I found the thread that I saved from the time I asked about PSU's on a hardware forum. I cut and pasted some of the thread to a text file, which I am now copying below. The guy named Paul is the expert who answered my original question about PSU's. In his longer, second reply below, he explains the things that need to be looked for if you really want to choose your PSU in the "correct" manner and not just go by overall wattage rating. It makes for some interesting background information, at any rate.
> On Sun, 09 Jul 2006 13:23:12 -0400, Paul wrote:
>
>
> >
> > A "darn good" video card, could draw from 4 to 10 amps from
> > the 12V rail. Naming the card might make a difference to the
> > recommendation.
> >
> > Your processor is 8.25A from +12V. If the supply has 12V@18A,
> > the best video card money can buy, puts you over the limit.
> > Your other periphs might amount to 3 amps, and I'd want another
> > 3 amps of margin. With a 4 amp video card, the 12V@18A is
> > a good minimal choice.
> >
> > The more solid info you put in your hardware inventory, the
> > more accurate the answer.
> >
> > The selection process is complicated by whether the chosen
> > power supply, is an ATX 2.0+ spec supply, or an earlier one.
> > The ATX 2.0+ spec, splits the 12V rail into two circuits.
> > Generally, that is more wasteful, as 12V2 only powers the
> > processor, and any excess capacity is not available to be
> > used elsewhere. So that is another factor that has to be
> > taken into account - splitting the computed load over
> > 12V1 and 12V2, in the case of the more modern version
> > of power supply (typically with a 24 pin connector).
> > While some supplies may cheat, and use one 12V circuit to
> > power both 12V1 and 12V2, the power supply choice can only
> > be made by the specs the manufacturer provides, and the
> > assumption is, that 12V1 and 12V2 are separate circuits.
> >
> > Paul
>
> Hi Paul
>
> thanks for your very detailed reply. However, as you can see from the
> information that my original post is lacking, I am definitely not up on
> the technology issues as much as you are.
>
> To answer your question, I have not decided yet on which graphic card to
> buy, but am sure I will only include 1 and not 2 graphic cards in the PC.
> Whatever power supply and graphic card I choose, I would definitely like
> to have a little more power available in the event of an upgrade to a
> graphic card later. So the PSU definitely should be overdimensioned a bit
> from the start. But what is enough? 430 Watts, as the Thermotake that
> another poster suggested? Given the general description I provided, it
> should be possible to put a number on the PSU size in Watts that I can
> safely go for.
>
> I thank you very much for the time you took to give such a detailed
> description, but I just have no understanding of the sizing in terms of
> volts and amps. Since I am purchasing a new PSU and not re-using an older
> one, I also wonder if I need to worry so much about how the power is
> divided into various cables. I would imagine that a new PSU produced from
> a reputable manufacturer would arrange the power in the way which is
> optimal for today's equipment. Or am I wrong? If I, for example, choose an
> Antec or Thermotake of a certain Watt power, do I also need to dig into
> the specs to make sure the power is split or not split correctly? I must
> admit I did not understand much of what you wrote.
>
> Thanks again!
>
> Regards,
> Steve, Denmark
The power supply has a label on the side. It specifies how much
power can be extracted from the outputs. The supply will have
+3.3V, +5V, +12V, as the major supplies, and each one has an
ampere limit. Generally speaking, a lot of the supplies have
enough 3.3V and 5V (and it is not possible to calculate all
the loads for those anyway). The major power consuming loads
work with +12V.
The Newegg listings of power supplies, include all the
amps ratings, and that is a convenient way to get the
necessary information.
I've copies a couple tables of supplies from a previous
posting. The first group are Antec Truepower2 supplies.
The second groups are Seasonic high efficiency supplies.
I'm including these, to show that each brand does things
a bit different, making it necessary to shop by "amps"
and not "watts".
