In preparation for PSU reviews soon to come, I did some prelim testing of a power draw meter I devised. This gizmo allows the AC current drawn by a system to be measured, from which the total power (watts) can be easily calculated. <BR> <BR>The system: <BR>CPU - AMD XP1700+ <BR>motherboard - ABIT KT7A-R <BR>video - ATI Radeon VE 32 mb <BR>ram - 256 mb PC133 sdram <BR>floppy disk drive <BR>4x 2x 24 CD-RW <BR>Seagate Barracuda IV 20G drive <BR> <BR>The surprise is how low the actual power consumption is. <BR> <BR>Max AC power consumption (100% CPU usage, full video, all drives working): 98W <BR>Min AC power consumption (idle, monitor on): 70W <BR> <BR>The numbers shocked me. I repeated the test several times with the same result. There could be errors due to contact and cable resistance, but it's doubtful these would be more than 10%. Even if the numbers are 20% higher, it's far less than what I expected to see. One wonders, then, about how much power these >300W PSU are REALLY capable of delivering...
[addsig]

Well, the maximum dissipation of an XP1700+ is about 60W. That would leave 40W for everything else. That doesn't sound out of the question for the components you're using as long as you're not writing CDs. I'm a little surprised that the low side isn't less, though. Everything except the disk drives is pretty much stone cold in my P4 2.0GHz Northwood system when it's idle. <br> <br>It does make one wonder about the need for 400W power supplies, it's true. There are power peaks that you may not be seeing with your meter (drives starting and such), but still... <br> <br>Having said that, I'll have to say that my ancient 275W PC Power and Cooling Silencer power supply doesn't sound very happy in this new system. I'd calculated that there was more than enough power available for a system with 2 Barracuda IVs, a DVD drive, a GB of memory and the CPU, but it sure doesn't sound like it when the load cranks up, even though I'm probably not much over the 100W figure you mentioned. <br>
[addsig]

ps: It's a little scary to be tagged a "silent apprentice" when I've been doing this quiet computer thing for 10 years now... <br>
[addsig]

Sorry about the apprentice tag <IMG SRC="modules/phpBB_14/images/smiles/icon_rolleyes.gif"> Tell you what: submit a great article for publication here & I'll be happy to promote you to honorary master <IMG SRC="modules/phpBB_14/images/smiles/icon_wink.gif"> <br> <br>BTW, I have to report that, as suspected, my numbers were a bit off -- they were about 20% low. Our lab advisor guru Tommmy Yee re-examined the setup & told me what I was doing wrong... Sorry if I mislead anyone. <br> <br>One thing I seem to have missed in my original post: The PSU is an Enermax 350W: EG365-VE(FCA). <br> <br>These are the correct figures, and to address your comment, the high number is simply the highest seen on the meter -- a peak, not sustained. The sustained high is about 110W. But as you say, it is possible even shorter ones the meter doesn't display could be happening. <br> <br>Max: 120W <br>Min: 86W <br> <br>They numbers are still far lower than the general 300W minimum recommended PSU for AMD systems. What it tells me is that the regulation of typical PSUs on the various is poor, and a high load on one voltage output likely affects stability in another, which causes mysterious system instability people often blame on other stuff -- like Windows (a great scapegoat, though).
[addsig]

I'll see what I can do about an article. Do you have a list of things you've got in mind? <br> <br>It occurs to me that my "doesn't sound out of the question" comment didn't allow for the loss through the power supply (20% to 30% or so). That fits with the 120W figure if one looks at the major power sinks (CPU, drives and video card). <br> <br>There's something that bothers me, though. It's beside the point of this study, but I really don't understand where the 86W minimum figure is coming from. Is there power management in the OS? It's a little hard to believe that's as low as it gets given the idle temperature on my system (running W2K Professional). If you ever get a chance to measure individual components, I'd be curious to see the result. <br>
[addsig]

<!-- BBCode Quote Start --><TABLE BORDER=0 ALIGN=CENTER WIDTH=85%><TR><TD><font class="pn-sub">Quote:</font><HR></TD></TR><TR><TD><FONT class="pn-sub"><BLOCKQUOTE>I'll see what I can do about an article. Do you have a list of things you've got in mind? <br> <br>It occurs to me that my "doesn't sound out of the question" comment didn't allow for the loss through the power supply (20% to 30% or so). That fits with the 120W figure if one looks at the major power sinks (CPU, drives and video card). <br> <br>There's something that bothers me, though. It's beside the point of this study, but I really don't understand where the 86W minimum figure is coming from. Is there power management in the OS? It's a little hard to believe that's as low as it gets given the idle temperature on my system (running W2K Professional). If you ever get a chance to measure individual components, I'd be curious to see the result.</BLOCKQUOTE></FONT></TD></TR><TR><TD><HR></TD></TR></TABLE><!-- BBCode Quote End --> <br>No list, but I suggest something you know well, or have tried, and works. There must be something you're eager to share with us. <br> <br>The 86W question is a good one. (Note: when the video card turns off [power saving], the min drops to 82W.) I looked up the efficiency spec of the Enermax: it is given as "70% at full load." That suggests that if the system is drawing 86W from the AC, then the components are drawing 57W. Subtract a couple watts for losses to contact & cable resistance. AMD's spec on typical current draw for the XP1700+ is 32.8A. Multiply by voltage (1.75V) to get power -- amazingly, the result is 57.4W... <br> <br>That, I think is just coincidence, as the "typical" rating is typical usage, not idle. At idle, I am sure the XP draws less power.
[addsig]

There's also the memory, motherboard chipset, sound card (an amazing power sink for what it does), the hard disk, the CD writer, ... So it seems like a coincidence. <br> <br>We're now wandering incredibly far afield, but you might be amused at this: <br> <br><!-- BBCode Start --><A HREF="http://eetd.lbl.gov/EA/Reports/46212/" TARGET="_blank">Energy Use of US Consumer Electronics at the end of the 20th Century</A><!-- BBCode End --> <br> <br>It shows average computer power requirements of 125W active and 80W idle. <br> <br>I may be able to borrow a power monitor for a few days. If so, I'll post some information on my system. <br>
[addsig]

That Energy Use doc is interesting. The numbers look like a close match till you realize they're including monitors.
[addsig]

Why AMD reccomends a 300w PSU is because the rails are lower powered.

