier, but still use the IBM utility that came with my thinkpad to choose whether I want the lowest/highest/whatever multiplier, whereas the RMclock utility takes control over everything.

Another thing to take into account is that NHC has many more functions, such as temperature monitoring, GPU undervolting, HDD SMART monitoring, and so on...

I also read somewhere (do not remember where, and I did not bother comparing myself), that the CPU occupation was lower with NHC than it is with RMClock (even though it has more functionalities).

Re:Been doing this for almost a year with CHC

[timecop]

My only complaint about CHC is that its a .NET based app (for the GUI) which simply wraps some win32 DLLs/installs a driver for controlling ACPI and CPU clocking. That makes it rather heavy to start up, and while its sitting in the tray for days, it usually gets all swapped out, so if I need to change a setting, it takes a few seconds to bring the GUI back up because of all the .NET junk. If the author would just keep using C/C++ for the GUI it would be a great app.

As far as RMClock goes, I just tried it on a centrino 2.13Ghz laptop, and it dimmed my LCD to unusable levels without even asking, and I spent 5 minutes trying to find how to turn the fucking thing off and restore my LCD brightness.

laptops already have step by step instructions

[method77]

at least my Thinkpad does. The 'access IBM' button explains everything for you or right-clicking on the taskbar battery icon gives you choices of battery saving which does everything mentioned in the article. I am not advertising IBM or anything. Only pointing that out. I am sure other brands have similar functions too.

Re:laptops already have step by step instructions

[binaryspiral]

It's actually build into the motherboard, cpu, and OS.

The motherboard (more directly the PCU) recognizes it's not on AC anymore and switches to battery. It then alerts the OS that it is on DC.

The BIOS checks its settings and alerts the motherboard to go into power saving mode (if it's enabled).

The CPU then drops in speed. My Thinkpad also throttles the GPU, NIC, Wireless, FSB, backlight, and colordepth. Anything that isn't being used is shutoff - like wireless, nic, IR, etc.

The difference between full on and power saving gives me on average an extra hour and 20 minutes of battery life without a noticable performance degration in normal usage.

If I want to game - all bets are off.

Optimization... for human usability

[rsborg]

This writeup is great. I just bought a cheapo laptop (Turion ML-32) and other than increasing RAM to 1GB and getting a faster (and cooler/more efficient) 5400RPM laptop HD, I think that reducing power output is one of the most important things for usability.

Other than just battery life, a reduced heat profile will move the laptop from a desktop replacement to a more usable all-around better box. Will still pale in comparison to my wife's powerbook, but hey, this was half the price and I still can't use OSX at the office (damn intranet IE only webapps).

What's New?

[Anonymous Coward]

My six year old Dell Latitude has bios settings that automagically step down the cpu voltage when the laptop is unplugged. What's new about this?

Re:What's New?

[Slashcrap]

My six year old Dell Latitude has bios settings that automagically step down the cpu voltage when the laptop is unplugged.

But it doesn't decrease the voltage while maintaining the same clock speed.

What's new about this?

See above.

Parent is a Bad Idea

[Anonymous Coward]

Whee... I mean, CMOS logic will never "try to compensate". There is no feedback. In a typical digital system, only switching power supplies will draw more current when their *input* voltage drops. However, Vcore is the *output* voltage of those, not input.

Re:Parent is a Bad Idea

[Anonymous Coward]

when's the last time you took and ee course, the 70's?

Re:Parent is a Bad Idea

[StikyPad]

TTL is not a current-intensive design. There is some amount of current, but the only reason your CPU draws a significant amount of power (current x voltage) is because there are millions of transistors. The transistors will not draw more current to compensate unless there's a current feedback loop (there isn't, otherwise you wouldn't need, or even be able to have, external voltage controls); they'll simply cease to function properly. There is a feedback loop for the power supply on the motherboard (just a regulator really), but that's what we're manipulating, so overcurrent issues shouldn't exist. More importantly, the regulators on the motherboard should prevent overcurrent conditions.

The designs are trade secrets, so we'll probably never know for sure, but it doesn't even make sense to put any sort of voltage/current regulation on the chip itself since a) real-estate is at a premium b) regulators need to be relatively large, since they handle ALL the current for the CPU and c) they generate a LOT of heat. I haven't looked at laptop motherboards, but on a desktop motherboard, you'll see usually 3-6 transistors mounted vertically screwed to heat sinks near where the power supply connects. Touch them if you want. The heat sinks should be grounded, so the only thing you'll probably hurt is your fingers.

Re:Parent is a Bad Idea

[techfury90]

But most CPUs since the late 1980s are CMOS.

Re:Parent is a Bad Idea

[StikyPad]

Sorry, I meant to say CMOS, but I was thinking about TTL at the same time.

Re:Parent is a Bad Idea

[imgod2u]

Have you ever taken an EE course? The impedence of a digital device does not "compensate". Whatever load your transistors exhibit depending on the input voltage is whatever load it exhibits and it usually varies very little (which is why they're transistors to begin with) when voltage changes *except* at the threshold when it goes from short-like behavior to open-like behavior. The ideal case, of course, is that it behaves like a short when the input is above the threshold and open when it's below. Not being ideal, silicon manufacturers would at least try to get close to this behavior. That would mean that, for a given set of transistor states (as well as transistor switching sequences), the impedence through the chip does not really change much with the core voltage (or should not). You'd have to look at the specific I-V curve under different operating conditions and extrapolate the impedence values but they should not change much. Temperature probably causes a large change in impedence than input voltage. That being said, I'm not sure how good an idea undervolting a chip can be. You certainly won't damage anything but digital systems are made with a threshold voltage in mind. For silicon, this should be around 0.7V. Usually, the core voltage should be somewhat higher than the rail to rail signal voltages so lowering your core voltage could cause quite a bit of instability as your transistors would not be able to switch as fast or at all even. I suppose you have a bit of wiggle room as the manufacturer almost always leaves headroom, but that headroom is there for a reason.

Re:Parent is a Bad Idea

[Savantissimo]

While the impedence does not change much with voltage within the operating range of the device, the voltage directly affects how fast the chip can be clocked and the amount of heat it produces, both of which substantially affect the impedence. The rest of your post is a bit confusing. There is no such thing as a digital transistor, or digital electronics for that matter. There is no such thing as a "threshold voltage" - there are three voltage ranges in a given semiconductor process intended for digitally-encoded signals- low, high and indeterminate. The overall voltage transfer function of a "digital" device, while seldom linear for over much of a range, is smooth and monotonic and can be (mis)used for non-digitally encoded analog signals.

I'm not sure which is more surprising

[Marxist Hacker 42]

That the CPU can run at a lower voltage- or that voltage of the CPU on a modern motherboard is SOFTWARE Selectable.

Re:I'm not sure which is more surprising

[evilviper]

That the CPU can run at a lower voltage-

Not surprising at all. It's a mainstay of the processor industry. Since the start of the PC, processor voltage has been decreasing every generation. Generally, the only difference between AMD's "desktop" CPU and their "mobile" CPUs is the stringency of the testing. A mobile CPU that can't handle extreme underclocking/overclocking gets labled as a desktop chip.

or that voltage of the CPU on a modern motherboard is SOFTWARE Selectable.

Everything you can set in the BIOS is software selectable. I can understand the concern, but all CPUs have built-in safety circuitry that should allow them to automatically shut-down and prevent damage.

Re:I'm not sure which is more surprising

[Marxist Hacker 42]

I guess I'm getting old- in my day, the 128k Rom Bios was for things like hard drive setups, memory maps, and clock information.

Re:I'm not sure which is more surprising

[kesuki]

yeah nowadays they have up to 4 MB flashable BIOS that run linux http://www.linuxbios.org/index.php/Main_Page.

"when i was a kid, to set the CPU clock we had to manually insert jumpers, And we Loved it!"

18% -- that's really funny

[Jason1729]

How do they come off saying a reduction from 78 to 64 degrees F is an 18% reduction in temperature? The Fehrenheit scale is arbitrary and does not have a meaningful zero point.

In celsius, their reduction is 26 to 18 degrees, a reduction of 31%

Why not define a new scale with the same degrees but 0 degrees (new scales) = 63 degrees F. Now on the new scale they've reduced the temperature from 15 to 1 degree, a reduction of 94%....wow that's way better than their lousy 18%.

