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

Did someone hear such kinds of advice:


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. o_O It sounds like it's meant to be the only a minimum requirement.

Regarding the "double requirement" it said:


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?

Well, since you did not find any article, WHERE did you quote from?

It is a "rule of the thumb" that has it's merits when you look at reliability, efficiency and possible heat peaks of the PSU, of the case, or of the room where the PC is.

Additionally, PSU somehow work like big cars, that can, at medium speed, actually burn less fuel than smaller ones.

You may want to read this article:
http://www.hardwaresecrets.com/article/181

And expecially pages 8 and 9
http://www.hardwaresecrets.com/article/181/8
http://www.hardwaresecrets.com/article/181/9

Very often, besides, it's the initial data that is a bit too high, the use of a calculator to more accurately sum up the amount of power actually needed is strongly advised:
http://extreme.outervision.com/psucalculatorlite.jsp
you may find out that the requirement to be doubled is less than what you expected.

jaclaz

Sorry. I mean there is no other (formal/reputable) article on the net which suggests this.
I read it from a page somewhere from an anonymous. It claims "Twice is the value I have gotten from lots of reading and communication in forums with people that help test and design power supplies." So there is some science to back up "double requirement" but I can't find any (except that passage).



I understand the merits of not over-stressing a PSU. But this alone doesn't justify we should ALWAYS double the needs of EACH RAIL.

The points mentioned are still too weak to justify the double requirement.



I use this to estimate my watt usage. It even calculates 3.3/5/12V rails. It's much better than the above lite calculator:
http://web.aanet.com.au/SnooP/psucalc.php

OK, after doing the calculation, if its recommendation is 300W, do I still need to double the requirement and go buy a 600W PSU?

PS: I know clearly that I shouldn't buy from bad brands. I know total watt is not the only requirement etc. I don't need to use "so much headroom" as a measure to keep me safe.

It's just one anonymous guy who says so, don't lose sleep over it.

Going for higher certainly is a good idea -- room for expansion/upgrades, not operating at the max capacity of the PSU (not the most efficient point in the curve, and also not making them heat up as much), as well as working reliably for longer (due to things like electrolytic caps aging).

I have yet to hear anyone knowledgeable (e.g. jonnyguru) say you should double any rating (and I'm not a complete moron either, I've even built simple SMPS'es before -- usually using Maxim's ICs, and also repaired plenty of them)


Nope.

And I can't say I care much for that other calculator myself for various reasons.

Yes you are right. The advice and the reasons are BS.

Talking about buying much more than needed for future expansion,
When I pick a new PSU, my choice is narrowed down into 2:
A. buy a more powerful PSU (eg 500-700W) to serve me 5-7 years. It should still be able to serve me after 3 years I upgraded most components. The power consumption can be lower or higher but I assume the worse case.
B. buy a PSU which is just what I want (eg 300-400W). Of course some headroom has been included. Then upgrade PSU again after 3-4 years.

These factors affect my choice:
- price
- usefulness of adapter (it can help my PSU to be more future-resistant)
- degree of aging (how much percentage of power is reduced after 5-7 years)
- average lifespan of PSU

I only upgrade once every 3-4 years. At that time I upgrade I find out I always need throw all components away except "HDD / FDD / DVDRW" and maybe "PSU".

Which choices should I pick?
What do you think?

Honestly, I'm not sure one can plan on 5 to7 years on a PSU right now. In the same system? Sure. But eventually connectors change. In the last 5 to 7 years, we've seen drastic changes PSU wise:
-the 4 pin P4 connector
-24 pin motherboard connectors
-SATA power connectors
-PCI-e power connectors
And that's not counting fundamental changes in their design, like the shift from high-amperage 5V bus in old designs to more power on the 12v rails, and more recently multiple rails, and more advanced features in nicer PSUs (things like active PFC weren't exactly common 7 years ago), besides the obvious increase in power ratings.

Just saying. I doubt I'll be reusing my PSUs in a new PC in 7 years.

What to chose exactly is a matter of budget, preferences and priorities. There's many good PSUs on the market nowdays, and plenty of garbage ones too.

How serious PSU aging is?
Aging occurs and this will affect the power supply.

1) How serious is aging? How much percentage of power supply will be lost after, say, 5 years of usage? Is there any article which research the aging effect?

