When you buy a PSU, you must buy twice as much as you need!!

[WaiWai]

When you buy a PSU, you must buy twice as much as you need!!

Did someone hear such kinds of advice:

"You should always get a [power] supply that supports about twice the load (on each given rail), in amperes, because that is how you better assure longevity, cool operation, quiet operation, max electrical efficiency (to save on electric bills), lowest ripple and noise current, and so on."

Twice? And also for each rail?

The statement looks strong. If my system needs 300W (3.3 & 5V: 40W; +12V: 240W), I must AT LEAST get 600W (3.3 & 5V: 80W; +12V: 480W) PSU to be safe. It sounds like it's meant to be the only a minimum requirement.

Regarding the "double requirement" it said:

Twice is the value I have gotten from lots of reading and communication in forums with people that help test and design power supplies.

From the sound of it, the "double requirement" is not some general advice. It has some science or research to support it. Yet I couldn't find one of them.

What do you think?

 

[GarfieldtheCat]

IANAPA (I am not a PSU expert ), but I think that advice comes from several reasons:

1. The fact that a PSU's max efficiency is usually around ~50% of it's max load. As the load moves to either the low end or high end, the efficiency isn't as good. That being said, most good PSU's don't fall off that much at reasonable load levels, so unless you run a 100W max system of a 1200W PSU, I doubt that there will be much of an efficiency difference.

2. Also, no-name PSU's are notorious for not outputting their max ratings, so again, this is probably playing it safe to keep generic PSU's from blowing up/failing.

3. Noise - If you run near the high end of rated power, the PSU will be hotter, and fan noise will increase. If you want a quiet PSU, oversizing a bit will keep the peak load ~50%, where the fan will be at it's min RPM.

Overall, I think it would be a good idea to oversize a bit, mainly for noise and efficiency, but you don't have to go overboard and buy a 1000W for a single GPU and CPU PC

 

[DarkRogue]

It's the PSU companies brainwashing to get people to buy their more expensive products.
It really irks me when people recommend stuff like this. How much you're utilizing your PSU's capacity has NO bearing on how much more or less heat it produces. The only thing that matters is efficiency. At a given load of 300W, a 350W PSU operating at 87% efficiency outputs LESS heat than a 800W PSU running the same 300W at 80% efficiency.

Just get a PSU that fits the requirements for your hardware, and runs efficiently. I was about to start a similar thread as I just got off the General Hardware forum, amazed at people telling others to buy 750W PSU's "because they can."

Here is a quick article from SPCR: Power Supply Fundamentals
That is page 4 of the article, scroll down for a quick graph. The entire article is worth a read IMO, as it dispels many rumors and myths people seem to have in their heads.

 

[HOOfan 1]

Twice as much is a strech, but you don't want to run your PSU at maximum capacity either. I have seen several people suggest that running at 66-70% of capacity is about ideal, and running over 80% capacity isn't recommended. (their words not mine) As the capacitors age in a PSU the ammount of power it can output will decrease over time.

Just because you can run a system on the bare minimum, doesn't mean that you should.

 

[DarkRogue]

I would agree with the 66-75% of maximum thing, but really, when you're using around 250W at full load and you buy a 750W PSU.... My rig runs around 200W at full load, and people were telling me to get a 550W PSU..

 

[WaiWai]

An example has been given in the passage:

You don't get a 500 watt supply for a 250 watt load only because of future proofing, you get it to make the supply last, run cooler (a big deal that would involve a lot of explanation here), lower ripple and noise (which are damaging to your mainboard, drives, and cards), and the other reasons.

And also:

Twice is the value I have gotten from lots of reading and communication in forums with people that help test and design power supplies.

From the sound of it, the "double requirement" is not some general advice. It has some science or research to support it.
Yet I'm unable to find any article/paper which suggest that.

The reasons it mentioned are as follows:
"longevity, cool operation, quiet operation, max electrical efficiency (to save on electric bills), lowest ripple and noise current"

Longevity. True but can't justify double requirement.
It's true that the capacitors will be aging when we use it, so does other components. If you want the PSU to serve you for ~5 years, it is a good idea to oversize a bit of your requirement. I have no data about the lifespan of PSU and how serious the aging is. but I don't think it is so serious that we have to double the requirement EVEN FOR EACH RAIL to combat it.

Cooler. Wrong.
DarkRogue pointed this out. He said, "How much you're utilizing your PSU's capacity has NO bearing on how much more or less heat it produces. The only thing that matters is efficiency. At a given load of 300W, a 350W PSU operating at 87% efficiency outputs LESS heat than a 800W PSU running the same 300W at 80% efficiency."

Quieter. True but can't justify double requirement.
The PSU will be louder under full load. But what about 70-80% load? Is it still so loud? Do we really need to keep the load about 50% or less to keep it quiet.
Anyway I don't care about noise so it's not my concern.