Their number is totally meaningless.

Also, "undervolting" is not a word.

Re:18% -- that's really funny

[Anonymous Coward]

No, what they meant was that the temperature now is -25 F. This laptop can also work as a refrigerator. Don't ask the physics behind it, because it doesn't apply to computers. We are now in the information age, right? Obviously they are just hacking the Matrix control computers.

Re:18% -- that's really funny

[Da Zeg]

We already have the kelvin scale where 0 is absolute zero - where all particles in a mass would have no energy. 0 K = -273 C All calculations involving heat and energy should be done using this scale or become invalid And I really doubt that the kelvin scale was used to give the figures quoted in the first post. They seem a bit high and arbitrary.

Re:18% -- that's really funny

[mattcoz]

You make the same mistake by using Celsius, it's zero point is only meaningful for water. The only truly meaningful zero point is absolute zero. So if you convert the temperatures to Kelvin you'll see that it was a 2.6% decrease.

Re:18% -- that's really funny

[Forbman]

The Fehrenheit scale is arbitrary and does not have a meaningful zero point.

Umm...*every* temperature scale (just like linear measurements) are arbitrary.

It just so happens that Kelvin coincides nicely with an unobtainable measuring point. You could just as easily factor a thermal scale that rises at 0.857347235 the "rate" of the Kelvin scale. Say, in the Stupid thermal scale, boiling water is at 5.2 deg Stupid, instead of 212 deg F, 100 deg Cel, or 373 deg Kelvin.

The *advantages* for one system or another are not always so apparant. Is it better to calculate trig-related math in degrees, gradients or radians? For trivial things, radians is a whole heck of a lot easier than degrees. Is it better to do geometrical maths in cartesian coordinates, or use some other coordinate system?

In the end, one system vs another explain the same thing correctly and to the same degree of "accuracy".

Re:18% -- that's really funny

[Formica]

The Celsius zero is just as arbitrary as the Fahrenheit zero. The only true "zero" is absolute zero, at -273C or -459F. Using either scale, the "percentage reduction" is around 2.7%, for what it's worth. It shouldn't matter what scale you use when talking about percentages, assuming you use the true zero. If an object becomes 10% lighter, it doesn't matter whether you use pounds or kilograms, does it? Of course, you use percentages even if it doesn't make sense. (78-64)/78 is around "18%", but isn't a very meaningful number. Switching to Celsius doesn't help here, but Kelvin (or Rankine for those Fahrenheit fans) does.

Re:18% -- that's really funny

[Anonymous Coward]

Why not define a new scale with the same degrees but 0 degrees (new scales) = 63 degrees F

Why not just throw out that pesky third "law" of thermodynamics [I never liked it, anyway], and define the zero point as 64 F? Wouldn't that give you a 100% reduction, using your calculations?

Re:18% -- that's really funny

[shawn(at)fsu]

I didn't know that temperature needed a "meaningful zero point" to be useful? How do you determine a "meaningful zero point" anyway. As long as long as you state the scale used what's the problem?

Re:18% -- that's really funny

[Mr. Underbridge]

It just so happens that Kelvin coincides nicely with an unobtainable measuring point.

No, it's actually much more fundamental than that. Essentially, it's the point at which a given substance has zero internal energy - which is to say, a temperature of 0. It's not a coincidence - absolute zero is the point at which a given substance literally has no temperature as we know it. I can't think of a more appropriate zero point for a scale.

Re:18% -- that's really funny

[boldtbanan]

My 0 degrees as 64 degrees, so my temperature reduction is (14/0)% = (infinity)%. I win. :)

Re:18% -- that's really funny

[cosmicdog]

Their number is not 'totally' meaningless. It's just not totally meaningful. I'm an American and Fehrenheit makes sense to me. If the article was addressed to a group of Physics students, I'm sure that they would have used more absolute, concrete numbers. Now, Kelvin would not be very useful for the average folk because of it's low 0 point. Temperature changes that humans can experience and tolerate would barely register as a change on that scale. That's just my opinion. As far as the term 'undervolting', it is a word. People coin words and phrases all the time. Underclocking is not in Webster's Dictionary, but is has meaning. The same thing with Undervolting. It has meaning and is distinct from the meaning of any other word in the language. It's a word, they made it up. (or someone did).

Re:18% -- that's really funny

[jandrese]

Wouldn't it make more sense to compare the percentage drop against room temperature. As in:

Pre: 24 degrees Fahrenheit over room temp
Post: 12 degress Fahrenheit over room temp, a 50% savings!

Obviously no amount of undervolting would ever get the processor to absolute zero, it's going to bottom out at room temperature (when reduced to 0 volts).

Re:18% -- that's really funny

[santiago]

Say, in the Stupid thermal scale, boiling water is at 5.2 deg Stupid, instead of 212 deg F, 100 deg Cel, or 373 deg Kelvin.

Just to nitpick, there's no such thing as "degrees Kelvin". It's just "Kelvins". e.g. "Water boils at 373 Kelvins."

Re:18% -- that's really funny

[Anonymous Coward]

At least do your math right. It'd be

((14 - 0)/14) * 100% = 100%

Re:18% -- that's really funny

[santiago]

The correct zero point to determine the reduction in heat output would be room temperature. Compare the difference between room temperature and the high-volt processor with the difference between room temperature and the low volt-processor. After all, if it were outputting no heat at all, it would be sitting at room temperature, a 100% reduction in heat output from its initial running state.

Re:18% -- that's really funny

[Politburo]

All calculations involving heat and energy should be done using this scale or become invalid

Please don't forget Rankine.

Re:18% -- that's really funny

[Glonoinha]

I agree with you, but here's one thought : take the temperature at which your machine quits working (overheats) as a baseline - if they have increased the distance from that temperature (downward) by a certain thermal range - that delta would be the same measured in any current normal scale (Celcius, Farenheit, Kelvin.)

I'm guessing, but something tells me that 18% isn't going to be that delta.

Re:18% -- that's really funny

[Anonymous Coward]

You do not win. You should have failed 1st year calculus. Division by zero does not equal infinity.

Re:18% -- that's really funny

[Martin Blank]

If you want to be pedantic, the percent reduction in temperature was:

25.6C = 298.8 K
17.8C = 291.0 K

(291.0 K / 298.8 K) = 0.974

That makes it a drop of about 2.6%, not 14% or 31%. Doesn't sound nearly as impressive, but it is more accurate.

Re:18% -- that's really funny

[HermanAB]

Well, I suppose undervolting is revolting...

Re:18% -- that's really funny

[teslar]

Now, Kelvin would not be very useful for the average folk because of it's low 0 point. Temperature changes that humans can experience and tolerate would barely register as a change on that scale. That's just my opinion.

Wrong.
Kelvin is exactly the same scale as degrees Celsius. The difference between the two is simply that the zero point offset by 273.15 units. So any change would register exactly the same on scales in either degrees Celsius or Kelvin, the numbers would just be higher in Kelvin.

Perhaps you were referring to the fact that relative increases may appear smaller. But as the parent already pointed out, looking at % increases in this way is completely meaningless.

If it is 0 degrees Celsius today, and tomorrow will be twice as cold, what will the temperature be tomorrow?

Re:18% -- that's really funny

[tedgyz]

Mod parent overrated.

That's not insightful. Their knowledge of math is sketchy at best.

Re:18% -- that's really funny

[Ed Avis]

You can use temperatures without a zero point, and compare them, and talk about differences between two temperatures. 10 degrees is colder than 15 degrees, and the difference between 10 and 15 degrees is the same as the difference between 15 and 20 degrees. What you can't do is talk about 'twice as hot' or 'twice as cold' or anything involving multiplication.

It's a common mistake though. I spotted a weather forecast saying that one city would be 20 degrees Celsius today, 'twice as hot' as another city where the temperature was only 10 degrees. But what if one city was +5 degrees and the other -5 degrees? Is it still twice as hot - or minus one times as hot? Is 10 degrees infintely hotter than 0 degrees?

You can only talk about 'twice' or 'half' once you have a meaningful zero of temperature, as in the Kelvin scale.

