Post your laptop's from-the-wall power usage under light load.

Procedure: Plug your turned-off laptop into a Kill-A-Watt or similar. Remove the battery, because if it starts charging it'll really throw readings off. Start it up, and record the highest wattage you get while starting up (#1). After starting up, record the idle wattage (#2). Next shut the lid to turn off the LCD, and record the wattage then (#3). Now load Firefox, and record the highest wattage you get when it starts (#4).

I've been using dynamic cpu scaling, and the drives are not being spun down.

My results:

Compaq Armada E500/Pentium 3 900mhz W2k
1. Peak Startup 37w
2. Idle 17w
3. Idle No LCD 9w
4. Light Activity 28w

IBM Thinkpad T40/Pentium M 1.5ghz undervolted Ubuntu7.04
1. Peak Startup 28w
2. Idle 17w
3. Idle No LCD 12w
4. Light Activity 22w

Edit: Upgrading the T40 to Ubuntu Linux 7.10 has reduced the idle usage by a couple watts:
2. 13w - Idle, Wifi
3. 10w - Idle, No LCD, Wifi

Some others:
7w - Idle, No LCD, No HD, No Wifi.
9w - Idle, No LCD, No HD, Wifi
10w - Idle, No LCD, Wifi
18w - Idle, Brightest, Wifi
24w - Peak (Prime95), Brightest

Everex NC1502 (Walmart cheapy)

15.4" Widescreen, Via C7-M 1.5 ghz, SATA drive, wireless network/internet, Windows XP.


Screen off, idle, WLAN off - 11 watts.
Screen off, idle - 12 watts.
Idle desktop (screen full bright) - 19~20 watts.
Browsing (light loads) - 19~20 watts continuous with 23~24 watt peaks.
CPU Burn - 29 watts.

Screen off, CPU Burn running - 21 watts.
File copy from DVD to hard drive - 23~24 watts
CPU Burn, file copy from dvd to hard drive - 33~34 watts.

Add 1 watt to the non CPU Burn figures above if any USB device is connected. See Anandtech article for a detailed description of Windows' USB driver preventing the CPU from sleeping - just by having a device connected. It translates to about a continuous, measurable 1 watt increase in my system. One that also shows up in performance monitor on my system, unlike Anandtech's system. Though Anandtech doesn't discuss it in the article, Vista's USB driver does the same thing.

Ironically, the silly bug is just enough to cause my cooling fan to cycle on and off with a USB mouse connected, where without it the fan stays off. Those CPU sleep states might only last for minute fractions of a second, but my C7-M is in sleep state 50% or more of the time - while the computer is in use. It adds up in power and heat savings.

Copper! long time no post! glad to know you still check in from time to time.

Thanks, Jaganath. It's good to be back and see you and everyone else.

I don't have as thorough numbers for this system. It's a work laptop I tested a little while ago.

Dell Inspiron 1300. 1.7 Pentium M Dothan. 15" screen. 512 MB RAM. Carries a Centrino label.

Display 'bright'; WLAN 'on': Idle 16 watts, Peak 33 watts. (Prime95 running for peak)
Display 'dimmed'; WLAN 'on': Idle 10 watts.
Display 'dimmed'; WLAN 'off': Idle 9 watts.

Dimmed shouldn't be mistaken for off. It's the lowest brightness setting like used in battery mode. Judgeing by the Everex numbers I'd guess that adding a DVD to hard drive file copy to the Prime95 test would add yet another 4 to 5 watts to the peak.

Dell Latitude D400, Pentium M 1.7 Banias, Windows 2000, 12.1" screen, 512 MB RAM. No Centrino label (Dell wireless card.)

Screen off, idle, WLAN off - 8 watts.
Screen off, idle - 9 watts.
Idle desktop (screen full bright) - 13 watts.
Browsing (light loads) - 13~15 watts continuous with 22~23 watt peaks.
Prime95 - 34 watts. (Windows 2000 doesn't recognize CPU Burn as a valid 32 bit app?)

Screen off, Prime95 running - 30 watts.
File copy from DVD to hard drive - No DVD/CD
CPU Burn, file copy from dvd to hard drive - No DVD/CD

I did some underclocking and volting with the Dell Latitude D400 above with some pretty darn good results. I also discovered that I didn't have my screen set to full bright during the previous tests. It took a while to figure out why my results were differing, but the old lightbulb finally went off. Drop two notches on the screen and my figures finally synced.