Efficiency. Wrong.
What does it do with double requirement? It assumes maximum efficiency is attained at around 50%. Even if it was true, what about 60%/70%/80%/90% load? Does the efficiency drops noticeably?
What's important, it isn't true. It depends on what PSU you get. Bad brand PSU are going to have low efficiency under whatever load (eg well below 80%). Good brand PSU manages to do very well from most kinds of loads.
I see one PSU managed to keep the efficiency above 80% from 15-100% loads.
See the chart: http://www.xbitlabs.com/images...x-psu9/odin550-eff.png

Ripple/Noise current. Wrong (??)
I'm not too sure about this. I think this affects the stability of a PSU. However as long as the PSU is made from a decent/good brand, I don't think stability is a concern. You don't need to double your requirement to keep your system safe and stable.

The better reasons to justify "double requirement" are:
* you get absolutely no or little clues what your build consumes, so you need to buy a much more powerful just to be safe.
* your system power consumption is low. You want to be cheap and buy no-name brand. (Note: Never recommend buying from no-name or notorious brands. Such no-name brand can never feed any mid-to-high consumption builds. There are also some risks even if your system has low power consumption)
* future-proof. You want to keep using that PSU for 3-6 years. You can still upgrade your computer with the same PSU and save money.

However they ARE NOT the reasons behind the "double requirement" according to the passage.

So... in conclusion, I don't really see the logic behind "double requirement" nor any article which suggest this. Adding some headroom is a good practice but we don't need to overdo it.

 

[WaiWai]

Originally posted by: HOOfan 1
Twice as much is a strech, but you don't want to run your PSU at maximum capacity either. I have seen several people suggest that running at 66-70% of capacity is about ideal, and running over 80% capacity isn't recommended. (their words not mine)

Agree.

I think 25% excess of the REAL power output is more than enough. That means if the system requires 300W max, your PSU should be able to TRULY output ~400 watt of power. It's the real output, not just the label claims!

But the passage doesn't say so. It suggests you should double your requirement FOR EACH RAIL as a safety bottleline. So it's 100% excess FOR EACH RAIL of what you need. It's too paranoid, I would say.

As the capacitors age in a PSU the ammount of power it can output will decrease over time.

Yes and the point is how serious. If the actual power output will be halved after a few years, then the "double requirement" will make sense. But isn't it so serious? Other components will be aging too but many can be kept using for quite a long time before another upgrade cycle.

I would say no.
Maybe just 10%-20% at most although its just an wild guess.

 

[jonnyGURU]

50% is a bit of a stretch, but it's not necessarily bad advice.

GarfieldtheCat nailed most of it.

Most PSU's are most efficient at 50%. Most PSU's are at their quietest at < 50%. And really cheap PSU's can barely do 50% of what the label says it can do! At least at operating temperature.

That said, how often will your computer being running with a great enough load to where you're producing 300W (for example). It's for that reason I'm going to agree with HOOfan. Your PC is usually idle, not playing games. And if you play a lot of games, how long do you play for and is the PC at peak power consumption that whole time? Answer: No. So 60%~75% is pretty fair.

That said, you have to remember that devices like Kill-A-Watts are notoriously inaccurate and so slow they don't pick up the many transient peaks of power that could push that "300W" system over 300W. So for that reason, it may be a good idea to side with a tad bit of caution.

 

[DarkRogue]

Tad bit of caution is very good advice, but the problem is everyone is recommending PSU's rated for over 3 times the maximum power draw of their machines, which is absolutely insane and useless. I can't say I agree with people saying it will be louder when it's closer to its max load either, because isn't fan speed nowadays determined by a temp sensor ramping up when the temp rises past a certain point? If this is true, then efficiency plays a role here as well - with a more efficient rating at a given wattage, less heat is produced, meaning the fan spins SLOWER, even if it's near the capacity of the unit.

One strong point that I'm unsure about though, is that people claim running PSU's closer to their capacity will reduce its operational lifetime - is this true, or complete BS? In my opinion, heat generated by inefficient power conversion will shorten its lifetime more than running at 80-95% power rating, but I'm just not sure and can't find much information about this.

I believe we agree that running at 60-75% capacity is the optimum choice though, but it would be nice if there was an effort to inform people about this, rather than let everyone recommend kilowatt PSU's for running an HD4870 over an 8800GT "just to be safe" (slight exaggeration, but I wouldn't be surprised if this recommendation has not already been made.)

 

[WaiWai]

50% is a bit of a stretch, but it's not necessarily bad advice.

Well it is if there is no good reason.

If we are simply afraid of novice buying bad brands so we should tell them to always double the need, it's not the right way to do. Two wrongs don't make a right. You should teach the novice not to buy the bad brands. Pick a decent/good brand. Tell me to check how much watt the PSU can actually deliver (don't just rely on the label. The ???W is only a logo in my eye LOL).

