Power Supply

[Leadfoot]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

KJ wrote:

| i suppose you missed this one:
|
http://www.newegg.com/app/viewProductDesc.asp?description=17-103-438&depa=1
|
| only $250!
| 660W
|

No I didn't. But comparing Price differences between a BRAND PSU, that
actually delivers the advertised wattage, and can do so for extended
periods, without ripple, is unfair.

That 24$ 600w would be sold by most Brand companies as a 400w or 380w at
most. It cant realistically provide 600w, even the person using it said
as much, that they fell 'safe' about it provided he doesnt pull more
than the 450-500w he currently consumes.

Actually it may produce 600W I simply don't feel like adding 6-8 hard drives
to prove you wrong

 

[Wes Newell]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Wes Newell wrote:
|
| Sorry, but I would also say buy generic. But I would also say always buy
| bigger than what you think you need, generic or not. I'm not sure, but
| genereics may or may not have a bigger failure rate, but it isn't twice as
| high, and nowhere near the 3-6 times more you will pay for name brand.


Well, at least this one contains some good advice 'But I would also say
always buy bigger than what you think you need, generic or not'

I dont know where the 'twice as high' MTBF comes into play, I never made
that claim.

I didn't mean to imply that you stated anything about failure rates. I
was speaking in general.

And as for the 3-6 times figure, yes, you *could*
theoretically pay 6 times more for a Brand PSU vs a Generic.

I didn't mean to imply that you stated anything about failure rates. I
was speaking in general.

Like that $24 600w vs a *REAL* 550w like this:

|
http://www.newegg.com/app/ViewProdu...-103-918&catalog=58&manufactory=BROWSE&depa=1

for woo, $100 US, so 4x the cost of a cheap PSU, maybe twice the cost of
a middle-line.

Yep, and I can buy 6 of these 550W PSU's for $96 ($16 ea.)

http://www.mrtechus.com/55powsupretb1.html

 

[Creeping Stone]

Hi Leadfoot,
I wish you luck but a $25 600W PS is like putting a junk yard engine in an
Indy 500 car <s>. The power supply is the heart of a good system. One
should *never* compromise on the PS.

Whats the bad aspect, the nominal 600w retail tag, the $25 dollars or the
absense of brandedness

With some brands you get unreported benefits like pre-overrating of the
equipments spec., premium design work, deluxe component selection for cool
tight spec operation. Some make sure the wattage is ideally represented,
any legal PSU will certainly have been securely safety and reliability
checked -If its sold as a certain wattage -its design has had at least that
much power running through it in aggravated test conditions, Production
runs have been sampled and passed.. by authorities with stringent remit to
avoid disaster.

Heres a p'dox - Antec say they safety test their units for higher demands
than theyre sold for - well that means /from one outlook/ theyre sold as
being weaker - so while they are reported as being more/less powerful -
that claim only has to be tested by Antec. A reportedly superb 450w Antec
has only to be officialy tested for that much gross output/
P'doxicaly, even its generic 650w competition has been officialy tested
for full 650w -though that power may not come in the right proportions, the
cheaper - seemingly inferior PSU has *been certified* for greater
electrical demands than the pre-underrated premium brand )

So many PSUs are manufactured and sold and plugged into the mains, I think
the lack of computer fire epidemics strongly witness the effectiveness of
standard compulsory PSU quality regulation, which is not necessarily so
urgent as all the other low voltage parts in the computer that risk
~environmentaly~ less catastrophic failures.

 

[Creeping Stone]

*thanking posters for eloquently advocating POVs which I enjoy refactoring*

I should qualify this, 'without *ATX specification acceptable or BETTER*
ripple'

'ripple'
-a generalised coloquial term refering to sporadic temporary voltage
fluctations
-a compontent of PSU output noise~

Better motherboards, often bought to accomodate more powerful cpus and
components, have more substantial on board power regulation than cheaper
boards with primary role to create on-board voltage lines and secondarily
keep them stable through non-intentional supply conditions (from main psu).

To test:
Whatever ripple my generic 350w psu gave out when I accidently flicked its
power switch quickly on/off was handled adequately (no reboot) by the
slighlty below ASUS's best standard voltage Regulation of my mobo (-x
version) of a7v8x.
Perhaps thats not how a ripple can be induced.
I cant see any ripples in the logs, the only tiny fluctations noticable are
at the limits of asus probes measuring capabilities - my 12v can be
observed dithering over the sensors least significant bit -no more
observable 'ripple' than that. (dependant on temp of sensor maybe)

Ripples or reverb?/?resonnance could be present at any frequencies over the
monitors interval that the onboard monitoring couldnt see, I suppose.