With this test I wanted to mimic a Banias 1GHz ULV model. Its voltages are listed as 1.004/0.844, maximum and minimum, respectively. I was able to beat these numbers using RMClock. They are 600MHz@.748v, 800MHz@.796v, and 1GHz@.844v. All steps were temporarily set as the highest step and tested for stability overnight with CPUBurn, which I was finally able to get to run on Windows 2000. I had to download the full version of CPUBurn with its installer and then it worked fine.

=================================

First, some new results with the screen actually at full bright and the CPU using its default clock speeds and voltages. Remember that in these tests the CPU is allowed to clock all the way up to its max, 1.7GHz. CPUBurn is being used to load the CPU:

Screen Off: Idle 9 watts. Load 33 watts. (You can see that CPUBurn puts a higher load on the CPU than Prime95 by comparing this loaded result to that of the screen off, Prime95 result above. The previous result was only 30 watts with Prime95.)
Screen On: Idle 15 watts. Load 38 watts.
Screen On, CPUBurn, & Defrag: 41 watts.
Internet Browsing: 15 to 20 watts continuous with occassional 25 to 26 watt peaks.

===================================

The results for ULV levels @ load are impressive. What is interesting is how little the idle and internet browsing states are affected even though the CPU was stable at an entire .2v lower than default. The Pentium M is extremely efficient at idle with its built in sleep states. Remember that in this test the maximum speedstep is set to 1GHz:

Screen Off: Idle 9 watts. Load 12 watts. (Only a 3 watt increase with CPUBurn running!)
Screen On: Idle 15 watts. Load 19 watts.
Screen On, CPUBurn, & Defrag: 21 watts.
Internet Browsing: 15 to 18 watts continuous with occassional 19 to 20 watt peaks.

=================================

As good as the peak loads are brought down by the undervolting and speed limiting, they don't make a whole lot of difference in casual use like internet browsing. Leaving the CPU set to 600MHz and undervolted nets a most impressive 17 watt then 11 watt CPUBurn load with the screen on and then off. I like the screen two notches down from full bright and I find the 600MHz speed to be more than enough for browsing and vids. This combo nets 13 watts idle and 15 watts peak. Shutting off RMClock and letting the CPU run at its default voltage, but set at "Battery" such that it doesn't speedstep above 600MHz nets the same 13 and 15 watts. No need for RMClock at this level.

Some partial results, a few of which I'm not sure of.

Toshiba A15, 2.4ghz Northwood P4, XP
15w idle, screen off
18w idle, screen dimmest
23w idle, screen brightest
30w light activity, screen brightest
70w load (Prime95), screen brightest.

Toshiba A75, 3.0ghz Prescott P4, XP
36w idle, screen off
38w idle, screen dimmest
42w idle, screen brightest
50w light activity, screen brightest
116w load, screen brightest

A Prescott in a notebook! Ouch. How well does the design do at keeping the laptop cool and quiet?

I have a second Latitude D400 with a 12.1" screen. This one carries a Centrino label and has a 1.8 GHz Dothan and 512 MB of RAM. In addition, it has a low power 8 GB Transcend SSD running Windows 2000. I did some searching around in Device Manager and found two additional power saving features that I hadn't found on my other Latitude. Both the sound card and USB hub controller have power saving features I enabled. Together they shaved one watt off the idle deskop, but don't seem to have made a difference under load or with the screen off and idle. Don't ask me why!

This system is set to speed step between two P states using RMClock, both on battery and AC. They are .6 GHz @ .716 Vcore & 1.4 GHz @ .908 Vcore.

Screen off, Idle: 8 watts.
Screen on, Idle: 13 watts (Was 14 before adjusting sound card and USB hub).
Screen on, Idle, WLAN off: 13 watts. (No measurable change. Way to go Intel!)
Screen on, CPU Burn: 19 watts.
Screen on, CPU Burn, HD Tach: 20 watts.

Overall, Banias to Dothan is a big improvement. Dell wireless to Intel wireless is an improvement. And, HDD to SSD is an improvement. Over my other D400, I can run .4 GHz faster for the same power draw. In fact, the power draw is actually less at an idle desktop and when running heavy disk activity.