After you know all actual output of PSUs you may choose, you will know what PSUs you should pick.
If your system needs 300W only, a PSU with 400W actual output is more than enough.

Most importantly, it seems even decent brand managed to keep pretty close to the rating they advertised. For a 500W, it seems they can manage to pump 460-470W!! Not bad.

 

[WaiWai]

Most PSU's are most efficient at 50%.

I think this statement is over-generalization. Perhaps things change so it no longer valid.

Before discussion, we only focus on decent-good PSUs. Few cares about the rubbish unless they are short of money.
* I don't see there is a noticeable peak at ~50%. Sometimes there is a smooth peak across 50-70% for example.
* The drop in efficiency is low towards 80%, or even 90%-100%. Assume 600W output, the difference of 85% vs 80% is only 30 watt loss. It's crazy to spend much more money (the higher-watt the PSU, much higher rise in price) for 30 watts or even less.
* I even see some PSUs whose peaks are much after 50%.
* One PSU managed to keep the efficiency above 80% from 15-100% loads. Read this chart: http://www.xbitlabs.com/images...x-psu9/odin550-eff.png
(Hey! How can i display a chart in the post directly?)

So it really depends on the PSU. But decent/good PSU seems to be efficient across most loads.
Also efficiency drop tends to be small (except very light ones) which isn't a big deal.

Most PSU's are at their quietest at < 50%. And really cheap PSU's can barely do 50% of what the label says it can do! At least at operating temperature.

Yes generally true. I believe the speed of the fan is the major cause of noise, and the load is related to the speed of fan.
But some PSUs use large fans and can still be quiet at high loads.
So again it depends on each PSU.

BTW I don't care about noise. I think my computer can be considered as "loud". But actually human has an ability to ignore what you don't want to hear.
If you learn not to notice the noise, you won't hear the noise. when I play movie or game, I hardly get interrupted by the noise.
Well, maybe I just don't care or I have filtered the noise away.

...don't pick up the many transient peaks

You mention transient peaks. This leads to some interesting discussions.

Personally I use http://web.aanet.com.au/SnooP/psucalc.php to calculate watt/volt/amp requirements.
Let's say it says my system needs 200W at most.
Would it be possible that some transient peak will occur which will draw much more than 200W?
If so, by how much?

Doesn't a PSU can do something to prevent an unexpected transient peak?

If I build a crazy system like extreme dual CF/SLI with extreme CPU, will the transient peak be much higher and much harder to tackle?

If transient peak is a serious threat AND they can draw much more than the so-called "max" suddenly, then I would agree "double requirement" is correct for extreme builds.

 

[Jessica69]

I'm just wondering where the "the more you draw from your power supply, it won't heat up because of efficiency" came from.

Despite excellent efficiency, even 85% across the board, you have increasing amounts of wasted electricity, which is converted into heat.

For example, take a PC Power and Cooling Turbo Cool 860. At 50% output, or 438W roughly, it consumes 515W from the outlet on the wall and has an exhaust temp of 49C. (This is from one of a fairly well respected testing site for power supplies.) Increase the power output of the unit to 75%, which equates to an output of 657W, it draws 776W from the wall and now has an exhaust temp of 50C...minor increase, but an increase none the less. And at full tilt....output of 856W, it's drawing 1027W from the wall outlet and now has an exhaust temp of 56C.

To claim that higher efficiency makes the unit cooler is true, to a point. But unless you have a magical power supply with 100% efficiency, there is always loss in the conversion and heat is produced. Ask your power supply for more output and more wattage is wasted and more heat produced.

From another vantage point.......a power supply is 90% efficient (not in reality, but easy to work with in round numbers.) The power supply is a 1000W unit. At 25% output, or 250W, it has to draw from the wall outlet 275W...so 25W is wasted. At 500W, or 50% output, it draws 550W and is now wasting 50W....and on and on. No matter how efficient, as you ask more output from the power supply, more electricity is wasted and turned into heat. And as this is an imperfect world, 85% is the typical high efficiency power supply.....so you're wasting 15% of the draw from the wall's outlet in making the wattage you need.

And as you need more, that 15% means an increasing amount of elect. is wasted......15% of 800W is more than 15% of 300W.

And as you create more heat the fan does indeed spin faster to ventilate and cool the power supply.

Very simple......and is a phenomenon that's been seen every time a power supply is put through its paces....no matter who tests it, Jonny G and his crew, Hardware Secrets, HardOCP, Hardware Canucks, Anandtech.......