2) For whatever reasons, BSOD has been triggered. Doe BSOD hurts any physical component?

3) If it somewhat falls below the maximum watt required by my system, what will happen? Will my other components be physically damaged?

4) When it is old or weak or next to death, is there any sign I can notice before tragedy occurs (eg fire, damaged CPU/RAM/mobo etc.)?

Thanks.

[ Topics merged. ]

Please don't create new topics when you've already started a similar one.

It's an interesting question you got there about PSU aging, I thought about it myself when checking out the "eXtreme PSU calculator".

2) No, it doesn't. However, it's usually caused by a bad driver or faulty hardware.

3) Your system may become unable to start, for example the hard drive not spinning up or the system won't even POST.
In the worst case the actual PSU itself will be damaged by overload (in rare cases and under too much load there have been PSUs catching fire).

4) Similar to 3). You can also sometimes hear some high-pitched or other unusual noise coming from the PSU. Having multiple components failing in a short period of time is a surefire sign that something i wrong.
Also if the system just shuts down (multiple times) for no apparent reason you could suspect the PSU to have something to do with it. But not everything like that is caused by the PSU, the mainboard may also play a part.

A bad PSU may damage your components; a few of my friends had a number of HDDs dieing on them because they used cheap noname PSUs. After switching the PSUs they haven't had a single problem of that kind.

I've often seen figures saying 10% of their capacity per year, assuming it's powered on 24/7. Never actually measured myself. But then again, quality varies a great deal when it comes to electrolytic caps -- another reason not to buy cheap no name PSU's. To offer dirt cheap products, they have to cut corners somewhere, and quality caps are usually one of the first things to go. It's been particularly bad in the last few years, with some manufacturers offering some very poor quality caps that go bad pretty quickly (you might have seen the tell-tale bulge on the tops of caps on a motherboard before...)

Personally, I always go for 30% extra to account for this. I've just gone through WAY too many power supplies before that were just barely sufficient. I'd rather buy one good quality and powerful power supply once for a little more $, than have to replace it every year and dealing with the problems you get when they go bad. Buying twice or more isn't really cheaper than buying a good one in the first place either. And again, going for something that sometimes looks overkill means room for expansion/upgrades and not running near 100% capacity (less heat, better efficiency), even with some capacitor aging.


A BSOD itself doesn't damage your computer. It's Windows' way to make you aware it has detected a serious problem with your system that it can't recover from (bad drivers or faulty hardware). It basically just stops.


If the PSU is too weak, your system can become unstable, reboot without warning, have various issues (like hard drives "disappearing"), not POST at all, etc.


Most faulty PSUs are found when you try to find out why a system has problems.

Anyways. Pick a good quality model, from a reputable company, with some head room, and you'll be just fine. It doesn't have to be a crazy overpriced model to be good either, look at specials, there's often some good deals on very nice models. I got a fairly decent deal last time, I got a Cooler Master Real Power Pro 750W for $74 CAD. It's made by Enhance, who makes pretty good products (the cheaper CM PSUs come from other not-so-great OEMs). No complaints at all! Like most things, it pays to research what you're buying.

If it's a rule that all topics about PSU must go to the same thread, then I agree.
Otherwise I don't see it appropriate. Read the title: "When you buy a PSU, you must buy twice as much as you need!"
It's a discussion of the "double requirement" suggested by an anon.
This question is solved, and the answer is BS.
The topic can be closed.

But this topic has nothing to do with the general questions about PSU I'm asking now. It isn't really related. The subject becomes a misnomer too.
If someone is asking those questions for convenience, I would think it's hijacking / off-topic.

Comments given. Thanks for listening. Good day.

What if it restarts suddenly?

I have set to show BSOD. Still it restarts suddenly without BSOD. If it's the case, it's very certain that the problem must lie on my hardware. Am I correct?




So as long as I'm using a PSU from a decent/good brand, apart from truly bad lucks, my PSU shouldn't kill my other hardware even if it can't supply enough power or suddenly shut down. Am I right?


The only figures I got is from Extreme Power Supply Calculator:
"Electrolytic capacitor aging. When used heavily or over an extended period of time (1+ years) a power supply will slowly lose some of its initial wattage capacity. We recommend you add 10-20% if you plan to keep your PSU for more than 1 year, or 20-30% for 24/7 usage and 1+ years."

