Help me diagnose this shit

[Ixarku]

he should need 485W and he has a 530W PSU

Honestly, when I saw this, it had me wondering if the PSU was getting overloaded even though the load was less than the rated maximum, though I didn't think the PSU was actually the problem.  So when building a machine, what's the tolerance that one should allow for on PSUs?  20% below the stated maximum?  Or does it depend on the quality of the PSU in question?

[rostheferret]

he should need 485W and he has a 530W PSU

Honestly, when I saw this, it had me wondering if the PSU was getting overloaded even though the load was less than the rated maximum, though I didn't think the PSU was actually the problem.  So when building a machine, what's the tolerance that one should allow for on PSUs?  20% below the stated maximum?  Or does it depend on the quality of the PSU in question?

I know that for peak efficiency 20% is a good margin to use. Else it seems to depend on the quality of the PSU. I've seen cheap 1500W ones which I bet suck balls. Just go with a reliable brand. I mean, they aren't expensive, and do you really want to quibble over a few quid when buying the one component capable of frying every other?

[Ixarku]

Yah, really.  If I'm going to spend the money and go to the trouble of building a PC myself, my philosophy is to spend money on quality parts, or don't bother with it at all.  I've personally had good experiences with Corsair PSUs, although I'm not as exacting about my system requirements as some people.

[buchno]

I know that for peak efficiency 20% is a good margin to use. Else it seems to depend on the quality of the PSU. I've seen cheap 1500W ones which I bet suck balls. Just go with a reliable brand. I mean, they aren't expensive, and do you really want to quibble over a few quid when buying the one component capable of frying every other? :P

I just added together 4 FX-8350 and 4 Geforce 690 in the PSU calculator, and they didn't reach 1500W.

Who needs such a PSU...?

[kitamesume]

Yah, really.  If I'm going to spend the money and go to the trouble of building a PC myself, my philosophy is to spend money on quality parts, or don't bother with it at all.  I've personally had good experiences with Corsair PSUs, although I'm not as exacting about my system requirements as some people.

not really, you have to compute by amperes not wattage, ignoring the 5v and 3.3v loads(they're too tiny to be of concern), you'll need to compute all the total wattage of components then divide it by 12v(this is to include all the loses induced since you're factoring in the efficiency with this setting), once you get your final amperage you'll need to distribute them to rails, load balancing as well.
you don't need to add much margin on top of the peak conditions, even on stress tests you barely see them go out of spec on their power consumptions right? at most only powerbugs can induce a rather disturbing power draw.

but in my opinion adding about 3-5A of margin just in case it  wouldn't hurt to add some more fans and harddrives.

[Pentium100]

is that 2x10A continuous or pulsed? because pulse rating is approximately 3-4times higher than continuous, and it should be obvious that a switching powersupply is pulsed with a certain duty cycle.

if it was a quality PSU, even if you overload one rail it shouldn't send off a magic smoke, the OCP sensor could get fried when repeatedly done though.

Continuous. Pulse rating is only applicable if the duty cycle is low (so the average current is not very high). If the output current is continuous 30A then the diode current will either be continuous 30A or pulsed, but much higher, so it would still overload the diode. Some diodes are actually rated in "average rectified forward current", but I am too lazy to find the datasheet of the failed diode.

And yes, if the PSU is high quality, it would shut off on overload. If it is cheap it will release the magic smoke or just blow up.

not really, you have to compute by amperes not wattage, ignoring the 5v and 3.3v loads(they're too tiny to be of concern), you'll need to compute all the total wattage of components then divide it by 12v(this is to include all the loses induced since you're factoring in the efficiency with this setting), once you get your final amperage you'll need to distribute them to rails, load balancing as well.

Also, if you have the choice, buy a single rail PSU instead of the multiple ones - no need to load balance.
If you want the power supply to be the most efficient, buy one that is twice as powerful as your load, because these power supplies are most efficient at around 50% load. Adding margin is also good because over time the power of the supply will drop because of aging capacitors. Higher margin is more time until you need to replace the caps.

I just added together 4 FX-8350 and 4 Geforce 690 in the PSU calculator, and they didn't reach 1500W.
Who needs such a PSU...?