 

[WaiWai]

Tad bit of caution is very good advice, but the problem is everyone is recommending PSU's rated for over 3 times the maximum power draw of their machines

In today world, the problem is no longer people who are not cautious but they are overly cautious.
It's common to hear overwatt suggestions rather than underwatt suggestions. Quite a few can be considered as "highly overwatt".

One strong point that I'm unsure about though, is that people claim running PSU's closer to their capacity will reduce its operational lifetime - is this true, or complete BS? In my opinion, heat generated by inefficient power conversion will shorten its lifetime more than running at 80-95% power rating, but I'm just not sure and can't find much information about this.

It's hard to know. We can only based on logic and some understanding about electronic products.

I believe overstressing (eg 95-100% loads) will reduce the life. But would there be difference between a 50% load and 80% load? I don't think so. The impact is minor to make a big difference.

What's more, sometimes using less can do harms. Think about a DVD player you don't play much. It turns out to die faster than some product you keep using it. I have been learnt I should play the dusty product once in a while to keep them healthy. My car has been used (nearly) daily (with moderate usage, I don't stress it too much!). It managed to live much longer than some people who used the cars 1-3 months on average.

Don't get me wrong that I'm implying that using a lot can increase their life. I just want to say a moderate to moderately high usage may not hurt the product more than the usual process of wear-and-tear. So 50% load and 80% load may makes no difference at all. Over-usage is however a bad thing and often hurts the product.

I believe we agree that running at 60-75% capacity is the optimum choice

Well I believe it depends on the PSU and what you mean by "optimum" What is optimal?
Noise reduction? Efficiency? Or else?

I won't set a hard number on it.

What's more, currently too many people are too conservative. One fact is we usually use the max watt to do the sum. Don't forget the system uses much less most of the time so they are going to get the wrong optimal capacity.

For example:

The system uses max 300W. Then 300W / 60% = 500W decent PSU.

Actually the system uses less than 150W most of the time.
So the TRUE capacity becomes 30% (=150/500) most of the time.

NO system has the same load most of the time. They keep fluctuating.
It's much better to choose a PSU which is efficient across most loads, while then focusing on getting 60-75% optimal load.

When people like to double or triple their needs, they are actually get much less loads most of the time, resulting in inefficiency and wastage => more heat.
They are really paying more for getting less, while PSU suppliers are more than happy not to stop you. How sad!

 

[WaiWai]

I'm just wondering where the "the more you draw from your power supply, it won't heat up because of efficiency" came from.

AFAIK, actually it should be "the more watt you draw, the hotter it is."

Heat is generated by:
- the amount of watt drawn
- efficiency

But NOT:
- PSU capacity
- % of load

Saying that 50% load helps to cool the PSU, as the quote of the first post claims, is false.

Even with 100% efficiency, PSU will still get some heat.

 

[DarkRogue]

Originally posted by: Jessica69
I'm just wondering where the "the more you draw from your power supply, it won't heat up because of efficiency" came from.

Did you read the link I posted?
Excerpt:

From a PSU heat waste point of view, the differences are significant:

At 200W load,

* the 300W model would generate 44W of heat (18% of 244W AC input);
* the 600W model would generate 50W of heat (20% of 250W AC input).

At 90W load,

* the 300W model would generate 23W of heat (20.5% of 113W AC input).
* the 600W model would generate 32W of heat (26% of 122W AC input).

Here, the 300W PSU has an efficiency of 82% @200W vs the 600W's efficiency of only 80% @200W.
At 90W, the 300W PSU is 79.5% efficient vs the 600W's 74% efficiency.

I never said that it would not heat up due to efficiency, I said that it would be COOLER than the one with less efficiency, as that excerpt clearly shows. It's true that the more power you draw, due to the fact that we do not have 100% efficiency, there WILL be heat generated. However, at any given load level, the MORE efficient unit will generate LESS heat than the LESS efficient unit.

You seem to be arguing that every PSU is capable of a typical efficiency of 85%. However, WHEN does it hit 85%? Most people claim it happens right around 50% capacity. Well then, for an 800W PSU with 85% efficiency at 50%, that means you need to draw 400W in order to reach that point. Most users who have a single video card and a single CPU in their system will not draw very much over 250W. By that measure, a 500W PSU @ 50% will be right at 250W - peak 85% typical efficiency, whereas the 800W PSU will be at around 31% load, normally at a lower efficiency rating, and thus producing MORE heat than the 500W unit at 250W load.

I hope that clarifies my position a bit.

Edit:
Also realize that most users' PCs will be idling most of the time, unless they're into distributed computing projects. I believe a good 'typical' idle wattage would be around 100-150W, which means it will impact the high capacity PSU's even more if their efficiency at this low end is very poor.

HOWEVER - IF a PSU is virtually 100% CONSISTENT in efficiency from 0%-100% load, then all these points become moot, and the only argument is how much headroom you want, assuming the efficiency LINE (not curve in this case) is quite high.

