I didn’t really post these up before, didn’t consider them relevant, and was too busy getting a licence to drive to actually run the tests.
But today, out of curiosity, among others, I decided to determine what uses how much power, and found some rather interesting results.
We did the testing by finding the base load, that is, the load of all the devices that would occur with no systems or monitors turned on, and we used that as “Base Load”, and simply calculated the draw after the base load.
We have 3 machines running and they have various tasks.
My point at the wall (where we measured), showed 38W with the two systems and two LCD monitors turned off.
So, base load was 38W.
I then turned the LCD monitors on, one and then the other, and found they draw 22 – 24W of power. Pretty good.
I then turned my system on after turning the LCD monitors off. Unfortunately, I have a flaky Linksys router, and its constant reboot loop caused a result between 68 and 80W. I called the system at 75W, the extra 5W is negligble. That means, my system uses 37W.
So, for my complete system, it now uses 85W when it is turned on. Not bad!
Actually, I think I question these results. They don’t seem right when compared to my partners machine.
I then decided to determine the servers load. After turning all off, I had a base load of 41W (3W higher than what was happening with my tests, Linksys router again). The server pushed the meter to 154W, giving an idle load usage (that is, after it had booted and no hdd activity was occuring). This gives a usage of 113W continuously, 24 hours a day, even when I’m asleep. I think we’ll put a fix to that.
Next on the chopping block was the reason I bothered starting testing, CRT compared to LCD power results.
Since my partners new machine still has CRTs connected, we ran the tests.
Her power points base load was 23W.
Her system when powered on, and idle sparked up 110W, bringing her System in at 87W. She’s got a near identical machine to mine, with one extra HDD being the only difference, so that was surprising, perhaps her video card was the blame for the minor difference as well?
Anyway, we moved on to test the CRT monitors. I figured the difference wouldn’t be too great.
Turned one CRT monitor on and we got from 110W to 158W, meaning that it used 48W for one CRT.
Turned the other CRT on and we got to 205W (we peaked at 210W), meaning that it had used 47W more.
That’s incredible, so each time she’s using her computer, the draw is similar to that of a 100W light globe being on, plus the additional 80W her system will consume on.
Amazing to say the least. I think the ideal solution to our power consumption is to reduce the usage wherever possible.
I’m pushing for her to get LCDs, but my little one likes to play games and so forth, and she is concerned he might attack the monitors. My solution to that today, was to perhaps add perspex to them, however, consider this, LCDs cost around $200 each. The average expected life from a monitor for me at least, is 5 years.
The power savings off 48W over around 10 hours a day is not equal to the $200 each monitor will cost. So, the upgrade will end up costing more.
The servers results are surprising, considering it doesn’t have a monitor at all! I think its high reading is simply because its got a pesky Prescott in it, notorious for running Hot, and that heat is seemingly due to poor use of energy at the core. Causing it to piss away more power.
A dual core (that is 2 x CPUs in one) uses less than a Prescott (a single CPU).
Upgrading the server would probably have a future looking lifespan of 3 years or so from now. Costs more to upgrade then it does to simply run it for a few more years longer until its time to upgrade.
Anyway, the best power consumption in a PC you can get is what we have in the dual core machines.
Probably could even lower the results.
Use a single HDD instead of 2 or 3. Don’t plug in as much RAM as I have. Get boards with onboard video if you don’t game.
But the savings in power at that point are small, you are dealing with litterally up to 10-20W.
Combined load at peak reaches 300W on one point, and 210W on the other, meaning we can potentially suck 310W, or even more if the systems are worked hard. They aren’t all that often though, meaning they use less power.
Recently, I’ve added policies to machines to reduce further consumption where it seems good. So hopefully the results translate into a better power bill.