People who run the computers at 1kW or more. Not running a component at maximum load increases its life and a switching PSU is most efficient at 50% load (so that 1.5kW PSU would be great for a computer that uses 750W).

[kitamesume]

I know that for peak efficiency 20% is a good margin to use. Else it seems to depend on the quality of the PSU. I've seen cheap 1500W ones which I bet suck balls. Just go with a reliable brand. I mean, they aren't expensive, and do you really want to quibble over a few quid when buying the one component capable of frying every other?

I just added together 4 FX-8350 and 4 Geforce 690 in the PSU calculator, and they didn't reach 1500W.

Who needs such a PSU...?

ah yes yes, i7 hexa core extreme OCed to over 4Ghz can consume a little over 200watts, pair that up with a dozen harddrives with a hefty quad HD7970GE(yes these things are so inefficient), set capacitor aging at 20% (roughly 3-5years down the road) and it will reach over 1.5KWH

(click to show/hide)

is that 2x10A continuous or pulsed? because pulse rating is approximately 3-4times higher than continuous, and it should be obvious that a switching powersupply is pulsed with a certain duty cycle.

if it was a quality PSU, even if you overload one rail it shouldn't send off a magic smoke, the OCP sensor could get fried when repeatedly done though.

Continuous. Pulse rating is only applicable if the duty cycle is low (so the average current is not very high). If the output current is continuous 30A then the diode current will either be continuous 30A or pulsed, but much higher, so it would still overload the diode. Some diodes are actually rated in "average rectified forward current", but I am too lazy to find the datasheet of the failed diode.

And yes, if the PSU is high quality, it would shut off on overload. If it is cheap it will release the magic smoke or just blow up.

depending on the diode usage, if the two diodes are in use at alternate load(eg each has 50% duty cycle) they can theoretically handle up to 20-30A(depends on the diode's specs) each.
but since its a cheap PSU, i won't be expecting a pretty diode anyway, most likely it's 60% duty cycle pulse rating is merely 20A or so.

not really, you have to compute by amperes not wattage, ignoring the 5v and 3.3v loads(they're too tiny to be of concern), you'll need to compute all the total wattage of components then divide it by 12v(this is to include all the loses induced since you're factoring in the efficiency with this setting), once you get your final amperage you'll need to distribute them to rails, load balancing as well.

Also, if you have the choice, buy a single rail PSU instead of the multiple ones - no need to load balance.
If you want the power supply to be the most efficient, buy one that is twice as powerful as your load, because these power supplies are most efficient at around 50% load. Adding margin is also good because over time the power of the supply will drop because of aging capacitors. Higher margin is more time until you need to replace the caps.

considering that the PC won't be hovering at 100% peak usage 90% of the time... well hitting a capacity target of 10-20% over the peak usage would be plenty, the PC would spend it's time at 50-60% power usage even during rough encoding, mostly benches are the only things that seems to push power consumption to peak power usage.

[Pentium100]

depending on the diode usage, if the two diodes are in use at alternate load(eg each has 50% duty cycle) they can theoretically handle up to 20-30A(depends on the diode's specs) each.
but since its a cheap PSU, i won't be expecting a pretty diode anyway, most likely it's 60% duty cycle pulse rating is merely 20A or so.

Well, if the two diodes are in use at alternate load (which they are - they make a full wave rectifier). each half can only supply 10A average. So, put them in parallel (or in alternate load) and you get 20A. 30A goes over the spec. The duty cycle does not matter because the shorter duty cycle will be the more current it will generate to produce the same output current (the inductor just averages the current, it does not step it up).

considering that the PC won't be hovering at 100% peak usage 90% of the time... well hitting a capacity target of 10-20% over the peak usage would be plenty, the PC would spend it's time at 50-60% power usage even during rough encoding, mostly benches are the only things that seems to push power consumption to peak power usage.

I do not like failed power supplies - they are time consuming to replace (especially if I tidied up the cables and especially if I do not have a spare of sufficient power) and can produce weird intermittent problems before failing. So, I'd rather buy a bigger PSU and have it be more reliable than save a few euros on the PSU now and curse a few years later. Besides, I never know when I'll decide to add a few hard drives or another VGA.

