Folding PSU Comparison

A forum just for SPCR's folding team... by request.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
mas92264
Patron of SPCR
Posts: 659
Joined: Fri Sep 26, 2003 5:26 pm
Location: Palm Springs, CA, USA

Folding PSU Comparison

Post by mas92264 » Sun Jun 13, 2004 10:58 am

Biostar M7NCG400, ver 7.0
Using igp
2400+ Thorton at 140 * 15, 1.65 vcore
Arctic Cooling Copper Silent TC
80mm Panaflo L blowing on nb hs
1x256 ram
Maxtor DM+8 30g hd, no fd or cd
Win 2K SP4
Ambient temp 27*C (80*F)
All components uncased

Code: Select all

                   Zalman ZM300A-APF  Seasonic SS-300FB  Seasonic SS-300FS  Fortron FSP300-60PN
                     Watts   Amps       Watts   Amps       Watts   Amps       Watts   Amps

Idle                   61     .53         49     .42         49     .74         62     .93  

Folding               117    1.01        101     .87        100    1.42        117    1.63 
Foxconn 865M01-G-6LS
Using igp
P4 2.6C, oem speed
Swiftech MCX478-V HS
PCP&C 92mm 1,600 rpm hs fan
80mm Panaflo L blowing on nb hs
1x256 ram
7200.7 80/8mb hd, no fd or cd
Win 2K SP4
All components uncased

Code: Select all

                                      Seasonic SS-300FB
                                        Watts   Amps

Idle                                      43     .38                                      

Folding 1                                 88     .76 

Folding 2 with ht                         94     .80

Unplug hd power, at idle                  37     .33

Data from Kill A Watt - p3international.com. Zalman 300A and Seasonic SS-300Fb (ver A2) are active PFC which appears to be the reason for the disparity between the watts/amps ratio above. An explanation of power factor. A layman's (like me) explanation would be that the electrical design of the non pfc psu's actually cause power to flow back into the power line during the ac cycle from positive to negative. It seems to follow that the amps shown on the Kill A Watt are the peak amount of power used and the watt usage reflects this flow-back of power. Whether the watts value is calculated by the Kill A Watt or determined some other way, I have no idea. If it all sounds a little like witchcraft, well, it does to me too.

As a practical matter, I'm not sure how much active pfc matters, other than the more advanced psu's from Seasonic have a clear advantage in lower power use. However, for now anyway, it doesn't appear that the advantage is so great that trash canning an older psu in favor of an active pfc psu is indicated.

All tests were run twice over a time span of about 3 hours. The second tests resulted in differences of zero to about 2%, perhaps because of a slight temperature change. The 80mm nb fan sits on top of the pci slots and is sort of aimed at the nb hs to cool it. The motherboards are either sitting on top of a cardboard box or piece of plywood and with practically no airflow over the nb hs's, I was getting daily lockups until I added the fans. Note that unplugging the fan, with the Kill A Watt connected, no noticeable reduction in power use could be detected.

My previous postings of amps using a digital multimeter turned out to be about 10% less than the same setup measured by the Kill A Watt. The previously posted watt usage, calculated by amps * volts was way off as my calculations didn't consider "power factor" as there was no practical way to measure it with a dmm. Hey, we're all learning here. :)

M

Edited for clarity
Last edited by mas92264 on Sun Jun 13, 2004 12:42 pm, edited 1 time in total.

peteamer
*Lifetime Patron*
Posts: 1740
Joined: Sun Dec 21, 2003 11:24 am
Location: 'Sunny' Cornwall U.K.

Post by peteamer » Sun Jun 13, 2004 12:27 pm

Nice one mas, thanks for your time and input, that's very interesting.

Hopefully someone with a more technical side than I can explain the watts/amps disparity. Like you said, it sure don't add up.

Though I do know some one who had a device he connected between his incoming mains supply and the meter box which enabled the meter to run backwards. He had to be carefull to not overdo it mind but I seem to recall that using a principle, that I think we're seeing with your results. Not that he used it what with it being illegal and all...