It's scary. I lose 10-20% for 1 year of usage. 20-30% for 1 year of 24/7 usage.
However I wonder it's a general advice. They don't really do some research to come up with those figures.



So do I. But if the aging is as scary as what the above suggest. My PSU won't survive after 3-4 years of usage.

The major worry about PSU aging is I can't detect it until it causes damages on my hardware.
But after hearing what both of you say, I am relieved. It happens the PSU will signal us by keep shutting down / rebooting. STILL it won't kill my hardware, perhaps unless I keep ignoring its signals.

Of course I never buy bad/ no-name brands. I only use decent-to-good brands (or rather OEMs).

PS: Actually my monitor is acting weird. Eccentric behaviour:
1) Restart. Suddenly off then on again.
2) Soft Off. I need to press the button to turn it on.
3) Hard Off. I need to unplug and plug the cable connector to revive it.

It only happens a few times in this 1-2 months. I don't think my computer causes it.
I suspect the cable. But how can I verify my "theory"?

@Wai Wai
Just to clear things up.

You asked a question, I gave you my best answer, citing what I consider a reputable source, Hardware Secrets.

Then, once you have read the article where the reasons behind the "rule of thumb" are explained, you are free to simply ignore it or comment it's contents, to trust what is written in it or confute it, but I find ignoring it or classifying it as BS only because it does not contain what you want to hear, to be simply unpolite to me and to the time I spent trying to help you better understand the problem.

You weren't able to find a single article about the question, I gave you one.

Go and buy your 300W PSU, you'll be happy with it, rest assured.

jaclaz

That's one of the few good places for sure (along with jonnyguru.com and a few others)


I think you misunderstood what he was calling BS. From my understanding, he was saying the "double the rating" nonsense is BS, not the stuff you said or linked to.

Also, the calculator you linked to is my favorite as well, it's not perfect, but I have yet to see a better one out there. I personally wouldn't use the one in post #3. Even side-stepping countless things that bug me (like only having a handful of CPUs on the list, and having to calculate your OC'ed wattage by hand, having to enter DDR2 speed for no real reason, being forced to pick a specific hard drive in a list where yours aren't, the limited max amount of hard drives, them expecting to lookup your fan's models even though it hardly makes any difference, the lack of many options, etc), the results are WAY off. Just try my current main box:

E2160 @ 3400/1.25v, 4x DDR2 800, GeForce 8500GT, 12x Seagate 7200.11 (SATA), 1 DVD writer, check TV Tuner - Satellite & PCI SATA RAID Card, 1 extra PCI-e x1, 4 USB devices, 6x 120mm & 1x 80mm fans, 90% load, 30% cap aging. Do the same on the other -- no E2160 on the list, so perhaps try with a E6300 (same 65W TDP, only 0.06GHz diff), OC'ed to 3400 too, and Arctic cooling fans. The results? Outervision says 756w, whereas aanet's calc says a 367w PSU will suffice, when I'm 100% positive it won't (a good quality 450w isn't quite enough -- I've actually tried!), and only 23.4 amps combined draw on the 12v bus, whereas each hard drives takes about 2 amps when they spin up, so the hard drives by themselves will use that much power, leaving none to the computer... It probably wouldn't even POST, or the PSU would die a very quick death. I'm very much un-impressed by that calculator.

Sure, I may have misunderstood that, but I must have misunderstood the whole thread as well.

How I would thought this thread would go (condensed):



....


How the thread has gone (condensed):




No problems though, just wanted to clear my view on the matter.

jaclaz

I wouldn't go as far as saying double. See, it's a tricky thing because when a power supply runs at full load, it loses a lot of efficiency to heat which compounds losing more efficiency to more heat. At light load, it's wasting a lot of electricity just by running, so that's no good either. I'd say the optimal power supply size IN MY OPINION (a lot of people seem to miss that part) would be about 30 % to 40 % over your requirements. So if you calculate you need 400W of power, 520W to 560W of power supply would be just about right.

It is not against the rules. However, given how similar your new topic was (also about PSUs), I felt it could be merged with this one. We like to keep things nice and neat. If you have any other comments or concerns about this decision, feel free to PM me or the other mods.

Now... Everyone, please keep it clean.

I'll agree with this, but throw another pinch of salt into the mix.