Re:18% -- that's really funny

[Kafir]

A temperature system with an arbitrary zero point can be useful, sure--I think in Fahrenheit, like most Americans--but talking about percentages to compare temperatures in such a system is meaningless. If the temperature changes from 50F to 75F, is that a 50% increase? Convert the same temperatures to Celsius and the numbers are now 10C and 24C--does that make the same change in temperature a 140% increase? In the absence of a universally accepted zero point, ratios and percentages just aren't a useful way of comparing temperatures.

Re:18% -- that's really funny

[Delph1]

I've updated with Fahrenheit values and made it a bit more obvious that the other values were in Celsius. Hope this clears things up a little.

Re:18% -- that's really funny

[Anonymous Coward]

Sausages! :)

Re:18% -- that's really funny

[Noehre]

Tell that to Mr. Peltier!

Re:18% -- that's really funny

[homesandgardens]

Actually, the difference is that Celcius is a linear scale and Fahrenheit isn't. 64 F is not half the actual temperature of 128 F. With pounds and kilos you can divide or multiply either by one number to get the other, not so with F and C.

Re:18% -- that's really funny

[HardCase]

If you want to be pedantic...

I don't want to be pedantic, but if I did, I'd consider the change in temperature as the important measurement. After all, there's not much chance that the computer will be working at anything near 0K. So, consider the "zero point" to be room temperature, or about 295K.

So, at normal voltage, the peak CPU temperature changed by 56K. With reduced voltage, the peak CPU temperature changed by 42K.

1 - (42K / 56K) = .25, a 25% difference in temperature change. And that's the important number.

-h-

Re:18% -- that's really funny

[MushMouth]

What gets me here is how many people can't figure out that what is being measured is the difference of change of underpowered against the normal (standard voltage). Which of course it is very close but slightly less than the difference in power consumption, all of the power stored in the battery is eventually converted into heat (thermodynamic energy) the slightly smaller efficiency is due imperfect measurement and containment. It doesn't matter what the scale or starting point is as long as the scale is linear.

Re:18% -- that's really funny

[Anonymous Coward]

Celsius zero is arbitrary? It's the freezing point of water.

Re:18% -- that's really funny

[Anonymous Coward]

The freezing point of pure water at 1013 hPa. It's arbitrary. Why water ? What has the freezing point of water to do with the temperature of a computer ?

Re:18% -- that's really funny

[multipartmixed]

Either minus 68 or plus 22.

Assuming, of course, that the question is being asked by a meterologist from Environment Canada.

Re:18% -- that's really funny

[h4x0r-3l337]

What you describe is not what determines whether something is linear or not. You can convert Celcius to Fahrenheit by multiplying by 9/5 and then adding 32. The fact that there's an offset does not mean that one is linear and the other is not. Either both are linear, or neither one is.

Re:18% -- that's really funny

[Odin's Raven]

Also, "undervolting" is not a word.

I'm sorry, but "undervolting" is a perfectly cromulent word. I'd agree that percentage was a poor choice for expressing temperature differences, but the primary point remains - undervolting can embiggen battery life.

Re:18% -- that's really funny

[cosmicdog]

Of course you are correct. I am referring to relative changes being expressed in a meaningful way. Telling me that my laptop will be 18% cooler, using Farenheit, let's me know that it will be noticeably cooler. If you tell me that it will be 1% cooler, using Kelvin, I will think "What's the point?" I didn't major in a Physical Science, so even telling me that the temperature will change from 299.15 Kelvin to 288.15 Kelvin really doesn't convey the intended message. I have to go searching the net, find a Physics book, or ask a Geek to find out what the heck you are trying to say. The sad fact is I would even have that problem with Celsius. I'm not stupid, I just don't usually need that knowledge to function. Call it ignorance, but I just don't have time to retrain my brain to think that way. So if this was a professional research paper, being pitched to engineers or other professionals, I agree, it needs more precise math. Since it is actually being marketed to the average user (even if, in the case, they are geeky enough to read SlashDot), those numbers are fine. The message is "Try this, it will make your laptop cooler and last longer." Message received. And of course, I don't have an answer to the 0 degrees Celsius question. Twice as cold is relative. What's cold to you and what's cold to me is different. If you are saying that day 2 will be half the temperature of day 1, I understand the question a little better. However, it is still relative. If we are talking about temeratures on Earth, we have a different standard than if we are talking about temperatures on the moon, the sun, Jupiter, or deep space. Since Earth temperatures never come close to approaching absolute zero, we shouldn't use that as a reference for judging half or twice as much as another figure. Right? Anyway, you are right about the numbers, the are arbitrary from a certain perspective (as are most things). However, in my case, at least, the convey, what I assume to be, the intended message.

Re:18% -- that's really funny

[ian_mackereth]

Verbing weirds nouns.

Re:18% -- that's really funny

[shawn(at)fsu]

Thank you.

Re:18% -- that's really funny

[Savantissimo]

To expand on your point, there are four types of measurement scale, each of which only supports the arithmetic operations of its own type and those of the more restricted types. They are, from most restricted to most general, with permissible operations:

Nominal Measurement Scale [e.g. phone numbers]- Counting
Ordinal Measurement Scale [e.g. serial numbers, class rank]- Greater than or less than operations.
Interval Measurement Scale [e.g.Fahrenheit, Celsius - scale divisions of equal quantity]- Addition and subtraction of scale values.
Ratio Measurement Scale [e.g. length, mass, Kelvin - interval scale with a unique, meaningful zero point]- Multiplication and division of scale values.

See http://web.uccs.edu/lbecker/SPSS/scalemeas.htm for more on scales of measurement.

Re:18% -- that's really funny

[Jason1729]

I'm not making the same mistake by using Celsius, I was saying both numbers are meaningless for the same reason, and giving my arbitrary scale as a further example.

Re:18% -- that's really funny

[Jason1729]

In what way is that math sketchy?

Just because it goes way over your head doesn't make it sketchy.

Re:18% -- that's really funny

[Anonymous Coward]

It doesn't matter what the scale or starting point is as long as the scale is linear.

But why use a percentage?

I.e., instead of saying 'method 1 had a y% reduction in CPU temperature', you can just leave it at 'method 1 resulted in CPU temperature reduction of x degrees Celsius.'

The latter alternative is better in every way.

Re:18% -- that's really funny

[tedgyz]

Just read some of the modded up responses below parent. The math is not over my head, it just doesn't make sense. FWIW, I have a BSMA.

Re:18% -- that's really funny

[Anonymous Coward]

Verbing weirds language.

If you're going to quote Calvin & Hobbes, get it right. Or you might be attacked by a tiger when you least expect it.

Re:18% -- that's really funny

[juancnuno]

78 - 64 = 14

14 / 78 = 0.179 = 18%

What's the problem?

Re:18% -- that's really funny

[Ed Avis]

On an interval scale like Celsius you can't add values. You can't add together 10 degrees and 20 degrees to get a meaningful result. However you can subtract one value from another to get a difference, and add that difference to another scale value.

There are some scales where you can arguably define less-than and greater-than operations but without being able to subtract to find differences. For example if you reckoned temperature just with 'warmer than' and 'colder than'. You might be able to say for certain that today is colder than yesterday, but not to find the difference between the two in a meaningful way and then add that difference to some third temperature. Also, you couldn't count in this scale because you can't measure equality; on some days I could say for sure that it's hotter than it was yesterday, but I could never say that it's *exactly* the same temperature. So for continuous quantities you can have less-than without counting.

Re:18% -- that's really funny

[Jason1729]

BSMA is not a real math degree it's a science degree ;). FWIW I have a B.Math from U of Waterloo.

Except for one mathematically ignorant post saying he's used to Fahrenheit so the percentages make sense to him and your reply which doesn't actually say anything, every other reply agrees with me.

Re:18% -- that's really funny

[tedgyz]

Well I guess you must be right then.

Re:18% -- that's really funny

[hvymtlsteve]

Why not use joules, which is a unit of energy? Hmm, I guess you still can't get a good percentage of decrease out of that, as it would be an amount of energy per celcius degree, and you'd still have to compare it to some arbitrary reference temperature... perhaps room temperature or zero on whatever scale you feel like...

What for?

[Poromenos1]

I don't think many people have a use for this. The processors shut down when they're not doing intensive work, and when they are (playing games, encoding) you more than likely have them plugged in an outlet. I don't know about heat, as I've never had a problem (I have an Athlon XP mobile).

Re:What for?