Wish I could remember more, I seem to recall it had something to do with phase change... 90 Degs out of phase rings a bell... so does power supply companies not being happy about certain power supplies that we use because they use more electricity than shows up on the meter and effectively steal electricity.. can't remember any more though..

Ther, hope that's cleared it up a little for everyone... :roll:

Nice one for what you've done Mas.


Pete

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Sun Jun 13, 2004 3:01 pm

I'm just confused!

What does my electric meter register; the reading given by the Kil-O-Watt, or the lower measurement produced via the volts * amps method? For folding, this is my only concern.

If the kil-o-watt isn't measuring what's being consumed at the meter then I wouldn't advise using it to evaluate different systems for cost/wu. The same goes for the volts * amps method I use.

ps, what happened to the 40 watt difference you initailly reported?

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Sun Jun 13, 2004 3:48 pm

I found this a little easier to read:

http://www.dansdata.com/gz028.htm

Apparently if the kill A Watt is measuring real power used rather than calculating from volts * amps it will give a truer reading. They make reference in the article that uncorrected power factor will throw off watt measurements via the volts * amps method. So, I guess anyway, if the power factor correction isn't perfect then using a multimeter will give erroneous watt measurements.

That sucks, 'cause I have enough things to buy!

CharlieChan
Patron of SPCR
Posts: 198
Joined: Sun Jul 13, 2003 2:57 am
Location: East Anglia, UK

Post by CharlieChan » Sun Jun 13, 2004 3:58 pm

Copper wrote:I'm just confused!
P = IV cos(theta) where cos theta is the power factor.

You need to know the phase relationship between the current and voltage in a AC circuit before you can workout the power. Use the Kill-A-Watt meter :D

mas92264
Patron of SPCR
Posts: 659
Joined: Fri Sep 26, 2003 5:26 pm
Location: Palm Springs, CA, USA

Post by mas92264 » Sun Jun 13, 2004 4:27 pm

Actually, I should have given MikeC credt for my first understanding of PFC as he mentions it in one or more of his power supply reviews. I couldn't find it this a.m., but, will try and link to it.

Got my Kill A Watt at J&R for $35 delivered. Not a plug for them, as I rarely buy anything there. Was the cheapest I could find. :)

And yes, adding to what CharlieChan said, it seems that the Kill A Watt meter "watts" is a reasonable reflection of what the power meter is recording. Good link to the Dansdata article. :)

M

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Sun Jun 13, 2004 4:29 pm

That's easy enough!! :D

mas92264
Patron of SPCR
Posts: 659
Joined: Fri Sep 26, 2003 5:26 pm
Location: Palm Springs, CA, USA

Post by mas92264 » Sun Jun 13, 2004 4:43 pm

ps, what happened to the 40 watt difference you initailly reported?
I was measuring the amps drawn on a board powered by a Shuttle XPC 200W (Achme) psu with active pfc. I was getting some really low amp numbers. Turns out that using the Kill A Watt, it's about + 10% of the SS-300FB watts numbers. Still pretty good, tho. I using it now to power the Biostar board.

M

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Sun Jun 13, 2004 5:04 pm

I guess I better get a Kill A Watt before I go replacing any of the hardware I've purchased. Much of the difference between the two systems may disappear with the measuring tool alone.

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Wed Jun 23, 2004 5:38 pm

The Kill A Watt arrived today. I made two measurements while having both Zens folding dual instances, first overclocked then not. In one I have a 3.0C undervolted to 1.4 Vcore. In the other I have a 3.06, first overclocked to 3.4 at 1.55 Vcore and then at the default clock and undervolted to 1.35 Vcore.

Default: 160 watts
Overclocked: 195 watts

Both measurements are with everything used by the computer plugged into a single powerstrip, with the powerstrip plugged into the Kill A Watt, and the monitor off.