Many online PSU calculators overestimate the actual power draw - and they do so horribly. Putting my system specs into the eXtreme Power Supply Calculator, it tells me that the recommended PSU wattage is 457W (when running everything at load). I can tell you, based on my Kill-A-Watt meter that at no point has my system ever drawn more than 350W of AC power (CPU, both video cards, and hard drives stressed).

The rated power on a PSU is the DC output. In simple terms, a 500W PSU can supply up to 500W of power to the components inside the computer. However, the actual power draw from your AC electrical system would be higher than that (because of the inefficiency of the AC->DC power conversion).

In my case, looking at the power efficiency curves at SilentPCReview's review of the Corsair HX520 (my PSU), my system draws a maximum of 300W DC at full load. In my case, I could probably get away with a 430W or 450W PSU for my system.

My rule of thumb is to find a PSU that will be at it's maximum efficiency during the majority of your usage (while still keeping at least 25% of "headroom" at full load). As you can see from the SPCR review, I'm running at pretty much peak efficiency all the time for my system.

jaclaz, I'm so sorry if my statements misled you.

Regarding HardwareSecrets, in effect I read this article before this topic.

If I get it correctly (you stressed on "Page 8 & "Page 9"), the points you want me to pay attention are "exaggeration of total watt" and "efficiency".

Please read the first post carefully. I pointed out the most important points for you:
- it suggests "twice the load" AND "twice the load on each rail (3.3V, 5V, 12V)"
- The reasons are: "[it's] 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."
- it's not meant to apply to bad/cheap brand cases which largely exaggerate their "total watt" rating. After all those PSUs should be avoided. Exaggeration is not their only problem. [Note: I find decent/good brands are pretty close to the advertised ratings. Some can really supply what it claims (eg Seasonic S12 500W/600W)]
- the "(at least) double requirement" is not some general advice. It is a rule of thumb or standard "gotten from lots of reading and communication in forums with people that help test and design power supplies"

About BS, it is meant to say I think the specific idea given in the first post and the reasons are too weak. This specific advice doesn't hold water. The wording of BS is too strong. Sorry!

Sorry I don't get why the article from HardwareSecrets is a source/proof of "(at least) double requirement". It's two different things between "we should add SOME headroom" and "we should AT LEAST ADD THE TOTAL LOAD AND ON EACH RAIL as the headroom".

That's scary. Unfortunately it seems to be the only site which calculates all sorts of requirements (total watt, +12V, +3.3 & +5V), together with UPS requirements. I couldn't find any site which will do this. They only give me total watt and that's all.
How could I pick a suitable PSU and UPS without knowing the other details?

I know their choices are limited. When it's so, I simply added the unaccounted values as a workaround.

What's your motherboard?
(I selected P35)

> "calculate your OC'ed wattage by hand"
Isn't that it will update and calculate itself?

> 12x Seagate 7200.11 (SATA)
WOW! So many SATAs.
(I selected 9+3 Seagate 7200.10. Anyway it shouldn't vary much from one to another.
I will pick those which have the same rotation speed)

> 1 extra PCI-e x1
There's no PCIe x1 as a choice.
(I selected 8600GTS as workaround, +20W than 8500GT)

> DDR2 speed for no real reason
Don't RAM size and RAM speed have an effect on PSU/UPS?

> 4x DDR2 800
what's the voltage?
(I used 1.8v)

> 6x 120mm & 1x 80mm fans
Normal case fans or...? how much watt/amp do those fans use?
You can input the numbers manually as the author suggests.
(I put 0.5 for both fans)

> 90% load, 30% cap aging
I believe it assumes 100% load when calculation.
There's no 30% cap aging. We have to add this percentage after the calculation.

> whereas aanet's calc says a 367w
You did take its recommended value (not the actual value), didn't you?


The values I got, after adding 30% aging by myself:


Thus it's 756W (Outervision) vs 572W (aanet). Which one is more accurate?

The pro version of the other calculator does that too, but it costs $2 to use (yeah, I know). Still better than one that offers such results, but whose results are completely off...