[Anonymous Coward]

My last 2 laptops (both HPs with ATI video cards- one w/ an Athlon XP, one with a P4) have had serious heat issues. If I leave my keys sitting next to the thing for more then a few minutes, they're nearly too hot to pick up. I guess it could just be me, or just be HPs, since that's what all my experience is with.

Re:What for?

[AngelofDeath-02]

He's talking about the videocard in this particular case.

But anyways, you do bring up an interesting point. usual power saving features do things like lower the clock rate when not in use, but lowering the maximum clockrate you would lower the speed of the computer, thus the max power it puts out. Knowing that you will be running the processor at max speed longer, you may or may not gain power/heat savings overall for long complex tasks, but I imagine for simple tasks you would.

now - I don't believe you need to downclock just because you lowered the voltage, but you may introduce system instability ...

That's usually what happens when you overclock without increasing the voltage as well.

Re:What for?

[Lehk228]

next ime you are looking to buy a laptop try an Acer TravelMate , not the Aspire line Aspires have horrible battery life my travelmate gets at least 3.5hours under average load with wifi enabled, and over 5 hours with wifi off

Re:What for?

[Anonymous Coward]

I have a pentium M 2ghz (model 770 i think) I can run the stock speeds at usually .2 or so less volts than default. At the lowest speed (800mhz) I can run .7 volts and like 1.1v at the highest speed (2ghz). I think the default is around 1.5 - I cant remember exactly as I did this a while ago now. I have been running like this for several months now and have had no stability issues. i use centrino hardware control to do it - it allows you to set a specific coltage for each speed the prcessor steps to in speed step mode. I have gained at least 30 mins to an hour of battery life depending on what I am doing. It really doesnt slow the system down at all and so long as it is still stable then you are good. I would suggest to anyone trying this to try with a pretty low voltage and work their way up until it is stable. THere really isnt any risk to damaging the components here, but be prepared for it to crash if you do run the voltage too low. At the lower speeds (like 800mhz to 1ghz - hah 1ghz 'low' now adays...) you should be able to run the lowest possible voltage it allows.

I also have my video card run very low clock speed (like 75mhz) in 2d mode - vs 380 in 3d mode. (Gefore 6800 GO btw). I wish you could have the ram change dynamically too - but right now the way it sets is the ram speed does not change with the 2/d3d setting, if you change it in one it changes in the other. I run the ram at 780 mhz.

Re:What for?

[Lehk228]

to communicate a wifi card transmits and recieves. yes they do use a lot of power.

Why?

[Eightyford]

Why not just underclock the processor? Adding more ram, dimming the screen, and using a virtual cd drive should also help considerbly.

I modded my Dual-Heel Processor

[digitaldc]

Laptops often comes with two Achilles heels, heat and limited battery time.

You know, I just found about this and I have modded my Laptop to the EXTREME!
I just went on a website and then tinkered with my new Dual-Heel Processor.
It's so EXTREME the battery catches fire 10 seconds after it finishes booting up.

It's 18% cooler only in Celsius

[RNLockwood]

If you use the American temperature scale (degrees Farenheit) you get a smaller temperature drop: 15%. If you measure in Kelvins you only get 4%.

For an excercise devise a scale that will give a 50% temperature drop, it's a lot more impressive and means about as much as measured in Celsius!

Re:It's 18% cooler only in Celsius

[pierreTheBear]

Ok, so my iMac G5 is running at about 50 deg C. That's 323 degrees Kelvin (ie, total thermal energy above absolute zero).

An 18% reduction in absolute temperature would reduce my processor to 264 deg K... that's equal to -9.15 degrees Celsius.

My kitchen freezer can't even get that cold! If I undervolted my iMac, I could be chillin' my b33r right now as well!

Re:It's 18% cooler only in Celsius

[Anonymous Coward]

You really should only measure the difference in *ambient* temperature vs the temperature of the CPU. e.g. If ambient temp is 20C and CPU before undervolting is 70C, then the difference is 50C. If after undervolting the CPU is 60C, a difference of 40C then the difference is 20%

((70-20)-(60-20))/50=20%

Re:It's 18% cooler only in Celsius

[Anonymous Coward]

You need a new freezer then, a decent freezer should be something like -18 deg C

Too much misinformation here.

[TheGuano]

Some facts:

Undervolting is NOT underclocking. You run the same clock speed, you just provide the CPU with less juice.

You do NOT need to underclock to undervolt, though if you're trying to hit a super-low voltage, a lower clockspeed will let you do it.

It can be perfectly safe. If you undervolt, and successfully run a Prime95 torture test for 24 hours, you're pretty much set. I'm currently running a 1.8Ghz Dothan Thinkpad at 1.134V (default at 1.8 is 1.340), and 0.700v at 600Mhz (default is 0.980 volts). That's on par or lower than those 1.0Ghz ULV Pentium-Ms!

Re:Too much misinformation here.

[TheGuano]

Oh, one more thing: undervolting is generally SAFER than overclocking, or overvolting to overclock. Providing less power to the CPU can cause errors or crashes, but it won't fry your CPU like overclocking/overvolting will!

Re:Too much misinformation here.

[Squash]

The empty can rattles the most.. The people with the least understanding of this topic seem to be the ones with the most to say. 95% of them don't even understand the title. Maybe its some crazy brain-fuck where "undervolting" looks like "underclocking".

I see 3 posters, including the parent, that understand the story. For the rest of you:

#1. Had you read the damn article, you wouldn't make yourselves look like such fools.
#2. Had you read the title, you wouldn't make yourselves look like such fools.
#3. Had you read the Slashdot summary, you wouldn't make yourselves look like such fools.

Unless, of course, you are a retard, in which case you should go back to humping something inanimate.

Underclocking is what you do to make your CPU run slower. Sometimes this means it runs cooler. SpeedStep, PowerNow! LongRun, etc are all forms of dynamic underclocking. Your windows control panel, BIOS power saving settings, etc are all forms of dynamic underclocking. This is NOT what this story is about. If you mentioned any of these things in your post, someone really should mod you OFF TOPIC.

Undervolting is what you do to make your CPU use less power at a given speed. You can usually get a good percentage drop before you hit instability, as long as you aren't overlocking. This will cause your system to run cooler, and your battery (on a laptop) to last longer,l without giving up performance! /me prepares to be modded down or pointing out the obvious

Lowering the voltage does not affect performance

[sammydee]

Lowering the voltage REDUCES current flow through the chip, reducing power consumption and heat output. The downside is, you can only lower the voltage to a certain limit before it goes below the threshold switching value for the transistors and the processor stops working. This causes no permanent damage, and is totally reversible by raising the voltage again. The lower the clock speed, the lower the voltage can be pushed. It is common practice among overclockers to try and push the voltage as low as possible for a given clock speed to reduce heat output.

I do this too.

[Anonymous Coward]

I find that if you disconnect the battery entirely, you end with 0 voltage draw on the battery. 0 amps are drawn, too. You can then go for many days without having to recharge the battery! This greatly increases overall battery life as well because of less wear and tear. With my Windows desktop environment being riddled with spyware and viruses, my productivity is only reduced slightly when I do this.

Re:I do this too.

[pscottdv]

And you get a whopping 68% lower temperature!

Nothing new for now...

[Razlor]

This procedure was described some months ago here, but without obnoxious "i spread my article over infinite pages in order to get more clicks" practice. I have been undervolting my Dothan a long time, using this little patch and some modifications to vidc. This keeps the fan off most of the time, saves some battery life and has no other impact whatsoever.

Undervolting?

[Anonymous Coward]

I'd comment on that malapropism, but I'm in an underwording mood today.

Undervolting is not underclocking.

[Destoo]

It seems a lot of people just assume that undervolting would be something akin to getting the inverse result of overclocking.

Here's the link to an interesting page about undervolting pentium M processors.

Experience shows that the processor may continue working correctly at lower-than-nominal voltages and frequencies, thereby reducing power consumption, heat and fan noise.

Even if your system seems stable, it may still suffer transient faults leading to arbitrary data corruption. In addition, errors in following these instructions (or changes between processor models) may operate the CPU above its nominal parameters, with effects up to and including laptop meltdown.

There's also a thourough discussion and user results from undervoltage on this thread.

Re:Undervolting is not underclocking.