I *blindly expect such noise in the main powersupply can be routinely
soaked up by on board regualtion as long as it doesnt include any very high
energy spikes -enough to make bottleneck components hot enough to blow or
degrade in the small unmonitorable moment of exposure.*not all blind folks r so unaware

| That 24$ 600w would be sold by most Brand companies as a 400w or 380w at
| most. It cant realistically provide 600w, even the person using it said
| as much, that they fell 'safe' about it provided he doesnt pull more
| than the 450-500w he currently consumes.
|

Hes paranoid that it would present a hazard to run the device at its safety
assured capacity -thats if he cant reference that capacity on a
label/manual/online support notes , but it I think its optimal to over spec
any psu by a fair amount (up to 2 twice powerful) when constructing a new
rig.
A large overrating margin will take care of the unknown underspeccing
premium of different 'luxury'/'status symbol' brands and allow
any PSU to run cooler, stabler than intended, even for elevated sustained
system loads.

| Please stop comparing the price of a 1965 Corvair to a 2004 BMW, they
| arent the same magnitude of quality.

hehe, that sounds bourgeois to me - To a man with always less to spend than
buy, deluxe labels are of no great use , economy is.
Its not having the best thing which counts, its having the requirement well
met -anything beyond is an imaginary extra.
(until all reqs met or something catches imagination)

Its okay to have a favourite brand, favourite things, specific
recommendations, practical pets; but to dump on plain old PSUs for the
plainess is ignorant of all the tech mentioned in these threads.
( I read Paul explain some of the innards too )
If you knew what really annoyed you about plain legal PSUs it might just be
the confusing labeling.

-if I can see a detailed lable, knowing the brand hardly helps.

 

[Jerry Wong]

Dear all,

I have one question:
On the same PC system and configuration, will the power consumption be
different if I use a 300Watt psu or a 400Watt psu?

I guess the power consumption is the same. Correct me if it is wrong.

Thank

 

[KJ]

the PSU itself will dissipate power (heat and mechanical from the fan(s), so
the simple answer is that the consumption will be different. the better
question is if the difference is of a concern to you.

most PSU's are only about 70-75% efficient, and often the cheaper units are
less efficient, so will waste a little more power than a higher quality
unit.

 

[DreamMaker]

the PSU itself will dissipate power (heat and mechanical from the fan(s), so
the simple answer is that the consumption will be different. the better
question is if the difference is of a concern to you.

most PSU's are only about 70-75% efficient, and often the cheaper units are
less efficient, so will waste a little more power than a higher quality
unit.

So to understand what you say, 20 to 25% of the current in the psu is
dissipated in heat, (normal maybe abit higher cause there is no work
done here, but the 120V to 12v and authers) resulting in a temperature
elevation. So less cheaper will dissipate less heat. In authers hands
my system will pull out, of the psu, always the same amount of current
to performe it's task. So what ever the psu you got 350w or higher the
result are the same, but the heat dissipation. Resulting in less noise
for bigger systems as for the low end systems...

Plus did you ask your self why a 2600xp baron core need about 39 amps
(64Watts) to get it work done ( at full load) vs a psu that cant
manage to give that amount of amps, (i would say todays psu work over
the spec rating) . Go see your CPus spec at the amd site (pdf format)
and see!
mostly all psu dont give that much of current...

http://www.amd.com/us-en/Processors/TechnicalResources/0,,30_182_739_3748,00.html
http://www.amd.com/us-en/Processors/TechnicalResources/0,,30_182_869_4348^7923,00.html

 

[Paul]

A modern motherboard will have an ATX power connector and a 2x2 12V power
connector for the processor. So, you need an ATX power supply to power it.
At a minimum, those two connectors have to be connected to the motherboard,
to make it go. The 2x2 (i.e. four pin) connector is used exclusively by
the processor, and there is a conversion circuit on the motherboard, that
converts the +12V from the ATX power supply, into the 1.500V or so, that
a processor might need.

The ATX power supply has six outputs. The consumption from the -5V and
=12V outputs is so low, that there is no need to gauge the consumption.

The +5VSB output on the supply is used for "keep alive" power. That
power is used to save the contents of the DRAM chips, while the computer
sleeps. It powers the LAN chip, in case a wake up packet is sent to the
computer. It can be used to keep USB and PS/2 keyboards and mice powered,
so they can be used to wake a sleeping computer. If you download the
manual for the motherboard you plan on buying, there will be some
estimates in there as to how much power might be required. These aays,
many supplies have 1.5 or 2A for this function, and you can control some
of the load on this supply, by changing the USBPWRxx or PS2 PWR header
jumpers.

The +3.3V and +5V outputs on most supplies will be in the 20A or so
range. There is generally not enough information on power consumption
of the various parts of the computer, to say how much is enough. But
I can tell you by the process of deduction, that the consumption won't
be too high.

At one time, the processor derived its power, by converting the +5V output
to the lower voltage needed by the processor. With the increased power
consumption of processors, this function has been moved to +12V, and that
is why, if you try to reuse an old ATX power supply, chances are the
output on +12V will be insufficient for a new motherboard and processor.