[kitamesume]

there is a technique to bypass the limits of the diodes, thats when you overlap the two duty cycles, the overlapped part would have theoretically double the current capacity. if lets say two 10A continuous when ran at 75% duty cycle would be able to handle 15A repetitive output, the overlapped max current would be 30A.
this would mean that 50% of the time the max current would be 15A while the other 50% would be 30A, with this you sacrificed 25% of continuous amperage for a boost of 50% for 50% of the time, although this causes power sags over the inductors and the imbalance in this can be seen in the harmonics of the output.
although as i've said, i wouldn't expect such a topology on a cheap powersupply.



no you misunderstood my claims, im not saying you should buy a lower capacity PSU, what im saying is that you don't need to get such overly exaggerated headroom for two main reasons.
1) your PC will only hit a certain peak power consumption during stress tests on all components. unless you stress test your PC 24/365 then i doubt you'll need anywhere near twice as much PSU capacity to hit the 50% mark where it is the best efficiency.
2) unless you're adding enthusiast cards in it you shouldn't even need anywhere near 300watts of headroom (i've yet to see one draw anywhere near 300watts is what i'm saying).

as for these two reasons, your PC would spend at 40-60% of it's max power consumption during heavy use(encoding and gaming), while it'll hover at 20% or even less during playback and surfing(unless people claim lies, hahaha) and less than 10% during idle.
if you had like 2x the max load then halve those power usage and you'll be seeing horrible efficiencies since you're practically under 30% of the PSU's capacity.
---
as for the durability of the PSU, capacity has nothing to do with it really, unless you're cutting close to the max PSU rating(PSU has it's own wriggle headroom, depending on manufacturer and model). what you'll need to consider about durability of the PSU is it's reliability, hey a cheapo 650watts like the Corsair VS650 wouldn't be as great as their Corsair TX650 right?

[mgz]

Yah, really.  If I'm going to spend the money and go to the trouble of building a PC myself, my philosophy is to spend money on quality parts, or don't bother with it at all.  I've personally had good experiences with Corsair PSUs, although I'm not as exacting about my system requirements as some people.

yea i switched to corsair for my last 4 PSU's and i love them. Have not had a corsair psu die yet had other brands die.
And the corsair units you can find really cheap and are solid performers.
currently have 4 corsair 750 supplies running 3 of them 24/7 for the past couple years

[buchno]

People who run the computers at 1kW or more. Not running a component at maximum load increases its life and a switching PSU is most efficient at 50% load (so that 1.5kW PSU would be great for a computer that uses 750W).

Ok, thought it was at about 70% load.

[Slysoft]

Yah, really.  If I'm going to spend the money and go to the trouble of building a PC myself, my philosophy is to spend money on quality parts, or don't bother with it at all.  I've personally had good experiences with Corsair PSUs, although I'm not as exacting about my system requirements as some people.

yea i switched to corsair for my last 4 PSU's and i love them. Have not had a corsair psu die yet had other brands die.
And the corsair units you can find really cheap and are solid performers.
currently have 4 corsair 750 supplies running 3 of them 24/7 for the past couple years

corsair tends to be solid no matter what it is. If only they made mobos and graphics cards I'd have a bulletproof PC

[Pentium100]

as for the durability of the PSU, capacity has nothing to do with it really, unless you're cutting close to the max PSU rating(PSU has it's own wriggle headroom, depending on manufacturer and model). what you'll need to consider about durability of the PSU is it's reliability, hey a cheapo 650watts like the Corsair VS650 wouldn't be as great as their Corsair TX650 right?

Yes, but a cheap 650W would be better than a cheap 300W. Similar with the TX series. As capacitors age, their ESR increases and so does the ripple at a certain output current. Bigger PSU = bigger/better/more capacitors = more time before their ESR gets high enough to not be able to support the actual load.