[rjstanford]

Even if your system seems stable, it may still suffer transient faults leading to arbitrary data corruption. In addition, errors in following these instructions (or changes between processor models) may operate the CPU above its nominal parameters, with effects up to and including laptop meltdown.

Other than that, Mrs. Lincoln, how was the play?

Re:Undervolting is not underclocking.

[runderwo]

You should get a MCE if the processor flakes out on you.

There is no performance difference

[Anonymous Coward]

Undervolting a processor without changing the clock does not affect performance. With a processor, the clock synchronizes the electric pulses which maintains a constant instructions-per-cycle rate. As long as the voltage is high enough to create adequate digital voltage differences, the processor will function properly. You're basically using a letter opener instead of a kitchen knife to open a sealed envelope. Both approaches get the job done, but one's more efficient than the other. And if asked to do so, you could open the same number of letters per hour with either tool.

Also, for the Gentoo users: HOWTO Undervolt a Pentium M CPU.

mnemonic_

been doing this for years

[diitante]

intel w/ speedstep and "powernowd" (in linux) support. my 1.6Ghz PM runs at 600-800Mhz 90% of the time. significantly lower heat

Undervolting is NOT the solution

[Anonymous Coward]

This sounds good but because of Ohm's Law where the volts(EMF) times amps(current) equals watts the device will demand so many watts to operate and if device doesn't have enough voltage it takes it from current. This increase in current will increase temperature since the circuits are designed for so much current and increasing the current will just produce heat. Also you can screw up the voltage reference for the digital signal since the each 1 needs so many volts for the circuit to know it is really 1 and 0 is really 0.
Most circuits in laptops are desgined to their limits of design and manufacturing production so I personally won't screw this up. Desktop or workstations have an little more leeway on design so I would play this theory on those first.

Re:Undervolting is NOT the solution

[Anonymous Coward]

Ohms Law doesn't work like that..

If you lower the voltage, the components will simply stop working as you go below their operational rate.

Put it another way.. as you approach zero volts, does current approach infinity? The answer is no.

Re:Undervolting is NOT the solution

[onkelonkel]

Say What?? Either you are trolling or you fundamentally misunderstand Ohm's law.

Resistive loads (which, to a first approximation, a CPU is), don't "demand...Watts", they "draw current". The load resistance doesn't change, so Ohm's Law I=V/R says that if you drop the voltage, the current decreases. Drop the voltage 5%, you DECREASE current 5%. Your total power (V * I) is now decreased by 10%.

As for "screw up the reference voltage", this is and remains system ground, or 0 V. Yes, at some lowered voltage, the CPU will cease to operate. Assuming the CPU still runs, a logic level of 2.85 V is just as good as 3.0 V.

Re:Undervolting is NOT the solution

[Anonymous Coward]

Your logic is fundamentally flawed because semiconductors are not purely resistive loads. If you have ever taken a college course in semiconductors or digital circuit design, you should know that ICs are very non-linear beasts that have very many capacitances and inductances in their inner workings along with the traditional resistive elements. there very well may be different load resistances at different voltages. think of something as simple as a diode; forward bias produces a very small resistance, but reverse bias produces a much higher resistance.

Re:Undervolting is NOT the solution

[Anonymous Coward]

None of which has anything to do with what he said about Ohm's law. His logic is fine. You are just be argumentative like a typical /. flunky.

dual boot?

[DarkClown]

So looking at this I can't tell what happens in a dual boot situation - it describes setting up 'autostart' - does this happen at the os level?
What i'm wondering is, since this things looks like a windows utility, will the changes stay in effect when I boot into linux? Is there a linux solution if not?

Re:dual boot?

[Anonymous Coward]

Linux has this built in. Works for desktop CPUs too, I use it on mine so I can sleep.

Re:dual boot?

[pmjordan]

Yes, changing the clock speed in Linux is no problem (where supported) and many distros even have automatic throttling and such activated. However, the article and GP are talking about changing the CPU's core voltage. Can Cpufrequtils change that? If not, is there another way of doing it? I suspect a very low-level hardware call needs to be made for this to be done.

~phil

Re:dual boot?

[Anonymous Coward]

From what I can tell, each time the computer is restarted, the values reset to default (or whatever values the BIOS will set).

Non-expert wants to try it!

[erroneus]

I would love to see if I could make my laptop a little less warm, so let's see if anyone can help:

I am running the following:

Dell Inspiron 8600, 1.6GHz, 768MB RAM, ATI Radeon 9600. Running Fedora Core 4 fully up to date and using proprietary drivers successfully.

The cpuspeed daemon seems to be doing it's thing properly by adjusting stuff on the fly so I suspect it is in /etc/cpuspeed.conf that I can effectively make adjustments but I don't know where to begin tweaking. So, anyone out there already know what I should do to give this a try?

Re:Non-expert wants to try it!

[und0]

There are a couple of kernel patch out there, if you look with google for "gentoo pentium undervolt" you'll find a wiki page with a enough information useful even if you use Fedora (i'm with Debian). Basically you need to patch the speedstep-centrino.c source and add a single line bash script to your boot sequence, if you want it. Expect no more than a couple celsius degrees on idle and probably 20 minutes of extended battery life... (my experience with a Pentium M 730 =)

No performance loss

[Delph1]

There is no performance hit here. The thing with undervolting is trying to find the sweetspot for the processor. I.e. the lowest possible voltage at which the processor works just as it is suppose to. If you are experiencing problems you've gone too far. Some users have managed to go as far as 30% with their Pentium Ms.

USB often has its corners cut...

[PHanT0]

I work in support at a hardware company which sells some USB products. On a related note to this article, the processor isn't always the one whose voltage is dropped. When one of customers call-up using a laptop, more often then not the device is fine and it's the laptop who is underpowering the USB port in order to save battery life which is causing the problem.

Just food for thought.

Done something like that

[jigjigga]

I run my 2ghz pentium m at 200mhz in the car to play mp3's. I get around 5 hours of constant playback with it downclocked to that level.

Re:Done something like that

[rjstanford]

Wow. That's almost as long as a laptop with a new battery running at a pretty high speed. Of course, you can also spend $50 or so (probably not a big deal if you've got a 2ghz M laptop) and get a mobile laptop power supply like one of the Targus ones... then it will last (effectively) forever.

Re:Done something like that

[jigjigga]

Well i didn't elaborate enough. I get 5+ hours with battery to spare. Usually in the car rides going home. I bet near 6 hrs is doable with powerstrip to declock the video card etc... Fun challenge.

Re:isn't this what speed step did back with the PI

[krgallagher]

"isn't this what speed step did back with the PIII"

This is what AMD did with their PowerNow!(TM) technology. It dynamically adjusts CPU power consumption based on CPU load. According to AMD, it can reduce CPU power at Idle by 75%. I know on my laptop, I can hear the fan speed up and slow down based on the load on the CPU.

Transmeta's LongRun technology

[yorktown]

Back in 2000, Transmeta started producing chips with Longrun technology, which automatically varied processor frequency and voltage many times a second in response to the current processor load. The technique is quite effective in reducing heat and increasing battery life.

Re:Transmeta's LongRun technology

[Anonymous Coward]

Hate to break it to you but every mobile CPU has been doing this for many years.

Exactly

[EVil Lawyer]

Seriously. I mean, I could pretty drastically "lower the voltage" and "improve batterly life" if I replace the LCD with a single LED, and the hard disk with a block of cheese... but that doesn't quite make it a good idea.

Re:Exactly

[orkysoft]

That entirely depends on whether the LED is blue, and the kind of cheese you use.

Re:Exactly

[Anonymous Coward]

Not if the LED is off.

Well, it can be done......

[AKosygin]

If they simply converted to Kelvins before doing the percentage..... but then, were they smart enough to use how many Kelvins?

Temperature Discussion

[JoeGugel]

There is little arguement to be made based on what temperature scale is used. Every temperature on any scale has a 1:1 equivalent on another scale.

The proper means to calculate a percent change in temperature would be to divide the change in operating temperature by the difference between the normal operating temperature and the inactive temperature. The inactive (no power) temperature is the "meaningful" constant everyone seems to be complaining about. "Zero" means little here.

([normal operating temp] - [new operating temp]) / ([normal operating temp] - [offline temp])

What scale you use for the temperatures is completely irrelevant because they divide out.