As for the other points, I'm not gonna quote them all...
P35 board << yes
Isn't that it will update and calculate itself? << it would, if your CPU is on the list, otherwise, you have to enter your wattage by hand -- it only has a handful of CPUs (47 vs over 750!) so you'll likely have to lookup the wattage of it with google first
WOW! So many SATAs. << I like/need my storage I'm looking forward to more/bigger drives still.
Anyway it shouldn't vary much from one to another. << precisely my point. Why offer countless options over something that essentially makes no difference? The same thing applies to fans, RAM speed, etc. It feels like say, if a there was a calculator for how much MPG you'd get using a certain car, this would be asking the color of the seats and type of carpet mats. They ought to just use a decent average value...
There's no PCIe x1 as a choice. (I selected 8600GTS as workaround, +20W than 8500GT) << it's more like 10W, probably less (extra NIC for VMware)
Don't RAM size and RAM speed have an effect on PSU/UPS? << it doesn't actually make a huge difference (power doesn't scale linearly with frequency), and again, it depends on the chips used, etc. Again, why not use a standard average i.e. 4 to 5 watts? And yes, 1.8v too (same story)
Normal case fans or...? how much watt/amp do those fans use? << same story again. They're whatever case fans came with the CM stacker 810 and the modules in it. There's just no way I'm going to waste time looking for the specs of each fan, when it'll only make a total change of a couple watts over the whole system build.
(I put 0.5 for both fans) << they don't use nearly that much power. That's 6 watts you've added per fan, so no wonder you calculator results are much higher. all the fans in the box combined likely use about 10 watts total.
You did take its recommended value (not the actual value), didn't you? << I'm going by "Recommended PSU wattage for decent brand PSU"
Your startup 12V1 draw needs a frightening 38.8A << non-issue This beastie has plenty of rails, and can handle 54A continuous on the 12v rails combined (more for a few minutes). This is exactly why the 367w figure it calculates isn't even close to being enough. (Also, if you don't enter 0.5A for each fan, it only calculates like 35A, not nearly 40)
So does your motherboard supports this technology? << Definitely not! But a $800 Areca ARC-1230 SATA RAID PCI-E controller would have it though... Been dreaming of one for ages, but it's a little on the expensive side

And even at 572W, it's not really what I'd want. A good 450w didn't work (it would at the very least need 500 or 550), so 572w wouldn't really have much left for upgrades, or to account for cap aging. I'd likely have to replace it within a couple of years...

Just saying. It's not exactly my fav calculator. It's not very convenient, asks for too much detail over things that makes hardly any difference, has very few CPUs, etc. And seeing the results, I don't really trust it. That's all.

Corrected the fan issues. I use 1A (12W) for the total watt consumption cost.
Updated result:


Thus it's 756W (Outervision) vs 508W (aanet).

========================================================



So do you pay for its pro version? Or you simply pick a PSU based on the total watt given by Outervision?

What's the brand/model of your PSU?




That's the problem of "free". It's free.

Can't really say it's my favourite or not. In effect there is no choice (except if I use paid service but I can't pay online). If we want more than just the total watt, it's the only one available on the net (so far). The choices left are "use it" or "not use it".

Just try to be cautious and add more headroom. Use the other one. Take the most conservative value.

Nah, I just use the basic. The results are perfect for "normal" boxes. This one said 750w, so I figured I'd look at the 750w PSUs at my usual shopping place, and the one I bought was on special for $74. I looked at the specs, and I knew 54A combined (4x 19A) on the 12v rails would be plenty (even if you remove 24A for the 12 HDs spinning up, that still leaves 30A for the CPU and such). The reviews were excellent, and it basically had everything I was wishing for (except perhaps being modular, but the cheapest good 750w modular PSU was like 3x the price, so too bad)


Cooler Master Real Power Pro 750w. It's powerful (it's not marketing lies -- it can actually supply 900w for a short amount of time, like drives spinning up), has ~85% efficiency, it has wires for everything (18 total; 6 of which are sata which I needed badly) which are plenty long (up to 31" long) even for a large full tower where the PSU is at the bottom, it has Active PFC, it's not missing a lot of the filtering stuff like many PSUs are, it uses a good design and quality parts (mosfets & diodes that can handle plenty of power -- nothing weak anywhere), it uses decent quality caps rated at good temps, it has excellent voltage regulation & low ripple, it handles brown outs just fine, it has pretty much every optional protection a PSU can offer, it runs quiet, etc. I'm sure there is even better out there, but not for $74

Thanks for your reply.

Cooler Master seems to be a "OK" brand only. Some people tell me not to pick this brand. I think, with some pre-cautions, we can actually find decent PSUs at a good price.