 

[WaiWai]

HOWEVER - IF a PSU is virtually 100% CONSISTENT in efficiency from 0%-100% load, then all these points become moot, and the only argument is how much headroom you want, assuming the efficiency LINE (not curve in this case) is quite high.

BTW I wonder a PSU with virtually consistent in efficiency from 0%-100% load exists.
I see a few PSUs manage to keep really flat between 30-100% load. But I don't see any PSU which can keep high efficiency at low load.
For people who like super-high-watt PSU, they are really paying more for less.

 

[beray]

Originally posted by: DarkRogue

Originally posted by: Jessica69
I'm just wondering where the "the more you draw from your power supply, it won't heat up because of efficiency" came from.

Did you read the link I posted?
Excerpt:

From a PSU heat waste point of view, the differences are significant:

At 200W load,

* the 300W model would generate 44W of heat (18% of 244W AC input);
* the 600W model would generate 50W of heat (20% of 250W AC input).

At 90W load,

* the 300W model would generate 23W of heat (20.5% of 113W AC input).
* the 600W model would generate 32W of heat (26% of 122W AC input).

Here, the 300W PSU has an efficiency of 82% @200W vs the 600W's efficiency of only 80% @200W.
At 90W, the 300W PSU is 79.5% efficient vs the 600W's 74% efficiency.

I never said that it would not heat up due to efficiency, I said that it would be COOLER than the one with less efficiency, as that excerpt clearly shows. It's true that the more power you draw, due to the fact that we do not have 100% efficiency, there WILL be heat generated. However, at any given load level, the MORE efficient unit will generate LESS heat than the LESS efficient unit.

You seem to be arguing that every PSU is capable of a typical efficiency of 85%. However, WHEN does it hit 85%? Most people claim it happens right around 50% capacity. Well then, for an 800W PSU with 85% efficiency at 50%, that means you need to draw 400W in order to reach that point. Most users who have a single video card and a single CPU in their system will not draw very much over 250W. By that measure, a 500W PSU @ 50% will be right at 250W - peak 85% typical efficiency, whereas the 800W PSU will be at around 31% load, normally at a lower efficiency rating, and thus producing MORE heat than the 500W unit at 250W load.

I hope that clarifies my position a bit.

Edit:
Also realize that most users' PCs will be idling most of the time, unless they're into distributed computing projects. I believe a good 'typical' idle wattage would be around 100-150W, which means it will impact the high capacity PSU's even more if their efficiency at this low end is very poor.

HOWEVER - IF a PSU is virtually 100% CONSISTENT in efficiency from 0%-100% load, then all these points become moot, and the only argument is how much headroom you want, assuming the efficiency LINE (not curve in this case) is quite high.

One basic thing missing is cooling capacity.

600W PSUs would have twice the cooling capacity of 300W PSUs when designed properly. At low power usage 600W PSUs will still have twice the cooling capacity of 300W PSUs.

600W PSUs will always be cooler than 300W at same workloads. Even 10% differ in efficiency would not make the 300W cooler.

 

[MrOblivious]

Originally posted by: WaiWai
An example has been given in the passage:

You don't get a 500 watt supply for a 250 watt load only because of future proofing, you get it to make the supply last, run cooler (a big deal that would involve a lot of explanation here), lower ripple and noise (which are damaging to your mainboard, drives, and cards), and the other reasons.

And also:

Twice is the value I have gotten from lots of reading and communication in forums with people that help test and design power supplies.

From the sound of it, the "double requirement" is not some general advice. It has some science or research to support it.
Yet I'm unable to find any article/paper which suggest that.

The reasons it mentioned are as follows:
"longevity, cool operation, quiet operation, max electrical efficiency (to save on electric bills), lowest ripple and noise current"

Longevity. True but can't justify double requirement.
It's true that the capacitors will be aging when we use it, so does other components. If you want the PSU to serve you for ~5 years, it is a good idea to oversize a bit of your requirement. I have no data about the lifespan of PSU and how serious the aging is. but I don't think it is so serious that we have to double the requirement EVEN FOR EACH RAIL to combat it.

It isn't just longevity why you have to increase the size it is a little thing called the derating curve. For instance one very popular OEM designs their units and rates them at 25c. Their old derating curve was 10w/c from 25c to 35c with no data above 35c. Their newer designs are 8.6w/c from 25c to 35c. This is rather relevant for a lot of us as my office the week I was testing that particular unit was 38.5c. So even if I wasn't running it at 45c and just had it in my office PC that 1000w unit would have been a 884w unit. Now scale that down to your lower powered units and all the sudden you have real problems.

Cooler. Wrong.
DarkRogue pointed this out. He said, "How much you're utilizing your PSU's capacity has NO bearing on how much more or less heat it produces. The only thing that matters is efficiency. At a given load of 300W, a 350W PSU operating at 87% efficiency outputs LESS heat than a 800W PSU running the same 300W at 80% efficiency."