Sacraficing speed for power?

[Dream1979]

By lowering the voltage to the processor you can not only drastically lower the heat dissipation, but also increase the battery time significantly. But wouldn't that significantly reduce the speed of the processor? If so it will take longer to perform the tasks, and that pretty much cancels out the longer battery life... No?

Re:Sacraficing speed for power?

[Delph1]

But wouldn't that significantly reduce the speed of the processor? If so it will take longer to perform the tasks, and that pretty much cancels out the longer battery life... No?

No, you don't sacrifice any performance, You just try to find the lowest possible voltage at which the processor will work just as well as it did before. Processors are simply set to work at a voltage at which all of them work well, but in fact many of them work just fine at lower voltages to.

Some of you talk about Intel Speedstep technology and similar which lowers the frequency and voltage when suitable, that is not the same thing. This should be considered an improvement of that as you try to go even further with even lower voltages. Of course without loosing any performance nor causing any instability.

Re:Sacraficing speed for power?

[VirionNW]

I've been using RM-Clock to run my 1500+ (1.33Ghz) in my laptop at 1.2V instead of 1.45V and haven't noticed any major slow-down, I would believe there is a slight decrease, but nothing near what running it at 530Mhz (lowest possible stable speed) and 1.2V would produce. I've found it to be a fantastic way to improve the battery life of the three year old battery (I could upgrade, but that costs money and RM-Clock is free.) I've noticed Sandra says my cpu can go down to 1.08V, but effectively to keep functioning (and not reduce the clock) it needed to be at 1.2V, minimum. The voltage and clock speed aren't directly tied, so it leaves some wiggle room, it would seem, as both experience and this article show. Also, a key thing I noticed is that even if the lower voltage may hurt preformance, my FSB is at full speed when I clock to max with lowered voltage, unlike the reduced FSB at lowest clock with lowest voltage, so even if I lose some preformance, it still beats the pants off the normal reduced power settings. Also, if anything, it at least keeps my fan quiet for longer, it's got a nice "replace me" buzz going and I haven't the time to do the work, so this works as a nice side effect.

Re:Sacraficing speed for power?

[Nazmun]

Good god please rtfa people... 9.9 out of ten people are totally misinterpretting things, thats based on +5 comments. Even for slashdot this is extreme.

This method does not lower performance, performance of the cpu is DIRECTLY based on clock rate not voltage. The higher the voltage the higher clock rate your cpu can operate. However, manufacturers leave a bit of leeway on the voltage rate.

Processors are made in batches. Not all processors even if they are from the same batch or same line will perform equally. After testing a processor line the manufacturers will rate them and set them at a higher voltage requirement just to be safe.

Manufacturers will often (as amd did a lot of back in the day, which made overclocking popular but i'm sure intel does it too) sell more capable chips at a lower speed rating then they are capable of running to meet market demand.

Because of these two reasons there is a good chance you can run your processor at a lower voltage (the decrease in voltage reduces power usage by a square for processors) then it is rated at. This is true even if the processor is using something like speed step or anything that comes from the manufacturers that throttles down the voltage and clock rate simultaneously.

Your cpu may throttle down to 800mhz and 1.1 voltage when it's using the lowest cpu power. However, there is a chance that it may run at 1.0 voltage as well, this often scales to the high end as well. 0.1 may not sound like much of a decrease in power requirements but the current drawn in the end is not linearly decreased.

The purpose of the article seems to be to find the minimum voltage your processor will run with 100% accuracy. Your mileage may vary but you should still see a reduced power requirement if your willing to spend the time testing (which can be set when you go to bed =P).

RESULTS ARE IN CELCIUS

[Anonymous Coward]

Look at the image
http://www.nordichardware.com/skrivelser_img/465/t empgraf.gif

See the big "C" at the end of the scale?

That is REALLY hot!.

But yes, the percentage is meaningless. They should have at least subracted out the ambient temperature which should have been 25-27 C giving a decrease in *temperature* of 26-27%

Further work would need to be done to determine the reduction in *thermal engergy* output in watts or BTUs.

Re:BushCo's State Of The Gulags: +1, Informative

[mjh49746]

Does this have ANYTHING to do with undervolting a laptop? The Politics section is four links down. Thank you.

"They are not in Baghdad. They are not in control of any airport. I tell you this. It is all a lie. They lie. It is a hollywood movie. You do not believe them."

PowerDemandign components

[pscottdv]

No. He's talking about the new PowerDemanding (TM) components
They're designed to increase current to meet their power demands.
If they can't get their power that way, they're designed to lobby
Congress.

Ok all that for 10 min of battery life?

[mdman]

All this for 20% does not sound worth it to me.. My time is worth more then the time it takes to save 10 min of battery life.. and my work is too important to risk my pc crashing because it is underpowered.

Re:Ok all that for 10 min of battery life?

[mkosmo]

So true. The time spent wasted if your proc also underclocks in the undervolt will likely be greater than the battery life saved, putting you at a net loss. Net loss means loss in productivity, which means it was worthless. Unless youre a hobbyist just messing around, of course.

Re:Ok all that for 10 min of battery life?

[cjb110]

urm, if it works at the slower speed, it works, unlike overclocking it can't 'half' work. The only time this would be bad would be if you need to run something that little bit quicker...but given that you've got a lot more time than the world has power I doubt that.

LOL relative temperatures

[NotQuiteReal]

I know, I know, there is an absolute temperature scale, but I'll bet most people wouldn't bat an eye at relative temperature comments like "wow, it's 50 degress today - it was twice as hot last summer when it was 100!".

Oh - here's a good one. In Farenheit - boiling water is only 6.625 times hotter than ice! Of course on the Celsius scale boiling water is infinitely hotter than ice!

Quick - what's the opposite of circle?

Re:LOL relative temperatures

[Anonymous Coward]

Hole?

Re:LOL relative temperatures

[robophilosopher]

Quick - what's the opposite of circle?

!circle

StarForce

[tepples]

Adding more ram, dimming the screen, and using a virtual cd drive should also help considerbly.

Using a virtual CD drive also makes your system incompatible with too many PC games.

Re:StarForce

[b0bx13]

...Which is why you get a program to "hide" the virtual drive software from the game. It's not hard.

Interesting... How much extra battery life?

[bdwoolman]

And how much noticibly cooler is it? Thanks. d:-b

Re:Interesting... How much extra battery life?

[TheGuano]

My own tests with Arctic Silver 5 on 15" Thinkpad T42p:

1.8GHz at 1.340V (default): Idle 40C. Load 58C. (Approx).
1.8Ghz at 1.134V: Idle 39C (there won't be much difference at idle). Load 51C.

600Mhz at 0.980V (default): Idle 35C. Load 41C.
600Mhz at 0.700V: Idle 35C. Load 39C.

I don't remember what the exact difference was in battery life, but I think I got about 30 minutes more out of a 12-cell battery (from 4.5 hours to 5 hours).

And older P3's underclock automatically

[wsanders]

Pre-Centrino P3's underclock automatically as well. My piece of crap Presario 1750 drops the CPU speed from 733 to 600-mumble when external power is removed. This feature extends my battery life from about 1 minute to nearly two minutes. (Compaq's battery charging logic in these Presarios was bogus and their batteries were all defective or ruined by the bugs in the charging logic.)

Revolting

[corngrower]

Also, "undervolting" is not a word.

But "revolting" and "revolted" are. We revolted the CPU to 1.05 Volts.

Re:Revolting

[Escogido]

now, that is revolting...

Revolting

[Anonymous Coward]

I suggest you cease undervolting your computing device and inrease the voltage back up to the recommended levels. See subject to reference to the name of this process.

Please use Kelvin to compare % temperature change

[Anonymous Coward]

> was able to boast 18% lower temperature

FYI: You can only compare temperatures this way if they are measured in Kelvin.

If it was not measured in Kelvin, then you must convert it to Kelvin before computing the percentage.

Even then, it can be risky to compare temperatures in this way, because there is sometimes a non-linear relationship between temperature change and thermal energy change.

(For example, when you heat up water, sometimes the added energy goes mostly toward increasing the molecular kinetic energy, but sometimes the added energy goes mostly toward breaking inter-molecular bonds. Those two phenomena have a very different impact on how much the temperature rises.)