AFAIK you can't simply add up the amps to form the combined amps.

For example the antec neo HE 430w dual-rail, has 12V1+12V2+12V3 = 48A. But look at the 12V combined rating. It's 384w, or 32A (384/12=32 - amps is watts / voltage and the voltage is 12V in this case). So the combined amp is about 32A (not 48A). We assume that rating is accurate. There is a much more detailed but complex way to work out the combined total of 12V amps but I skip it here. -- Details.

However combined amps have some other problems. Your system may shut down sometimes even if your combined amp is over the amp required by the system. It usually happens when you system has high-end graphic card(s). They draws frightening amount of watt. The transient load can be the cause of evil. Single rail will not have such kinds of problems.

The thing is, like a LOT companies who sells PSUs, they don't actually make the units themselves. The low-end Cooler Master PSUs are built by a different OEM, which is of not so great quality. There's a large number of companies who sell products like that, ranging from low-end/not exactly great PSUs, to more expensive quality units. Then again, there is somewhat reputable brands (e.g. enermax) that makes some no so great products (I've seen so many dead enermax PSUs it's not even funny, and many of their designs aren't that great, like the 1000w galaxy, very high ripple on that) It all comes down to who made it, the design they used, and the parts used (if they skimped somewhere to save a buck). Whose name is put on it (rebadged) doesn't matter really, it's just a sticker.


I didn't 4x19A = 76A, but combined is 54A like I said (648W total -- the limit here is how much power the transformer's secondary can deliver). Like I said before, I'm not a complete moron I've built SMPS'es from Maxim ICs + International Rectifier's MOSFETs (good old IRF series -- great for H bridges too) and even classic chips like the ICL7660 (you know, back in the day when we still had ECG and NTE books?) Designing & building power supplies and audio amps was one of my hobbies in the 90's


Actually, not so much. Multiple rails is in fact better in most cases. This way, even if my hard drives spinning up suck the power so badly that the voltage went too low or such (not the case, but let's pretend), then the CPU is still unaffected, as it's fed from another rail (they have individual current limiters). So it all comes down to what you plug on which rail... I'm not too worried about the load of my totally l33t 8500GT that never sees a workload above 1% either

Why, everyone knows that cars with Red Connelly Leather interiors do eat a lot of gas....





jaclaz

Yes Antec does that too. It doesn't make any PSU. Theirs are built by different OEMs.

The problem is it's hard to know which model is produced by which OEM. Even the same OEM, it appears it can provide both good and bad PSUs to different companies / different product lines (perhaps they give budgets/orders to the OEM ?!).

Cooler Master has a mixture of mostly bad and ok and some good PSUs. That's why its fame is not good.

Achel creates its own PSU. The price is reasonable with long warranty. It produces PSUs for other firms too. But I have no ideas how good/bad their PSUs are.



LOL. It proves that either my math sucks or I was too tired when I read your reply.




Well it seems like they will not make any much difference under the modern technology. Either way fits both your and my non-gaming budget build.

The single rail can keep the voltage within the specified limits so I'm not worried about one big rail either. But single rail seems to be better to deal with transient load of ultra/very-high gaming build. So it seems it's slightly better upgradability-wise. Another option is multi-rail but no pre-rail current limits. Those are essentially no difference from single-rail.

Regarding who makes what power supply, you may want to take a look at : http://www.10stripe.com/articles/who-made-...ower-supply.php

If you want to read some really in depth, rip 'em open style, reviews of power supplies, you may also want to read about that here : http://www.jonnyguru.com/

I'm interested to know what makes such a huge difference so I started to take a close look at how they come with such a very different answer.
I even contacted the author of aanet and asked for explanations. I found something interesting between two calculators and I want to share what I found.

Capacitor aging
Outervision has this factor while aanet not. If we take this factor out, Outervision will be 582W vs Aanet 367W.

Asking aanet to include this factor into account, the author told me that he had included this into consideration in the calculation. He explained the 'generic' style of psu has already accounted for this with the huge headroom figures. Decent brand psu's use proper capacitors that last for a very long time before reducing in a capacitance/ESR enough to impact on psu's performance, so he don't add any headroom for this.