A good modern 350w power supply is exceptionally unlikely to have an efficiency of 87% with any kind of regularity while most good modern 800w units will exceed 80% at 300w. Beyond that, a 800w unit will have a larger cooling capacity so it will be cooler...in general.

Quieter. True but can't justify double requirement.
The PSU will be louder under full load. But what about 70-80% load? Is it still so loud? Do we really need to keep the load about 50% or less to keep it quiet.

The lower the load the quieter...so um yes? At 70-80% can power supplies become loud? Subjectively, from listening to lots of them at 50% and 75% load, yes the break seems right about 75% for most fan controllers. By 80% almost certainly they begin to become much louder.

Anyway I don't care about noise so it's not my concern.

Great but that doesn't invalidate the point.

Efficiency. Wrong.
What does it do with double requirement? It assumes maximum efficiency is attained at around 50%. Even if it was true, what about 60%/70%/80%/90% load? Does the efficiency drops noticeably?
What's important, it isn't true. It depends on what PSU you get. Bad brand PSU are going to have low efficiency under whatever load (eg well below 80%). Good brand PSU manages to do very well from most kinds of loads.
I see one PSU managed to keep the efficiency above 80% from 15-100% loads.
See the chart: http://www.xbitlabs.com/images...x-psu9/odin550-eff.png

I can find exceptions to every rule, but in general power supplies are most efficient between 40-60%. Period.

Ripple/Noise current. Wrong (??)
I'm not too sure about this. I think this affects the stability of a PSU. However as long as the PSU is made from a decent/good brand, I don't think stability is a concern. You don't need to double your requirement to keep your system safe and stable.

Ok so you don't know what ripple/noise is but that really isn't important for the point. The point is the higher the load the more ripple/noise you get. Now as long as it is in specificaitons it should be fine but the point that running a power supply near capacity increases ripple/noise is completely true.

The point here is that all of the points are valid reasons why you don't want to run near capacity but the real reason for the 2x suggestion by many people is for an unknown power supply you know NONE of these variables. If you can find a good review of a unit now you know those variables and it goes out the window.

 

[WaiWai]

It isn't just longevity why you have to increase the size it is a little thing called the derating curve.

Yes that's true. Cheap manufacturers like to exaggerate their total watt supply in unrealistic situations.
In effect the causes of all evils are the exaggeration of "total watt" label.

Seasonic S12-600: http://www.silentpcreview.com/article247-page4.html
They are able to draw 600W so they are honest to give an accurate rating under normal situations.
We have to take derating effect into account when manufactures rate their total watt at unrealistically low temperature.
For a Seasonic S12-600, it's 100% load with a hotter PSU. Still it can actually supply 600W. You don't need to take any number out unless you are working hotter than normal situations.

After all, the best is not to make guess. Don't base your judgement on the advertised "total watt".
Check how much it can really supply in normal situations.
It's much better than trying to overestimating everything. This will end up buying an unnecessary PSU running at low loads => inefficiency => more heat and shorter life.

A good modern 350w power supply is exceptionally unlikely to have an efficiency of 87% with any kind of regularity while most good modern 800w units will exceed 80% at 300w.

True but it's only an example given by DarkRogue to illustrate a concept. He didn't claim that a 350W PSU generally has 87% efficiency. After all the concept still applies regardless of a bad choice of number.

[My personal preference on noise issues] Great but that doesn't invalidate the point.

Very true. I think you should get that my personal preference has nothing to do with invalidating the point.
You did see I says "True but can't justify double requirement."

Don't forget it's meant to be a minimum requirement. It claims we should stay below 50% (or perhaps well below) to keep the PSU quiet. But it isn't generally true. It isn't hard to find one PSU which has a virtually flat line of noise about 0%-70% load.

If it changed into "we shouldn't overload the PSU to keep it quieter", then I would agree.

I can find exceptions to every rule, but in general power supplies are most efficient between 40-60%.

This is not exception! It's an example. I can always find more.

Claiming that 2x requirement is a minimum (a must) suggests that very few to no PSU can be efficient at other loads (eg 30%-40%, or 60%-80%). Otherwise why must we stick with 50% load?

After all we can't keep the load at about the fixed percentage. The load keeps changing. Are we able to keep your computer stuck at 50% load most of the time? What's more, we are doubling based on an extreme value (ie the max power your system will use).

As far as I read, at least decent-to-good PSUs manage to stay above 80% in most loads. There is no distinct peak or noticeable drop of efficiency in most loads (about 30-80%). Is it really a big issue when the efficiency only drops slightly in other range of loads?