[kitamesume]

yes but it doesn't mean you need that much headroom, if you've done your math you'd see that if you factor in 30% capacitor aging you'll only need about 25% headroom(a 500watt setup would need 625-650watts).

although i've yet to see a table for aging%-per-year, i can't say how much years would it take to hit the 30%, but it shouldn't be hitting 20% capacitor aging at less than 3years.
and a fixed aging rate doesn't make sense either, otherwise you'll see PSU thats still working after 5years should be supplying near-nil of power.

as for an "overkill" psu, i don't think its a good idea either, unless you modulize your rig instead of swapping the whole thing that is.
for me though, i'm more in peace if i had most of the components as brand-new rather than having an overkill 10year-old PSU. as people says, nothing lasts forever.

[kureshii]

So what happened to OP friend’s system in the end

[kitamesume]

^ it was confirmed that its the PSU giving the headaches, its either a new one is being shipped in, or its already been swapped.

[Pentium100]

and a fixed aging rate doesn't make sense either, otherwise you'll see PSU thats still working after 5years should be supplying near-nil of power.

I had one. The 400W PSU could barely power a single hard drive. Some time later it couldn't. After I replaced the capacitors now it can supply the full power.

as for an "overkill" psu, i don't think its a good idea either, unless you modulize your rig instead of swapping the whole thing that is.
for me though, i'm more in peace if i had most of the components as brand-new rather than having an overkill 10year-old PSU. as people says, nothing lasts forever.

I do not replace good components. For example, I have a hard drive (15kRPM 36GB) that I took from my old main PC and placed in my new main PC when building it (no point in buying another hard drive that is exactly the same). Now my main PC also has a 146GB 15kRPM drive and when I'm building my new main PC I'll probably move those hard drives to it (or buy SAS 15kRPM drives, depending on their cost at the time).

I also tend to use computers for a long time, so that 10 year old overkill PSU might still be in use in the same system with me being very happy that I did not need to shut that system down because of PSU failure.

[Tatsujin]

Yah, really.  If I'm going to spend the money and go to the trouble of building a PC myself, my philosophy is to spend money on quality parts, or don't bother with it at all.  I've personally had good experiences with Corsair PSUs, although I'm not as exacting about my system requirements as some people.

Corsair is a very good PSU, I recommend to look at their HX and Enthusiast models.

[Ixarku]

Yah, really.  If I'm going to spend the money and go to the trouble of building a PC myself, my philosophy is to spend money on quality parts, or don't bother with it at all.  I've personally had good experiences with Corsair PSUs, although I'm not as exacting about my system requirements as some people.

Corsair is a very good PSU, I recommend to look at their HX and Enthusiast models.

I actually bought an HX850 when I built my current PC just about a year ago.

[kitamesume]

as for an "overkill" psu, i don't think its a good idea either, unless you modulize your rig instead of swapping the whole thing that is.
for me though, i'm more in peace if i had most of the components as brand-new rather than having an overkill 10year-old PSU. as people says, nothing lasts forever.

I do not replace good components. For example, I have a hard drive (15kRPM 36GB) that I took from my old main PC and placed in my new main PC when building it (no point in buying another hard drive that is exactly the same). Now my main PC also has a 146GB 15kRPM drive and when I'm building my new main PC I'll probably move those hard drives to it (or buy SAS 15kRPM drives, depending on their cost at the time).

I also tend to use computers for a long time, so that 10 year old overkill PSU might still be in use in the same system with me being very happy that I did not need to shut that system down because of PSU failure.

i cant say much as we're pretty much at the realm of preferences.

although, old components has a far higher chance of giving out than a brandnew component, with in mind that the average person upgrades once every 4years, or stingy persons sticking with theirs for over 10years(people who still uses pentium4s... no offense)

sure you can repair the unit if it starts to give out but how much would it cost to repair(excluding the trouble of repairing them, since pretty much repurchasing one is equally troublesome)? at around 4years it's worth has been well used, selling it for 30-40%(70-80% if you're lucky) of the original price and repurchasing an either better one or the same thing for much less cost(since you're reinvesting the money back).

theres also a good point of switching them instead of retaining them for so long, that is their efficiency diminishes over the years of usage that could've saved money through electricity bills.

and the part that you can sell a whole rig much easier than by part is one of the reasons why i'd sell them as a whole, although it depends on what the PC has inside. i usually keep most of the harddrives and leaving one of the oldest.