Take a typical 'low end' but not junk capacitor like a Teapo sm. Teapo are very popular, from pcp@c to low end thermaltakes. It's rated for 3000 hours at 105 degrees C @ maximum rated ripple current before 20% change in capacitance is noted. It will last for 192,000 hours at 45 degrees C. This would mean 21.92 years of 24/7 usage.
Ref: http://www.cbs.it/acrobat/teapo/sm.pdf, http://www.low-esr.com/endurance.html-ssi .

In other words, as long as you buy a decent brand of PSU which uses proper capacitors, you can use your PSU over 10+ years easily. Capacitor aging is a bluff. Adding 30% is utterly unnecessary.

However if you want to use "this factor" for other purposes (eg as a way to add headroom for future expansion), you may do it. But you don't use it simply because you are worried about "capacitor aging".


Specs vs Actual Consumption
I tried to compare the value given in each component by each calculator. I found something interesting. Outervision is trying to be very conservative and will usually take the extreme/max values. It does what other simple PSU calculator does (trying to be way overestimate just to be safe).

Aanet is very special in that it tries to be as realistic as possible (while being safe not to underestimate). It uses the actual wattage from the test results to do the calculation. It breaks down the steps and does not give only one single answer. Instead of reading the recommendation only (which is a wrong way to do), you should read all parts in "Actual System Power Consumption"!!

Some comparisons:
Outervision = (OV)
Aanet = (AN)

Product ======================== OV ===== AN (Difference)
CPU:
AMD Athlon 64 X2 3800+ (Manchester) ==== 89 ===== 47 (-42W)
AMD Athlon 64 X2 4800+ (Toledo) ====== 110 ===== 70 (-40W)
Intel Core2Duo E6300 (Allendale) ======= 65 ===== 44 (-21W)
Intel Core2Quad Q9300 (Yorkfield) ===== 103 ===== 58 (-45W)
Intel Core2Extreme QX9650 (Yorkfield) == 130 ===== 85 (-45W)
[Note: Regarding the watt usage in AN, the default wattage used by AN is the peak wattage at heavy/full load from actual tests]

Graphic card:
ATI HD2600Pro ================== 25 ===== 22 (-3W)
ATI HD3650 ==================== 45 ===== 37 (-8W)
ATI HD4850 ==================== 99 ===== 99 ( -- )
ATI HD4870 =================== 144 ==== 124 (-20W)
nVidia 8600GT ================== 38 ===== 38 ( -- )
nVidia 9600GT ================== 54 ===== 51 (-3W)
[Note: Regarding the watt usage in AN, the default wattage used by AN is the peak usage. The actual value is adjusted a little to be closer to what the graphics card would be continuously drawing in gaming]

Hard disk drive (HDD):
SATA HDD: 24/HDD (OV)
Generic HDD: ~7.875/HDD [when gaming], ~11/HDD [for fileserver] (AN)

Others:
RAM DDR2 (800) ================ ~4.5 ==== ~1.5 ( -3.0W)
Floppy drive =================== ~4.6 ===== 0.0 # (-4.6W)
DVDRW Drive ================= ~26.8 ===== 0.0 * (-26.8W)
Blue-ray Drive ================= ~28.8 ===== N/A
PCI (average) ================= ~13.4 ==== ~2.0 (-12W)
USB device ==================== ~2.3 ==== ~0.5 " (- 2W)
Firewire ====================== ~7.2 ===== N/A
#: The author excludes the option of FDD because seriously they only use power when accessing the floppy drive, so unless your running crysis off your floppy drive it's not going to make a difference.
*: Only the 5V Amp is changed (+0.3A).
": Low-powered USB device consumes 0.5W, high-powered USB device consumes 2.5W.

Reference Power Consumption:
http://www.tomshardware.com/reviews/power-...ide,1611-4.html
http://www.behardware.com/art/imprimer/670/

Comments
Outervision is a conservative calculator. Expect to buy an overwatt PSU that you may not need. In addition it appears this site has partnership with various PSUs (ads!). I don't know if it's part of the reasons why it is considerably conservative.
- CPU: It simply takes the values from the manufacturers'. It's the safest but the manufacturer usually overstates their watt usage. For example Intel states 65W TDP across many Core2Duo products. In reality it is very normal that the actual wattage is a lot different from the quoted TDP value of the CPU
- Graphic card: It's ok. However the wattage when both cards run in SLI/F is inaccurate. It's wrong to simply multiply the wattage by twice. The actual wattage is more complicated, depending on the card, motherboard etc.
- HDD: 24W per SATA HDD looks crazy. I think hardly any HDD needs over 20W. As an example, WD Caviar SE series needs less than 10W while reading/writing according to the manufacturer specifications. Slightly less (~0.5W) when it is idle.
- Others: It goes too far when measuring the possible watt usage under "Others". They are way over-estimating. For example, there are few USB devices consuming 2.3W. Devices like USB keyboard, mouse, RAM stick etc. usually need less than 0.5W.