Comparing energy or power levels is usually more meaningful.

dell battery

[Anonymous Coward]

great, now my 1.5 yr old shitty dell battery will last 5 minutes instead of 4!!

My x40 gets ~6 hours anyway

[Superfarstucker]

I think the processor is a LV or ULV pentium M (1.4 GHz), but the whole premise of the machine is battery life, which comes at the cost of some functionality, such as no cd drive, well worth it though. I rarely need to worry about finding a wall for juice, only If I forget to dock the bloody thing for like 3 days straight. Not to mention it weighs ~ 3 lbs, which doesn't really matter when you're packing around 35-40lbs worth of books.

Hello,

[garbletext]

I have a first person point of view perspective to give about my laptop, and a specific time estimate, as well as a comparison with another product that was mentioned. Would you like to refute my post with a vague, pseudotechnical one that basically amounts to a useless bitch? Fantastic, anonymous Coward. I commend you to the next life.

Very Bad Re:Bad Idea

[Anonymous Coward]

Very dangerous. Lowering the voltage below the specified level increases sub-threshold conduction and dramatically increases the rate of charge implantation - eventually damaging the chip. If you have a marginal CPU, this can quickly kill a part.

I dunno about you...

[Khyber]

But undervolting (Not a word people, it's 'under-powering,' as you're reducing the power if you reduce either voltage or amperage.) can cause some serious problems. I've severely damaged digital cameras and wireless headphones by using lower-powered batteries. (1.2v as opposed to 1.5v) Why could this not happen to a processor, hrm?

Re:I dunno about you...

[Anonymous Coward]

Because the processor will not draw more amperage to compensate, while your digital camera probably has a voltage step-up circuit that will use more current at lower input voltages. If you think about it, there really aren't many parts out there that can run at the 2.2-3V that a couple AA's provide - there's almost always a voltage step-up involved.

However, some chips will go nuts after the voltage drops below a certain point - both the pull-up and pull-down gates will switch on at the same time, effectively short circuiting the chip, but you'll probably be well beyond the operational range of the chip before you run into this.

CPU efficiency vs. heat rejection

[murderlegendre]

Help me out here.. I sure one of y'all know this:

If a CPU (let's take a 486DX as an example) is using 10A @ 5V, it's said to be consuming 50W of power. Question is, what is it doing with all that power?

In relative terms, electrons are pretty light little things. Sure, there are brazillians of them moving about the CPU, but their cumulative mass is still negligible. It can't take much real power to shuffle them around (nothing like rolling a giant boulder to the top of a hill) - unless of course, you account for electrical resistance (which we will).

Of that 50W consumed by the CPU, how much is lost as heat? I'd tend to think that pretty much all of it is.. 99% let's say. Is that right, or is there some other 'work' being done that I'm not aware of?

If CPUs were inherently superconducting devices, would they only use a tiny fraction (like 1%) of that nominal 50W?

Re:CPU efficiency vs. heat rejection

[Pyrrus]

It's pretty much all lost to heat. The "work" done by the electricity it to provide a signal where high voltage indicates 1 and no voltage (ground) indicates zero. Every time a transistor switches either from 0 to 1 or from 1 to 0, current travels through it, using power which is released as heat. The higher the clock speed the more transistions, thus the more power consumption. Lower voltage reduces power consumption (power = volts x current(amps)), but as the "high" voltages becomes lower, the transistors much be more precise (it's easyer to tell the difference between 0V and 5V than it is to tell the difference between 0V and 2V). This is why overclockers usually increase the voltage, since at higher than spec frequencies there is more signal degradation which could (and does) make the system less stable.

Re:CPU efficiency vs. heat rejection

[Anonymous Coward]

If Brazillians has become acceptable as a number, let me propose a new -- similar sounding word that at least has more giggle value:

...

brassiereians

long life battery

[cdn-programmer]

What we need is a cartoon of a lap top mod that connects a 60 AMP HOUR deep cycle battery. This would be so funny and you can get DAYS and DAYS of service from it. haha.

These batteries are about the size of a lunchbox so they are totable.

what about replacing some parts...

[josepha48]

.. like if you have an internal cdrom, do you get any benifits by removing it or disabling it?

Also what about replacing the hard drive with a flash drive, would that be better or worse?

Re:what about replacing some parts...

[Meostro]

Disabling your CD will help, but just keeping it from spinning up is probably fine.

You could also replace your HDD with a flash drive, or better yet, get yourself a hybrid drive. In the past couple days I saw something about IBM trying this, but I don't remember where and I can't find the reference. Just Google for it, there are tons of references out there.

Similar human procedure

[No2Gates]

George W. Bush was the first recipient of this in human form. The synapses were reduced by 80% to save energy.

Re:Similar human procedure

[Mike Buddha]

Haha! Yeah! The President is STUPID! LOL! ROFLMAO!

Re:Similar human procedure

[Anonymous Coward]

The President is STUPID!

*So* true.

Log scale?

[Atario]

Shouldn't the reduction be considered on a log scale as well? Since a 50-degree reduction from 105 degrees above room temperature to 55 above is far easier than a 50 degree reduction from +55 to +5?

Rovclock

[wiresquire]

For you Linux-ers who have ATI cards with no PowerPlay (it's disabled in my video BIOS - bastards!!), I'd recommend checking out rovclock.

While it doesn't actually reduce voltage, it can be used to underclock GPU and memory speed. My somewhat unscientific testing has shown no major differences between fglrx and radeon + rovclock with 2D, but I did note a 27% decrease in battery draw for 3D using the fglrx driver.

Of course, you're trading performance for battery life, and why you'd want to eg, play a 3D game on battery I wouldn't really understand

YMMV
ws

Re:Rovclock

[Anonymous Coward]

Sorry about the anonymous coward thing, but, as for playing a 3d game on battery power on my laptop... half life 2 and WoW while i'm on my lunchbreaks at work. There aren't any outlets nearby. But i still play, since i'm bored.

Maybe too much

[Poromenos1]

Not only do laptops underclock automatically, but they report the wrong processor speeds too. My 2800 XP reported some 700 MHz and I had great trouble running programs that checked to see if the processor was fast enough (I couldn't get it to speed up to full no matter what I did).

Re:Maybe too much

[Anonymous Coward]

google for SpeedSwitch XP

Re:Maybe too much

[Tolookah]

rar/unrar something, play a video, virus scan or find a file on your computer

All of those things will step your processor back enough for the software to check (I have this issue while installing, mine scales to a 600MHz processor when its bored)

Such Language...

[wernst]

"Undervolting" is a perfectly cromulent word.

GRAMMAR POLICE

[Anonymous Coward]

It should be affect, not effect! You need to order a book that covers the basics of grammar.

Re:GRAMMAR POLICE

[Anonymous Coward]

It should be affect, not effect! You need to order a book that covers the basics of grammar.

And on that note... It should be 'vocabulary', not 'grammar'!

Half of slashdot is clueless on this one...

[black hole sun]

No shit? I have an Acer Travelmate 8104 and I have the same control panel you do. All Pentium Ms come with thermal throtlling. The point, dear friends, and what makes this useful, is running the laptop AT FULL SPEED but with a lower voltage. My max speed is 2.0GHz, with a default voltage of 1.308 V. I can safely reduce this to 1.068 V.

I can also take my min speed voltage -- 700MHz -- and reduce it as well, from 0.988 to 0.700 V.

The REASON for doing this is that Intel gives a generous amount of power to their CPUs--enough to make sure ALL (or at least 99%) of their wafers from the factor work correctly. More often than not, you can decrease their "safe" value an appreciable amount to raise battery life and lower thermal output.

Re:Way to be a dick!

[Anonymous Coward]

Or stupid, lazy and irritating people, aparently.

Easier Solution:

[thepotoo]

All you do is drop voltage to 0. Once you've done this, you'll see a drastic increase in battery life, and virtually no heat output.

The downside is that it takes like forever to load Word...

Re:Please use Kelvin to compare % temperature chan

[Jim_Callahan]

Or Rankine, which also has a zero corresponding to 0 entropy.

Re:Good luck...

[Anonymous Coward]

Maybe, just maybe, some adults will read this article. You know, adults -- people who don't put fucking games at the top of the list in terms of requirements when buying or improving a computer.

Some day, you'll understand.

OS X Tools?