The point is the higher the load the more ripple/noise you get. Now as long as it is in specificaitons it should be fine but the point that running a power supply near capacity increases ripple/noise is completely true.

I know running a power supply near capacity increases ripple/noise current. But this fact alone isn't enough to support its claim.

What I understand is the PSU is fine as long as it is in specifications. The function of a PSU is to feed my computer. Either +/-5% or +/-3% difference is not going to fail the PSU. This isn't one of the factors when I pick my PSU (because all decent brands pass it!).
So I don't understand why it's a good reason, let alone a good reason to support x2 requirement.

the real reason for the 2x suggestion by many people is for an unknown power supply you know NONE of these variables.

Yes but only for a low-watt system.
It is still a very bad idea to pick a low-quality PSU. Largely exaggerated total watt is only one of their problems. They will die faster, more unstable etc.
2x is a bad way to find a suitable PSU. Why not, for example, using a good online PSU calculator as a way to get one close to your needs?

 

[WaiWai]

<sorry, deleted, wrong post!>

 

[MrOblivious]

Originally posted by: WaiWai


After all, the best is not to make guess. Don't base your judgement on the advertised "total watt".
Check how much it can really supply in normal situations.

Um I actually do that since I have the ability to and access to the pertinent data. End users do not.

It's much better than trying to overestimating everything. This will end up buying an unnecessary PSU running at low loads => inefficiency => more heat and shorter life.

What? Let's look at an example shall we?

http://www.hardocp.com/article...w0LCxoZW50aHVzaWFzdA==

At 25% load or 177w DC that unit is 83.88% efficient and has an exhaust temperature of 46c when the input temperature is 45c. That is a 1c delta. Not exactly inefficient, hot or shorter life span. Now move to 75% what happens? Efficiency is 83.08% and temperature is 55c. 80% and what happens? Efficiency is 82.23 and exhaust is 61c.

So do you get the point? Lower efficiency at lower load levels is preferable from a heat stand point than lower efficiency at higher load levels as users who push past 50% advocate as the total heat generated at higher loads is larger than that at lower load levels. At 25% it was 34w. At 75% its 103w.

True but it's only an example given by DarkRogue to illustrate a concept. He didn't claim that a 350W PSU generally has 87% efficiency. After all the concept still applies regardless of a bad choice of number.

It was unfair scenario and is incorrect for the reason stated.

You did see I says "True but can't justify double requirement."

It does to some people and is technically correct, therefore your opinion on the subject does not impact tis veracity and should not be considered.

But it isn't generally true. I see quite a few have a virtually flat line on noise about 0%-70% load.

You see them CLAIM you don't actually listen to them.

Claiming that 2x requirement is a minimum (a must) suggests that very few to no PSU can be efficient at other loads (eg 30%-40%, or 60%-80%). Otherwise why must we stick with 50% load?

No it was stated that peak efficiency occurs between 40-60% and this is correct. There really is no if and or buts about that.

After all we can't keep the load at about the fixed percentage. The load keeps changing. Are we able to keep your computer stuck at 50% load most of the time? What's more, we are doubling based on an extreme value (ie the max power your system will use).

Nope but if you center around 50% you can stay in the most efficient part of the efficiency curve. Hence the point of trying to stay around 50%.

As far as I read, at least decent-to-good PSUs manage to stay above 80% in most loads. There is no distinct peak or noticeable drop of efficiency in most loads (about 30-80%). Is it really a big issue when the efficiency only drops slightly in other range of loads?

Depending on how you grade you load steps there most certainly is a distinct peak and with every unit there is a peak value. Some are flatter than other but all units peak. Is it really a bi issue? Depends on how many power supplies you have and it depends on how much heat you are now wasting when you get past 50% as I stated above.

I know running a power supply near capacity increases ripple/noise current. What I don't know is why this is the reason to justify x2 requirements. Explain if you understand why.

It is called minimizing the increase. Except for total crap units there are rarely cases when a unit begins to approach the ATX12v specification limit before 50%.

the real reason for the 2x suggestion by many people is for an unknown power supply you know NONE of these variables.

Yes but only for a low-watt system.
It is still a very bad idea to pick a low-quality PSU. Largely exaggerated total watt is only one of their problems. They will die faster, more unstable etc.
2x is not a right way to pick a good PSU anyway.

I never said to pick a low quality power supply so don't put words in my mouth.

Take it as you will, but picking units that are close to their total capacity is begging for trouble especially when you guesstimated your draw using a Kill-A-Watt.

 

[DarkRogue]

MrOblivious, do you have any data about various units at the same power draw (not percentage) and their heat output?
For example, instead of a 300W and 800W at 50% load, what about a 300W and 800W at 200W.
Everyone keeps talking about units at the same percentage here, but no matter what PSU people use, their consumption will be the same as long as their other hardware remains the same. I'd like to know if the higher capacity units really are cooler at a low point when they're less efficient at it than a smaller unit operating near its peak efficiency.