Badly. The fan runs continually, and it's much louder than the Northwood (whose fan turns off time to time).

IBM R52, 1.86ghz, ATi X300, XP
First wattage is from the wall, the occasional second in parentheses is the ACPI wattage from the battery.

Thought I'd post my results, though the collection of P-Ms/Centrinos on here make mine look like a power-hungry demon...

My laptop: Acer AS5002WLMi, Turion 64 MT-40 Lancaster (came with ML-30), useless SiS M760GX graphics, 2x512MB Infineon DDR333 RAM, 15.4" widescreen, 100GB 4200 rpm Toshiba HDD, WinXP Pro

Peak startup: 52W
Peak CPUBurn: 53W (55W once fan comes on and throttles to 2nd speed)
Idle, LCD bright, light usage: 29W
Idle, LCD bright, idle: 27W
Idle, LCD dim: 23W
Idle, LCD dim, WLAN off: 22W
Idle, LCD backlight off (LCD still on though), WLAN off: 21W
Idle, LCD off, WLAN on: 19W
Idle, LCD off, WLAN off: 17W

I should note that my MT-40 will not tolerate undervolting. I run CrystalCPUID with the following voltages: 800MHz 0.900V, 1500 MHz 1.025V, 2200 MHz 1.225V (it actually isn't even stable at the rated 1.200V, though it boots at 1.300V)

IBM Z60m, P-M 2ghz, ATi X600, Ubuntu 7.10
Powerplay not used because I didn't feel like installing the official ATi drivers.

Might be nice to add one more measurement -
#0 (or #5) Phantom power
- plugged in, turned off, fully charged battery (or no battery)

i.e. how much does it use when it is "off"

The laptops were pretty frugal about this, typically it says 0w most of the time. So it's not easy to measure, since it's some fraction of a watt. As opposed to my desktop that takes 7w doing nothing.

I can report pretty much the same as aaa. Not only do all 4 I measured read 0 watts plugged in, but off, they also all read 0 when asleep (not hibernating). Laptops seem very efficient sleepers too.

They are obviously consuming some power when asleep, but the Kill-O-Watt meter I use is only accurate to one watt and reads in one watt increments.

According to my T40's ACPI, it's drawing somewhere between 0.2 and 0.5w when asleep.

Toshiba Portégé R500-100

Peak startup: 28W
Idle: 16W
Idle (WLAN off): 14W
Idle LCD off: 15W
Light activity: 21W
Peak (loaded with TAT): 25W

Havent had this machine very long. I'll get back after some tinkering.
I'll probably try some undervolting aswell.

Tell me if you want more numbers

Well I want more numbers from other laptops. As you can see, it's mostly old Centrinos here, which is getting redundant. Already I can see the interesting fact that my undervolted old Centrino T40 uses less power than newer laptops (10w vs 15w for the Portege, idle/lcd off). My theory is that it is the result of the undervolting limits Intel put on the newer stuff, but there aren't enough numbers to be sure. So, people out there, if you got an Athlon/Turion(x2)/Core(2)Duo/etc, please post!

My portégé has a U7600 which I havent seen undervolted before.

I hope it can go down. Even just a little bit.

BTW. Is there a NHC undervolting guide somewhere?

I know the regular T series cpus can't go below 0.95v. Basically no reduction in idle power, but improvements at load. Don't know what the deal is with the U series, though (do post your experience!).

For NHC:
Enable advanced mode to show the voltage tab
Set it to *not* start automatically at startup
Lower the voltages as much as you can, until you get an error/crash
Set the voltages to two slots above where it crashed
When you're sure it won't crash, set it to startup automatically.

Will do. Thanks.

I found an interesting graph on the rmclock forums:
http://img442.imageshack.us/my.php?imag ... re2wv2.jpg
from http://forum.rightmark.org/topic.cgi?id=6:1409

If you have a Socket P Core 2, apparently you can get lower voltages. Beta versions of RMClock let you use the "SuperLFM" voltage (0.8-0.85, for when the FSB is reduced) anytime you want. The FSB does not have to be clocked down for you to use it. Mind you, it's still a beta that has crashed for some people.

Toshiba Portégé 3500 Tablet
-21-24W BTing w/ screen on
-15-17W BTing w/ screen off
-18-20W copying stuff w/ screen off
-3W off, power brick warm

Hallo!

