I've determined power usage on my desktop and on my server, but I hadn't gotten around to testing the power usage of my laptop. As battery life is always a concern with a laptop, I was particularly curious to see which parts of the laptop draw the most power. So I ran a few tests with my trusty kill-a-watt on my Dell Inspiron 300m. It's an ultaportable 2 pound laptop of late 2003 vintage with the following specs:
3dmark shouldn’t use up full processor load though (or is this another “feature” of intel’s onboard?), so it’s probably a few more watts at peak.
Thanks for the link, Peter, I found RM Clock supports my core duo now. =D It already defaults to as low as it can go, but there are still a few tweaks that can be made.
I’d have to say that I think these numbers would probably hold true…unless you have a huge monitor. I’ve a got a 17" widescreen and dimming the monitor usually doubles my battery life.
I’ve also discovered that if I forget to plug in and run Visual Studio, IIS and SQL Server 2005 with my monitor at full brightness while using wireless I can crush my battery in just under 45 minutes. I’d be interested to see how much power my laptop is using during that.
Thanks for an excelent article, I alwayts wanted to know this but I didn’t have the tools.
Too bad CPUs can’t scale speed well. With the speedswitch XP, I am able to have my centrino pentium M 1,7GHz operate between 595MHz and 1700MHz, but I can scale it so it is 200MHz to 600MHz. Hey, Intel, I want a 100MHz-max scale
So in your other post about 450W desktop computers, you state:
The 300gb SATA hard drive consumes 4 watts
And in this post, you state your sleeping laptop drive consumes 14 watts.
Either you’re doing something wrong, or laptop manufacturers need to start putting superfast 300GB desktop drives into laptops.
I think the 14W is the total draw on the ammeter. Remember it was 15W with the machine on and sitting idle. So you can see that the 18W defragging is +3W for the disk at full use over idle or slightly better than the 300GB’s 4W.
Its probably hard to use these types of measurements to really isolate the power consumption of individual devices and subsystems, though. Defragging will use some CPU and lots of memory, for example, so you can’t just measure the increase in power and attribute it all to the disk.
Still, its a pretty reasonable approximation and Jeff’s takeaways are good: you save very little trying to optimize your hard disk’s awake time and save a lot if you can put up with a dim screen.
… via the BIOS, but I hadn’t considered doing that for my laptop. Thanks for the tip.
3dmark shouldn’t use up full processor load though (or is this another “feature” of intel’s onboard?),
Actually, 3dmark does use 100% CPU. Games always fully load the CPU and video card. Granted that the Prime95 load tends to be a bit more intense, but it’s definitely in the same ballpark.
With the speedswitch XP, I am able to have my centrino pentium M 1,7GHz operate between 595MHz and 1700MHz
This one drops to 1/2 clock, 600mhz, when it’s on battery. I guess I should have recorded power usage when it’s on battery to see the difference. Hmm. I’ll do that tonight.
So you can see that the 18W defragging is +3W for the disk at full use over idle or slightly better than the 300GB’s 4W.
Right, the notebook drive does do a little better. And remember these numbers are plus or minus 1 watt estimates based on total draw from the wall.
I do think they’re reasonably accurate. You could check the spec sheets for desktop 3.5" SATA and laptop 2.5" drives to compare-- the manufacturers do list power draw on their spec sheets.
I don’t know-- this laptop has no internal CD/DVD whatsoever. There is an external one bundled with it, but I am not sure the power usage of an external CD/DVD device will be the same as an integrated internal one!
Most of the drives failed after being powered for 6 months
Wow, that’s interesting. How many is “most”, and what was the sample size?
I sort of assumed people would be hibernating, sleeping, and/or generally powering off their laptops from time to time.
That’s certainly far, far worse than typical 3.5" desktop hard drives do. I’ve had drives on for years (in servers) without incident.
Like estimated power draw, the manufacturer’s spec sheets should have guidelines for the MTBF (Mean Time Between Failure) we could typically expect. This is usually expressed in # of hours.
Which then lists three states. I didn’t realize I had to go up, determine that I should use “Laptop On, Idle at Windows Desktop” as the baseline and perform math - so what then do the other numbers mean? there are 3 numbers for Wifi - are these with the drive sleeping, at idle, or in use? Etc.
Be careful about not putting your drive to sleep. At the company I previously worked for, we used 2.5” laptop drives that were permanently powered on in one of our products. Most of the drives failed after being powered for 6 months. The manufactures design the drives specifically with the idea that they will be powered down periodically. You may want to increase the delay time before the drive is put to sleep, if waiting for it to spin up annoys you when you’re working, but let it go to sleep if you’re not using the computer for an extended period. It may not save much power, but the drive will last longer.
I’m not trying to be a jerk, but if you label a section “how much power does the hard drive use” then list the power. or the deltas from spun down. Don’t make me do math to figure it all out, or rewrite your headers.
Hard drive spun up 1w
Hard drive max seeking 3w
How much power does wifi use?
On, idle 1w
On, constant transmitting from memory buffer 4w
I’ve been coding for small mobile devices for a number of years now. On pocket PCs, Palm handhelds and mobile phones the hierarchy of power consumption puts the display in first place, not the CPU. While this observation is due in part to the nature of the hardware, it’s also due to the programming environment and the applications on them.
Applications spend the majority of their time waiting for user actions, and the OS knows that it can put the CPU into a low power mode. Properly written programs yield as much as possible. Poorly written games run in a tight update/paint cycle, which keeps the CPU at full utilization. The shortened battery life is obvious.
CPU power consumption is also minimized by the types of applications you see on these devices. You don’t have a handful of applications grinding at large amounts of data while a number of background processes do other useful things. Instead, background tasks like maintaining a connection to a wireless network are handled almost entirely by dedicated hardware. There’s effectively only one application at a time, and even then the user isn’t going to spend an hour straight using it. The use model lends itself to long idle cycles.
