AMD Athlon 64 (and Pentium)

[kony]

I think you are greatly overestimating the efficiency of the
onboard conversion. Just allowing for the 0.7 v. drop of the
rectifiers into 3.3 v. show 21% losses, for a max efficiency of
79%.

MOSFET voltage drops can be much lower than 0.7V, under .2V
IIRC.

I did see a figure of 93% "somewhere" but at the moment I
can't recall where/what/who/etc... maybe too vague to make
assumptions about, I may know more specifics in the future.

If I get a chance I'll look into this further, though I
don't think it's possible for the max to be as low as 79%.
For one thing the parts aren't 'sinked anywhere near well
enough to dissipate that much heat on modern systems/CPU,
not do that AND stay as (relatively) cool running as they
do.

And that is only forward drop, with no switching losses. If
the diodes are schottky, with an 0.3v drop, the max efficiency is
still no better than 90% and the reverse and transformer/inductor
losses still have to be accounted for, also the input switcher
losses.

I do intend to spend more time on digging up data but for
the moment I"ll concede that I can't (yet) factually support
the 93% figure, but that I'm MUCH more confident that over
85% is reasonable, so at 85% that's still only 6.8A total
for the aforementioned example CPU.

 

[Fitz]

Although the course in electronics is interesting, I have neither the
expertise nor training to interpret a lot of it. My approach is a little
more basic....Overkill. A PSU from a reputable manufacturer with a
reputation for quality and a good warranty. Enough power to meet any demand
I may place on it, and a check of a few reviews to verify the advertised
specs are met. Take into consideration noise levels and looks and make a
choice. I probably spend more money using this approach than I need to, but
so far it's worked.

If this was my system, I would opt for a more potent PSU (not necessarily
the 420W rating, but specifically the amperage values on the 12V rail...and
since dual +12V are available, I'd go that route). Not knowing the OP's
budget, I said I didn't know if it was adequate. Personally, I'd replace it.

Fitz

 

[CBFalconer]

MOSFET voltage drops can be much lower than 0.7V, under .2V IIRC.

I did see a figure of 93% "somewhere" but at the moment I
can't recall where/what/who/etc... maybe too vague to make
assumptions about, I may know more specifics in the future.

If I get a chance I'll look into this further, though I don't
think it's possible for the max to be as low as 79%. For one
thing the parts aren't 'sinked anywhere near well enough to
dissipate that much heat on modern systems/CPU, not do that
AND stay as (relatively) cool running as they do.


I do intend to spend more time on digging up data but for
the moment I"ll concede that I can't (yet) factually support
the 93% figure, but that I'm MUCH more confident that over
85% is reasonable, so at 85% that's still only 6.8A total
for the aforementioned example CPU.

I'm not quite as dogmatic as I was yesterday. Use of FETs for
switching into an inductor could avoid most of the diode drop
losses, getting down to the 0.2 to 0.3 v. area, possible for dc/dc
conversion of a single voltage. This could also allow those drops
to apply to the source, rather than the destination, voltage.
There is no getting around inductor losses, nor supervision
circuitry.