A very expensive recipe for folding, but I decided I'd rather have the money sitting in hardware rather than give it away over time to the power company. I have another 3.06 that shold arrive tomorrow. I was going to overclock it, but I think I'll take advantage of it's good undervolting and take it easy on the electric bill. That overclocking adds a lot more juice than it returns in points.

mas92264
Patron of SPCR
Posts: 659
Joined: Fri Sep 26, 2003 5:26 pm
Location: Palm Springs, CA, USA

Post by mas92264 » Wed Jun 23, 2004 8:47 pm

Default: 160 watts
Overclocked: 195 watts
Yikes :!:

Just plugged my kill a wat in my SB62G2 - 2.8C at 214*14, 7200.7 200g, 512mb ram, igp, SilentX 250w psu (active pfc,) Win XP.

Consumption in watts:

Code: Select all

Idle        58
Folding 1  114
Folding 2  127  with ht
M

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Thu Jun 24, 2004 4:16 am

I bet we're pretty close. I only measured with both computer's running at the same time though. Overclocking just the one CPU made a 35 watt difference to what both computers were consuming while running simultaniously.

mas92264
Patron of SPCR
Posts: 659
Joined: Fri Sep 26, 2003 5:26 pm
Location: Palm Springs, CA, USA

Post by mas92264 » Sun Jun 27, 2004 8:41 pm

Oops. Didn't realize that you were running 2 boxes. :) Those are good numbers for 2 boxes.

M

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Mon Jun 28, 2004 7:44 am

You probably noticed the other thread where I have the same boxes running 3.06 cpu's slightly underclocked to 2.56. The underclock allows a nice undervolt, bringing the total combined watts down to 106.

The overclock is what seems to be the most penalizing, producing a much higher percentage increase in watts consumed than in ghz gained. But the alternative, buying more or faster hardware to get the increased ghz, is mighty expensive.

I bet if you ran your 2.8 at stock speed and undervolted as low as you could, and still be stable, your watts would be well under 100... I bet as low as 80. But do you realy want to loose the extra hz.

mas92264
Patron of SPCR
Posts: 659
Joined: Fri Sep 26, 2003 5:26 pm
Location: Palm Springs, CA, USA

Post by mas92264 » Mon Jun 28, 2004 11:04 am

I should have said "those are excellent numbers for 2 boxes."

Yep, I'm following the info in your other thread. Nice work. Just for fun, I'm gonna lower the fsb to 200 on my 2.8c and see what the results are.

M

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Mon Jun 28, 2004 6:12 pm

I OC'ed both of the 3.06's to take another watt reading and I get 228 watts to power two boxes, or approx 114 watts each. OC is 3.4 Ghz.

You can see the difference from my underclock to my overclock:

5.12 Ghz to 6.8 Ghz is an increase of 32.81%

106 watts to 228 watts is an increase of 115.09%

If we bought hardware and doubled my current underclocked set up it would put out 10.24 Ghz and consume only 212 watts, still less than the overclock. Undervoltable motherboards are worth their weight in gold to the folder.

Edited to add, my clocking and Vcore experience is limited to P4's. I don't know how things pan out with AMD's, though, I suspect they are pretty simular.

Edited yet again to correct the total watts, and thus percentages. I forgot to shut the monitor off before I took the reading.

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Wed Jun 30, 2004 4:59 pm

Even that nasty Prescott can be tamed with a little underclock and undervolt.

3.0E @ 2.49ghz & 1.0625 Vcore:

Total of two boxes (one 3.06 @ 2.56 ghz & one 3.0E @ 2.49ghz) = 118 watts, up 12 watts from dual 3.06 @ 2.56 ghz.

Time to try the 3.0C with the same underclock as the Prescott.

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Thu Jul 01, 2004 4:45 pm

The 3.0C is doing well at 2.49 ghz. The two boxes together are pulling just 108 watts, up only 2 watts.

Copper
*Lifetime Patron*
Posts: 587
Joined: Sun Feb 22, 2004 8:43 am

Post by Copper » Sat Jul 03, 2004 1:51 pm

I have the Intel board up and running again with a 3.06 in it and powered with the Coolmax 350. This time measuring with the Kill A Watt I get 119 watts while folding, instead of 165 with the amp meter.

Looking at the 127 watts you have with the OC'ed 2.8 I'd say you aren't doing to bad at all.

Post Reply