To work out the numbers for +12V, seeing as it is critical, I use

Processor
P4 3.2Ghz/FSB800/512KB cache = [email protected] = 8.4A@12V
Athlon 3200+/FSB400/512KB = [email protected] = 6.4A@12V
including 80% conversion efficiency, the required current is
10.5A or 8A for a top end P4 or Athlon respectively.
Hard drive
2A during spinup of the disk, 0.5A while sitting in Windows desktop.
Allow 0.5A for a CD. If you don't have a lot of drives,
don't worry about spinup current, and concentrate on idle current.
Fans
Allow 1 amp for case and CPU fans.
Video card
Low end video cards use no +12V. An Nvidia FX5900 or an ATI9800
have a separate +12V cable, and as the video cards draw up to 70W
when gaming, a maximum of 6 amps would be required. Unless you
are buying one of these, a lesser number is more appropriate.
Total = 15A for a basic system, with some margin. If buying a
video card that requires extensive cooling, this number is more
like 20A.

Now, a reality check. The last computer I measured (2.6GHz/800 P4
865GE Northbridge) needed a total of 55W while idling in the Windows
desktop, and 120W while gaming (wall power). So, you can see that the
calculation above is quite pessimistic. That system didn't have a video
card, as it used the builtin graphics, so consumption could rise by
another 70W, if the system had a decent gaming video card in it.

Here are some sample products. The first table is for some
PCpowerandcooling.com products, considered the cadillac of power
supplies. For my purposes, the 350ATX meets the minimum +12V current
I would be after, so that is the smallest supply I would buy. You'll
notice that the bigger supplies mostly increase the capacity of the
rails you don't care about, so the 425ATX I would purchase to run
a FX5900 or ATI9800 has got a lot more +3.3V than I would ever need.

VOLTAGE ---> +5V +12V +3.3V -5V -12V +5VSB

Turbocool 300ATX PFC 30A 12A 14A 0.3A 1A 2A +5 & +3.3<150W
Turbocool 300 Dell 30A 12A 14A 0.3A 1A 2A +5 & +3.3<150W
Turbocool 350ATX 32A 15A 28A 0.3A 0.8A 2A +5 & +3.3<215W
Turbocool 425ATX 40A 20A 40A 0.3A 1A 2A +5 & +3.3<300W
Turbocool 510ATX 40A 34A 30A 0.3A 2A 3A total<510W

Here are the Antec Truepower series (antec-inc.com) - one step below
a cadillac.

VOLTAGE +5V +12V +3.3V -5V -12V +5VSB

TRUE330 30A 17A 28A 0.5A 1.0A 2.0A
TRUE380 35A 18A 28A 0.5A 1.0A 2.0A
TRUE430 36A 20A 28A 0.5A 1.0A 2.0A
TRUE480 38A 22A 30A 1.5A 1.0A 2.0A
TRUE550 40A 24A 32A 0.5A 1.0A 2.0A

Here, even the True330 is enough for a basic system, and the True430
is enough for a FX5900/ATI9800 gamer system.

Use a similar comparison with bargain supplies. Gauge them by output
currents and not total power.

HTH,
Paul

Some refinements on the calculation. In researching the answer to
another post, I found some power numbers for a few video cards.
The ATI9700 is estimated to be drawing [email protected] and [email protected], which
means it doesn't in fact draw the majority of power from +12V.
This changes the numbers to 12V@15A for basic system, and maybe
12V17A for a higher end card - at least until I can find some
FX5950 numbers (they are possibly worse).

http://takaman.jp/psu_calc.html?english (a power calculator)

"Homie" (the cap repairman) and day to day measurements from his PS:

http://groups.google.com/[email protected]

Measured board total power consumption - no breakdown on currents
(unless you pay a fee and download their complete results ???)

http://www.tecchannel.de/hardware/907/images/0011358_PIC.gif
http://www.tecchannel.de/hardware/907/images/0011359_PIC.gif

Measurement of a 9700:

http://forums.viaarena.com/messageview.cfm?catid=31&threadid=48125&FTVAR_MSGDBTABLE=

HTH,
Paul

 

[Paul]

i suppose you missed this one:
http://www.newegg.com/app/viewProductDesc.asp?description=17-103-438&depa=1

only $250!
660W

Now, that is an excellent example of the principles of power
supply selection.

+3.3V@ min 1.5 max 38A,
+5V 5 42A
+12V 14A
-12V 0.8A
+5VSB 2.7A

Notice how the +12V is barely adequate for a basic install.
The other rails are strong enough to use for arc welding.

So, you cannot judge this supply by price (you pay $250 but
can barely run a basic system). You cannot judge by total
power (660W and it won't have enough power to run a large
number of disk drives).

This supply is from a previous generation, where processor
power conversion is based on +5V. It would have easily run
a dual board from that generation.

HTH,
Paul