[Kichigai Mentat]

I've heard a lot about Linux tools, Windows Tools, but what about Mac OS X tools? I'm assuming that this isn't something possible only with x86 chips.

Re:OS X Tools?

[Anonymous Coward]

I'm betting there is a possibility, but no actual software that does it. Hell if I have any idea how to put it there.

On the other hand, G3's and G4's run relatively cool as it is (ever seen the crappy heatsinks on the 60x's? yech...), and OS X does a pretty bang-up job of saving power wherever possible (then again, I'm comparing this to Linux which does it terribly if you don't have things set up perfectly, and Windows which is merely okay).

But if someone figured out which virtual switches to pull....

Re:OS X Tools?

[Kichigai Mentat]

Well, I'm aware that the G3's and G4's run pretty cool. Wasn't that the reason Apple went with the PowerPC chips instead of x86 in the first place? That, and I'm also aware that idling, making my XBox average 124F requires the fan to be constantly spinning at about 0.5x, but making my iBook (1GHz) run below 120F on battery requires next to nothing. Running on the power cord, it only hits 120F+ if I start having it pull heavier work loads, and it's never gone over 150F before.

However, even OS X's tight power management system (I've pulled 7 hours once) I'm sure can be tweaked a little. And with all the open source apps for Linux out there, I thought perhaps someone figured out how to make it work on OS X.

Re:Please use Kelvin to compare % temperature chan

[typical]

I would think that you'd care about percentage deviation from ambient temperature, not the deviation from absolute zero.

Firefox users are immune to this

[typical]

but without obnoxious "i spread my article over infinite pages in order to get more clicks" practice.

If you use Firefox as your browser, you have the antipagination plugin available, and no reason to be bothered by people that spread their articles over multiple pages.

(Also, try reading webcomics using antipagination. Woo!)

Re:Counter productive maybe? HYuuh HYuhhhhh

[davidsyes]

You'd end up with FrankenTop, hucking and bucking on the table... A laptop one electrode short. You might damage your laptop, too.

But, for those who are into pr0table Pr0n, I've got some regenerative gripping pads that you can frantically rub while you discharge your shorting probe. You just have to stroke the laptop a bit faster so you can keep up the differential to the battery recharge circuit. Just make sure you don't short-circuit your laptop in the process. No warranties provided. Order in the NEXT FIVE MINUTES to get your FREE re-purposed dental dams to to to to enhance protection of your laptop display.

And, if you act NOW, we'll even show you how to convert your member into a Tesla coil so you and Frankentop can GET IT ONNN!!! Listen to Al Green whyle u go GANGreen... If you follow our instructions, your static discharges may come so exceed those of the Flies Erectr0nix in Fremont, CA store. With our coil, you can REALLy blast your caps, all three of them... Get our customized jzzzhhh... jzzzhhh... jzzzhhhhh ... jzzzzhhhhzzzzzh ring tones for 1/2 off and get a free defibulator, too.

And, if you REALLY act now, we'll send you the Sizzler Amplifier and a pair of pink earmuffs and accessorized mickey-mouse-ears with rifles painted on the left and a bazooka on the right ear.

If you correctly answer the trivia question, you are elegible for a prize. How well-equipped was Frankenstein. Was he built in the eyes of his master's blaster?

Orders will be taken only between 0001 and 0003.

Undervolting my zv5000z since July 2004

[Brian Stretch]

I've been undervolting my HP zv5000z Athlon 64 notebook since July of 2004, and I got the idea from Silent PC Review where people were undervolting desktop Athlon 64s. I switched to using CrystalCPUID to manage speed and voltage since that initial post was written and most people have switched to RMclock. Lots of people on R3000 Forums and the HP forum at NotebookReview.com have been undervolting their notebooks. I believe I was the instigator on both of those forums.

Anyhow, I'm still undervolting, now with a Mobile-class Athlon 64 3200+ CPU that I swapped in some time ago (HP only used DTR-class chips). AMD is very conservative with their default voltage levels so there's lots of room to work with. I've set my CPU to run at roughly Low Voltage-class levels, close to what a Turion ML is rated to do. The machine is solid. I got great battery life before and even better battery life now. I was able to play DVD video for 3 hours (12 cell battery) and get closer to 4 hours in general use. That's with a 7200RPM HD too. Not bad for a desktop-replacement behemoth.

Note that if your machine usually runs at AMD's 800MHz idle speed, undervolting won't buy you much (if anything). AMD's PowerNOW! is already extremely efficient in normal use. If it jumps to full speed a lot, undervolting does wonders.

So, there's no need to suffer with a 32-bit-only Intel CPU and their awful integrated GPU if you want a long-lasting notebook.

(Anyone know how to undervolt under Linux? That's the one thing I've been missing...)

Dynamic voltage and frequency scaling

[akuma(x86)]

*Dynamic* voltage and frequency scaling is already available in processors from both Intel and AMD. System software can take advantage of this. Look up P-states and C-states. I think Vista supports these features. I'm not sure about XP. Linux of course supports it - look up power 'governors'.

The CPU designers went through a lot of trouble to provide software with the hooks to manage the voltage and frequency. Please don't go destabilizing your systems by manually "undervolting".

Power ~ Capacitance * Frequency * (Voltage)^2

Reducing frequency reduces power linearly.

Reducing voltage is a MUCH bigger lever as it reduces power CUBICly. Reducing voltage linearly also reduces frequency linearly which is why you get a cubic relationship.

There is only so far you can go in reducing voltage. Once you hit the threshold voltage of the transistor, it won't switch properly - so at this min-voltage, all you can do is frequency scale lower.

Re:Dynamic voltage and frequency scaling

[Use Psychology]

Reducing voltage is a MUCH bigger lever as it reduces power CUBICly. Reducing voltage linearly also reduces frequency linearly which is why you get a cubic relationship.
wrong... it's a quadratic scaling... cubic is power 3.

Re:Dynamic voltage and frequency scaling

[akuma(x86)]

Uhmm...Yeah cubic = 3

1) Frequency ~ Voltage
2) Power ~ Cap. * Freq * (Voltage)^2

where ~ means "proportional to"

Because of #1,

Power ~ Cap. * (Voltage)^3

You might want to design some processors before you start saying something is wrong first.

Re:isn't this what speed step did back with the PI

[Zog The Undeniable]

I'm sorry, but I couldn't read all of your comment. All those repeated ®'s damaged my retinas.

Old news

[kahunak]

As you can read here this is really old. I've using it for a long time without problems, I've even patch the linux kernel to use the minimum-safe voltage I obtained on windows. There is no performance hit, as you are just making a fine adjustment of voltage based on your real CPU instead of Intel safe adjustment based on mean CPU values.

The main benefit for me is lower temperature, I have a Dell D800 with a 2.1Ghz centrino plus a GeForce 4200Go, using also powermizer under Windows (or nvclock under linux) to underclock the GPU I get no fan activity at all when surfing or editing files. When I need power it just run faster at higher voltage (still lower than default for that multiplier).

- german

Saved me having to sell my Laptop!

[rehashed]

What a brilliant article!
So good in fact, I've registered and posted for the first time in many years of my slashdot reading because of it.

I have a mitac AMD64 "Desktop Replacement" laptop. which has been having horrible trouble overheating. It basically means that I have been unable to play any great games, as the CPU would inevitably overheat (hitting 90'C according to mobmeter), causing the laptop to crash just as gaming was getting interesting.
By implementing the techniques shown here, I can keep my CPU running at the maximum speed, yet a LOT cooler - meaning the laptop isnt crashing!. Its now running rock-solid, and I can finally play games!!!!

Thank you! Thank you! Thank you!

Guild Wars here I come!

Anyway - THERE is the true value of this article to me - an extra 10% battery life may not be worth all that much, but killing monsters sure as hell is!

marketing

[willCode4Beer.com]

"Why don't laptop manufacturers do this?"

Probably because a "faster" CPU generates a higher price. People tend to pay more for higher MHz, regardless of actual performance.

Battery downside

[fastgood]

Rechargeable AA batteries have been doing this by changing the 1.5V standard to 1.25V.

They've taken it one step further lately so that my 6.0V devices are now powered by 4.8V
(you need 5 rechargeables to properly replace 4 standard batteries)

The only benefit is that Eveready [Ralston Purina] saves money by not investing in R&D.