VOLTAGE Â +5V + 12V1 +12V2 +3.3V -12V +5VSB (Antec)
TPII 380W 35A Â 16A Â 16A Â 28A Â 1.0A Â 2A
TPII 430W 35A Â 17A Â 17A Â 28A Â 1.0A Â 2A
TPII 480W 38A Â 18A Â 18A Â Â 30A Â 1.0A Â 2A
TPII 550W 40A Â 19A Â 19A Â 32A Â 1.0A Â 2A (both normal and EPS12V version)
VOLTAGE --> 3.3V Â 5V Â 12V1 Â 12V2 Â -12V Â +5VSB (Seasonic)
S12-430 Â Â 30A Â 30A Â 14A Â 15A Â Â 0.8A Â Â 2A
S12-500 Â Â 30A Â 30A Â 17A Â 16A Â Â 0.8A Â Â 2A
S12-600 Â Â 30A Â 30A Â 18A Â 18A Â Â 0.8A Â Â 2A
First of all, the currents shown cannot all be drawn
simultaneously. But, what we know about the computer, is
one output really gets hammered, while the other two are
not at their limits. The way the computer works, aligns with
how the power supply is limited. The overall "power rating"
is just one of those limits, and there are also limits on
the total amount of +3.3V and +5V, for example.
You will notice in the two product families above, that
the 3.3V and 5V are close to 30A each capability. The last
motherboards I measured, used about 15A from one of those
two rails, so there is a good deal of margin there. The
controller boards on hard and optical drives draw 1A from
+5V each, so there is a little more current drawn there.
You will notice that the 12V1 and 12V2, only have a
modest expansion with the power rating of the supply.
So, when we say "I'll buy a 350W" or "I think a 450W is
enough", really we need to look at the amps instead.
Notice how the Seasonic 430W is only 14A and 15A. while
the Truepower2 is 17A and 17A. That is a significant
difference between them. Thus the watts are not a
very good barometer of "value".
Your processor connects to 12V2. You need around 8.25A
for the proposed processor. With a little margin, like
about 3 amps, anything with 12V2 of 11A or more, should be
OK. All the power supplies in the above table are good
enough for that.
Now, we need to work out 12V1. 12V1 is on the main ATX
24 pin power connector. It powers the fans on the motherboard.
It powers the video card slot. By means of a Molex or
PCI-E power connector, it may also power the auxiliary
connector on the end of the video card.
For a basic system, my 12V1 looks like
Fans 12V @ 0.75A (three fans, a good one on CPU)
Video card 12V @ 4A (6600GT)
or 12V @ 10A (X1900XTX)
Hard drive 12V @ 0.6A (idling hard drive)
CD/DVD 12V @ 1.5A (spinup media, or maybe burning)
Margin 3A
Total (w. 6600GT) 6.85A + 3A (12V1)
(w. X1900XTX) 12.85A + 3A (12V1)
With the X1900XTX, all of the Truepower supplies can manage
16A, so anything from a 380W - 550W TPII would work. On
the Seasonic, the 500W or 600W would be needed. Thus, you
can see that buying a Seasonic based on a "watts" recommendation,
might not give the desired results. That is why the amps
thing is so important.
If you bought a 6600 video card, or a 7600, about 10A on
12V1 provides a comfortable margin. Now, your cheesier brands
of supplies could be used. But again, you have to check
the amps on 12V1 and 12V2 to be sure. Using the Newegg
power supply adverts, you should be able to find a
12V1 @ 10A, 12V2 @ 11A supply, for a budget video card.
You can get a lot of video card power measurements here.
Divide the watts by 12V, to get amps. So a 60W video card
would need 5A or so.
www.xbitlabs.com/articles/video/display/gpu-consumption2006.htmlI hope that explains a bit why it is better to shop by
the output amps, rather than simply blindly grabbing a
bogus watts number. No two companies do their outputs
quite the same way, meaning one guy's watts aren't the
same as the next guy. (The worst example I can think of,
is a bargain supply 2 years ago. It was rated for a bit
more than 500W, but had a 12V@10A output. That amount of
power is not suited for any Pentium 4 based system, and
yet the "500W" sounds so impressive. I'm sure many of
those were bought by unsuspecting customers, and ended
up as door stops or boat anchors.)
HTH,
Paul