Dell 1520
Core duo T5200 - 1.6Ghz, 2x1gb ram, 160 gb 5400rpm hdd, Nvidia Geforce 8600gt,

Undervolted with Rmclock 2.35 from 1.25v to 0.950v stable

Idle- 800@0.950v - max lightness - 26w (min lightness - 20w)
Load Orthos dual core 1.6@0.950v - 39w (max lightness)
Load gpu - 3dmark 2001- nature loop- 50w - (max lightness)

I wish more post with performance video graphics here.
Please past consumption with 3d load. Thx

Just bought a Fujitsu Siemens Xa3530 with:

Turion 64 X2 RM-70
17in screen, and
ATI Mobility Radeon HD 3650 hybrid with integrated ATI Radeon HD 3200

These are the figure that I got for the laptop when I tested it at desktop (idle):

Silent Mode on, HD3200 on-board graphics - No battery or DVD in drive

Connected to external screen with laptop screen off - 9w
Both screens -13W
Just laptop screen on -12W

Silent Mode off, HD3650 discrete graphics - No battery or DVD in drive

Connected to external screen with laptop screen off - 12w
Both screens -15W
Just laptop screen on -14W

Battery in place and fully charged, laptop off - 0W
Silent mode off, HD3650 discrete graphics, both screens - 15W

The silent mode is not fully silent but underclocks both cores to 500MHz and undervolts to 0.8V without further tweeking. Depending upon load the fan will switch off at idle desk top and is reasonably quite anyway. These are currently available for about £500 in the UK.

I guess it's about time I added my other laptop, which I picked up last year for $450 at Best Buy. It's an Acer Extensa 4420-5963, and came with the following: 14.1" CCFL LCD, AMD Athlon X2 TK-57 1.9 GHz (65nm Tyler Turion core with half cache), AMD RS690M/SB600M chipset (Mobile AMD 690G), 2x1GB DDR2-667 RAM, 160GB 5400rpm Hitachi Travelstar 5K250 HDD, Vista Home Premium SP1 32-bit. One of the nifty features of the RS690M is that it has the ability to clock down the IGP from 400MHz to 130MHz when on battery, which has a noticeable effect on power draw, and consequentially, battery life. I also keep the wired LAN turned off whenever it's not in use.

I've since upgraded the RAM to 4GB of DDR2-667 4-4-4-12, 320GB 7200 rpm Hitachi Travelstar 7K320 (I do wish I had stuck with 5400 rpm now, but it was only $2.50 more for the 7200 rpm model at the time), and it's now running Windows 7 RC 64-bit.

I also undervolt the processor, which by default runs 800 MHz @ 0.900V, 1.6 GHz @ 1.025V, 1.8 GHz @ 1.050V and 1.9 GHz @ 1.075V under the standard PowerNow driver, but I disabled that and use CrystalCPUID to run 800MHz @ 0.775V, 1.4 GHz @ 0.850V and 1.9 GHz @ 0.975V. It will actually run 800MHz @ 0.7375V if I select it in RMClock, but since I have never gotten RMClock stable on an AMD X2 system, I stick with Crystal.

Power:
Idle, LCD 100%, GPU 400MHz, WLAN on, HDD spun up, fan on: 24W
Idle, LCD 100%, GPU 400MHz, WLAN on, HDD spun up, fan off: 23W
Idle, LCD 100%, GPU 400MHz, WLAN on, HDD spun down, fan off: 22W
Idle, LCD 0%, GPU 130MHz, WLAN on, HDD spun up, fan on: 17W
Idle, LCD 0%, GPU 130MHz, WLAN on, HDD spun up, fan off: 17W
Idle, LCD 0%, GPU 130MHz, WLAN on, HDD spun down, fan off: 16W
Idle, LCD off, GPU 130MHz, WLAN on, HDD spun up, fan on: 11W
Idle, LCD off, GPU 130MHz, WLAN on, HDD spun up, fan off: 11W
Idle, LCD off, GPU 130MHz, WLAN on, HDD spun down, fan off: 10W

2xCPUBurn, LCD 100%, GPU 400MHz, HDD spun up, fan on and sped up, 0.975V: 51W
2xCPUBurn, LCD 100%, GPU 400MHz, HDD spun up, fan on and sped up, 1.075V: 60W