 

[MrOblivious]

Originally posted by: DarkRogue
MrOblivious, do you have any data about various units at the same power draw (not percentage) and their heat output?
For example, instead of a 300W and 800W at 50% load, what about a 300W and 800W at 200W.
Everyone keeps talking about units at the same percentage here, but no matter what PSU people use, their consumption will be the same as long as their other hardware remains the same. I'd like to know if the higher capacity units really are cooler at a low point when they're less efficient at it than a smaller unit operating near its peak efficiency.

Not really no one has ever cared enough to ask before.

 

[HOOfan 1]

Originally posted by: DarkRogue
MrOblivious, do you have any data about various units at the same power draw (not percentage) and their heat output?
For example, instead of a 300W and 800W at 50% load, what about a 300W and 800W at 200W.
Everyone keeps talking about units at the same percentage here, but no matter what PSU people use, their consumption will be the same as long as their other hardware remains the same. I'd like to know if the higher capacity units really are cooler at a low point when they're less efficient at it than a smaller unit operating near its peak efficiency.

At jonnyguru.com they measure input and output temperatures on their units. These are just a couple I picked up but you can go look at their review for more.

Mind you these are Cold load tests with the PSU sitting out in the open as opposed to closed up in a computer case with other heat producing equipment

Delta 300 Watts at 195W load input 25C output 32C Power Efficiency is 74% that is a delta of 7C
NorthQ Giant Connector 850W (this is built by Enhance) at 191W load input 18C output 21C power efficiency is 84% that is a delta of 3C

 

[WaiWai]

Um I actually do that since I have the ability to and access to the pertinent data. End users do not.

They can read reviews to get the figures.

At 25% load or 177w DC that unit is 83.88% efficient and has an exhaust temperature of 46c when the input temperature is 45c. That is a 1c delta. Not exactly inefficient, hot or shorter life span. Now move to 75% what happens? Efficiency is 83.08% and temperature is 55c. 80% and what happens? Efficiency is 82.23 and exhaust is 61c.

Low load here means somewhere between, say, 0-20% load.
When the load is too light, the efficiency drops significantly.

When people tend to largely overestimate what they need (don't forget they double their requirement based on MAX watt usage), the normal watt usage will be well under it. It needs to inefficiency => more heat and shorter life.

True but it's only an example given by DarkRogue to illustrate a concept. He didn't claim that a 350W PSU generally has 87% efficiency. After all the concept still applies regardless of a bad choice of number.

It was unfair scenario and is incorrect for the reason stated.

It has nothing to do with unfair scenario.

Let's say an apple costs $5. What is the price of three apples? The answer is $15.
Then someone was arguing that the calculation was wrong because apple never costs $5 in reality.

What he tried to say is:
"For the same amount of watt drawn, other things being constant, a higher capacity but lower efficiency of PSU generates more heat than a lower capacity but higher efficiency of PSU."
In this regard, the capacity *alone* has nothing to do with the heat generated.

It does to some people and is technically correct, therefore your opinion on the subject does not impact tis veracity and should not be considered.

What is technically correct?

Fact 1: Load between 50% and 75% is optimal to reduce noise
Fact 2: Higher capacity helps to keep the load lower
Conclusion: we should buy TWICE AS MUCH AS what we need ON EACH RAIL.

While I agree lower load help to make PSU quieter, I failed to see why they are the reasons to justify the above claim.
Don't forget our original discussions are arguing about whether the above-mentioned reasons supports the "double requirements".

Nope but if you center around 50% you can stay in the most efficient part of the efficiency curve. Hence the point of trying to stay around 50%.

Even if it's true, it can't justify the "DOUBLE ON EACH RAIL" advice.
You are doubling based on the MAX watt you will use. It's the extreme case. Your computer won't be loading at around 50% if we follow this advice.

I never said to pick a low quality power supply so don't put words in my mouth.

Which statement gave you the thought that I said it was you who suggest picking a low-quality PSU?

You said: "the real reason for the 2x suggestion by many people is for an unknown power supply you know NONE of these variables."
I said" Well, this case is only applied to cheap/no-name brand. Only then do we need to take the 2x advice. In other cases it's not. After all we are not going to suggest people using cheap brands."
You said: "I never said to pick a low quality power supply so don't put words in my mouth."

Hope it's clear now.

Take it as you will, but picking units that are close to their total capacity is begging for trouble

Very true.

But it seems you failed to see the difference between "we should add some headroom" and "we should add the total watt required for each rail as the headroom (2x suggestion)".

After all, I don't really understand what you are trying to argue.
Do you agree with the author that "at least DOUBLE ON EACH RAIL" is a pre-requisite when we pick a good PSU?