See the page 11 spinup power requirements. It's a lot more than idle or running.

http://www.seagate.com/support/disc/man ... 7200pm.pdf

The thing is, the system is filled with low end compotents.
Care to replace the motherboard and add DDR ram as well as a not-so-outdated video card and you'll quickly add 60w!

Then again, it's an ok system for normal computing tasks (mail, surfing etc).
Still .. my EPIA consumes less .

OK Gandalf -- here is an update with a more powerful system, with AC power measured using Kill-a-Watt:

P4-2.8 @ stock volatge/clock
AOpen AX4GE Max mobo
512mb PC2700 DDR RAM
GF4 4800 VGA 128 mb
1 SCSI PCI card (for scanner)
3 Panaflo fans
1 CD-ROM
1 CDRW (Yamaha 40x)
2 Barracuda Seagate IV singl-platter HDD
Seasonic 300FS-APFC

max AC power during boot: 135W
AC power folding @ home: 137W
max AC power CPU burn: 159W -- No other combination of drive access, CD burning, MP3 video playing, scanning a doc w/acrobat, etc, could top this power consumption figure.

Assuming 70% efficiency for PSU, the DC power delivered to the components is: 94.5W, 95.9W, 111W.

Still amazingly low, eh?

remember that the power supplys are rated at 300/400/500 Watts MAX - This is not a continuously rated supply. They do not state the maximum continuous load. They generally over-rate much more on the cheaper power supplies.

quality goes a long way with power supplies, although there are only a few completely different ATX PSU designs, avoid the cheaper ones at all costs.

Mike if you or anyone gets a chance and are interested, I've always wanted to know the difference between power consumption using windows is idling, and when windows is idling using cpuidle or vcool or other software cooler.

I have a guess that using the machine as much as I do would save about $25 a year in electricity using s software cooler - 40w x 16 hours x 360 days x $0.10kwh

Just tried it after seeing your post. With and without CPU idle (with the system just idling) there is no difference in what Kill-a-Watt registers on the aformentioned P4-2.8 system. 93-94W AC draw.

Thanks for checking. I would guess that cpuidle also did not lower the temps? It works on most amd boards but maybe not intel.

Mike, do you have access to an EPIA board you can do some power measurements on? This recent Dan's Data review had some scathing things to say about the power consumption of the platform:


One of his claims is that the 1GHz C3 CPU alone was responsible for 20W jump in power consumption when taken from idle to full load: "So much for that ten watt CPU power figure". After reading that review, I wouldn't mind seeing a more comprehensive test in this area.

Seaker, There were some detailed power dissipation notes in the M10000 review I did some months ago. Bottom of the linked page:

I don't have an EPIA system set up & running right now. I don't recall seeing anything like a 20W jump due to CPU power for any EPIA board. They don't actually claim 10W max tho; there's little reason to doubt their max 15W (?) power rating for the 1GHz Nehemiah core.

The last set of power tests I did on an EPIA M1000 system showed absolute max AC power draw at 57W, which means DC power was ~34W. Curiously, this was while running Office benchmarks. In idle, that system pulled 32W AC or ~19W DC.

What sort of "gizmo" are you using? I find that your common garden variety multimeter on the 10 or 20a setting works just fine for measuring current with little if any chance of mistake and no calibration required. Also lets you measure the voltage instead of assuming it is 110 or some other value.

Just under 100w for the system you described is not out of the question, especially with your description of the synthetic load you provided "100% CPU usage, full video, all drives working". A program like prime95 or memcheck86 will do a better job of loading the processor to near its capacity. Multitasking that processor to try to keep the drives busy as well as play a video may show 100% cpu utilization (actually, 0 processor available, a subtle but significant difference), your processor is actually spending a lot of time waiting for main memory and not generating much heat. That and the whatever program you are using to keep the drives in constantly seeking end-to-end is not going to do much but wait for the PCI device.

While knowing that your system is generating 100-150 watts of heat in operation is helpful in designing a quiet cooling system, it doesn't do a whole lot in helping you in power supply selection. For that you need to measure the current on each of the rails and select a supply with adequate capacity on each. A 300 watt supply rated at 10 amps of 5v and 20 amps of 12v won't run your system very well since abit derives vcore for the processor from +5 on that motherboard whereas a 200 watt supply rated at 20 amps of 5v and only 15 amps of 12v will run it just fine. That same 300 watt supply with an over abundance of 12v capacity would run a different motherboard that derives vcore from 12v just fine.

good luck on your project.

A multimeter was the basis of my first gizmo. The current device I use is Kill-A-Watt. Search the web. Gives very accurate results on AC power draw. DC power delivery depends on AC/DC conversion efficiency, which varies from model to model and power level. The most demanding load is usually CPUBurn, which results in the highest measured AC power.


I disagree. Doubling the peak output of your system and using that as your guide for PSU power rating (honest rating) will be a perfetly good, conservative way to choose. You really don't have to get that scientific or specific. As others have said, when in doubt, buy gigger but make sure you're really getting bigger (rather than getting taken by advertising...).

That's still quite amazing yeah ..
I'll test my PC in a minute .

im at both ends ot the spectrum, via 866, and tbred at 2.3