Aanet is an aggressive calculator. It wants to get as close to the max watt usage our system really needs, with some headroom on top for safety.
- CPU: It is much more realistic and accurate. The values are the peak values at 100% load taken from the actual tests.
- Graphic card: It's slightly better than Outervison's. However the wattage when both cards run in SLI/F is also inaccurate. It's wrong to simply multiply the wattage by twice. The actual wattage is more complicated, depending on the card, motherboard etc.
- HDD: the values are taken from actual tests. The values are most realistic if your HDD is list. Otherwise you can select "generic hard disk" or any specific HDD close to yours. But then it is safer to add a few watt per HDD above its calculation, especially when your computer uses a lot of HDDs, acting as fileservers or having heavy hard disk activities. 10-15W would be a safe while realistic range of figures for 3.5" HDDs. It would be 3-5W for 2.5" HDDs.
- Others: It is being more realistic on the actual consumption of the components under "Others". It even takes RAM overclocking into account while Outervision lacks it. Although it can cover most typical or normal situations, it may get too aggressive in a few cases (eg non-typical systems which are heavy on PCI bus). Fortunately nearly all decent PSUs can accommodate your system with more than enough watt from 3.3V/5V rails, though the total watt required (in the worst scenario) will be slightly underestimated.

Except, you can't even do that. Not only they use bad caps, but most of them extremely overstate their specs (and dies at half the rated watts in a test), use junk parts that will soon break regardless, are often rated at unrealistic temps, and you just can't account for that kind of stuff. And often at the very first time the line is non-optimal (e.g. brownout) they die.


Even the best caps age, albeit not as much.


Yeah, @ 45C perhaps. In real life operating conditions, with some load temps are gonna be higher than that, especially on a lower wattage PSU that's always running nearly maxed out. And that "life" is rated for a 20% loss in capacity, which again might be too much on a barely sufficient PSU (i.e. you didn't actually add overhead for this). Plus, if every cap loses 20% capacity, in the end, your total power available likely isn't just 20% less. In real life, you see barely sufficient PSUs failing MUCH quicker than that (often only a year). Very, very few PSUs make it anywhere near 20+ years of 24/7 use without breaking down.


Simply not true.


Like I said before, they DO underestimate. On the 367W they calculate for my rig, it likely wouldn't even POST! (it just might put on a pretty fireworks show) A quality 450W has proven to be insufficient.


They let you change the % of TDP to anything you want if you think a conservative value of 85% is too much.


It's not crazy at all. You have to look at when the drives spin up. And at that time, they easily use 24W (in the very first few miliseconds, it's a lot more than that actually, like for any motor start current) Using anything less than 20W for this is plain wrong.


And the results come to LESS than what your system needs. How good is that?

Again, how good is a calculator that severely under-estimates how much wattage your system needs? And he claims to have included "headroom on top for safety" LOL. With a quality PSU ~100W more than what they calculate, my computer didn't even work right, so there... 550W perhaps might have sufficed (then add a minimum for cap aging, and some room for expansion, and 750W isn't really that overkill), but 367W isn't ANYWHERE NEAR what it needs. I don't go as far as doubling watts, but using aanet's as-is is a great way to buy something inadequate. You can try to defend them all you want, but that still makes their calculator fundamentally broken and 100% useless as far as I'm concerned. I'd sooner pay $2 for the outervision pro calc than rely on aanet's...

At this point, I'll have to borrow jaclaz's line: "Go and buy your 300W PSU" I'm not going to waste further time arguing over a broken/useless calculator...

Not to chime in and come off as arrogant but using wattage as a basis for your arguments is a complete waste of time as it doesn't paint a true picture. Wattage is the sum of voltage times amperage and anyone with some basic math skills knows that in a formula such as x * y = z you can have different values for x, y and still equal the same z.

To quote one of my replies in a previous thread that discusses this same issue: