PSU Maybe to big

PSUs: The source of DC power for all components in the PC & often a big noise source.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee, Devonavar

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

PSU Maybe to big

Post by fastturtle » Fri May 20, 2005 7:11 pm

I'm looking at building a very low power Pentium M (Dothan) based system using these System Specs

What I got to thinking about is the fact that I'm going for an ultra low wattage system and that the PSU may be :lol: To big for it. Since I'm thinking that, I want something that's going to be silent (fanless maybe) and very efficient as I don't plan on adding extra fans because I want a silent or near to it setup with decent performance. On that note I have also considered going with the 700 mhz Bananias Pentium M due to the Ultra Low Wattage of it (-4 watts I think).
Last edited by fastturtle on Sat May 21, 2005 7:09 am, edited 1 time in total.

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Fri May 20, 2005 7:27 pm

I you want a quiet system, I doubt that the PSU that comes with the case is the best choice. I would go with the Seasonic S12-330.

Maybe you can sell the OEM PSU on eBay, but don't expect to get much since the case with included PSU that you selected sells for $31.50. Actually, even though the OEM PSU is rated for 350 watts, it probably does not put out anywhere near the amount of power as the Seasonic S12-330.

alglove
Posts: 363
Joined: Fri Feb 06, 2004 11:21 am
Location: Houston, TX, USA

Re: PSU Maybe to big

Post by alglove » Fri May 20, 2005 7:48 pm

fastturtle wrote:On that note I have also considered going with the 700 mhz Bananias Pentium M due to the Ultra Low Wattage of it (-4 watts I think).
Negative 4 watts? That is ultra low!!! :P

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

Post by fastturtle » Fri May 20, 2005 8:22 pm

m0002a: I guess you didn't understand my concern: the existing PSU is TO BIG so what I need to do is downsize to something smaller.

The lowest wattage rating I saw was (I think) +2 watts for the 700 mhz bananias version. That's actually down in the range of the Via fanless 533 Epia CPU and with far better performance then any of them. 8)

That's why I'm know that the 350 is going to be too big a PSU.
O'Kay a few changes have been made
  • CPU 15 watts
  • Burner 30 watts
  • Hard drive 25 watts
  • Motherboard 25 watts
  • Geforce2 MX400 25 watts Killed
  • Sound Blaster PCI sound Card 10 watts
that's a total maximum draw of 105 watts. Well below the 45% minimum load for a Seasonic S12-330. I suspect the real world is going to be even lower then that, more around the line of 100 watts maximum; which means something around 200 watts. For the total draw, that would be almost perfect efficiency since it's right at 65% loading, leaving 30 watts (15%) headroom for upgrades/addons.

Don't even mention a DC/DC PSU such as the mini-itx can use. This system isn't quite that low power, although I guess it wouldn't take much to get it down below 100 watts. :lol:

Good Catch alglove: :oops:
Last edited by fastturtle on Sat May 21, 2005 6:37 am, edited 1 time in total.

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Fri May 20, 2005 10:02 pm

fastturtle wrote:m0002a: I guess you didn't understand my concern: the existing PSU is TO BIG so what I need to do is downsize to something smaller.
I understand exactly what you want. But you apparently don’t understand how PSU's work.

A 350 watt PSU does not draw 350 watts of power unless the power is needed by the components connected to it. The main consideration about getting a PSU that is too large (besides wasting money) is that a given PSU is usually more efficient (less power is wasted as heat) when it is running near the middle to upper range of its capacity.

However, the Seasonic S12 is more efficient across the entire power range (about 80%) compared to most other PSU's (the cheapo PSU like the one in your case is probably about 70% efficient at best) that even the 600 watt model of the S12 would use less electricity than the 350 watt PSU you have selected (for the same components). Not only does an efficient PSU save electricity, but it produces less heat (which indirectly translates into noise) than an inefficient PSU.

It would probably be good for you to spend some time reading the reviews of PSU’s and list of recommended PSU’s on this website, so that you have a better understand on how PSU’s work and what is the most important factors with regard to efficiency (wasting electricity) and building a quiet PC.

If Seasonic made an S12 with fewer watts than 330, then maybe I would recommend that. This forum is dedicated to Silent Computing, and IMO the Seasonic is the best choice for that end on your system, unless you want spend a lot more money on a completely fanless model like the Antec Phantom 350 (which is also a very efficient PSU).

lenny
Patron of SPCR
Posts: 1642
Joined: Wed May 28, 2003 10:50 am
Location: Somewhere out there

Post by lenny » Fri May 20, 2005 10:24 pm

fastturtle wrote:that's a total maximum draw of 130 watts. Well below the 45% minimum load for a Seasonic S12-330.
Where's this 45% minimum load requirement from?

My Seasonic S12-430 hardly ever draws more than 150W from the AC. In fact, when CnQ kicks in and I underclock the ATI when in 2D, it draws only a little over 75W. Assuming a very optimistic 80% efficiency that's only 60W DC. No problem whatsoever.

Besides, total wattage is not everything. You got to make sure that each voltage rail is sufficient. Unlikely to be a problem with your configuration though.

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Fri May 20, 2005 10:47 pm

fastturtle wrote:
  • CPU 15 watts
  • Burner 30 watts
  • Hard drive 25 watts
  • Motherboard 25 watts
  • Geforce2 MX400 25 watts
  • Sound Blaster PCI sound Card 10 watts
that's a total maximum draw of 130 watts.

Don't even mention a DC/DC PSU such as the mini-itx can use. This system isn't quite that low power, although I guess it wouldn't take much to get it down below 100 watts. :lol:
There are DC/DC PSUs with 200W capacity. Such could be ideal for you.

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

Post by fastturtle » Sat May 21, 2005 6:35 am

A few think that the AC demand of a PSU is the same as it's output. I'm afraid not; The calculation is simply Amps time Voltage to get watts. Simply put this is 12@mps x 12 volts provides you with 144 watts of output. It's the same for 5 and 3 volts. This is a basic calculation called watt's law and is directly derived from Ohm's law. Google for more info:

O'Kay All:

Here's the rules of thumb I'm operating from:

NOTE This does not apply to Mini-itx or laptops

Most ATX PSU's are designed with a minimum demand load for stability. This seems to be between 42-48 percent of designed maximum output. The reason for this minimum is the fact that a motherboard normally draws at least 15 watts and some can go to 35 watts, average is 25 watts. This is the motherboard itself mind you.

When you add in a CPU, video chip and memory, you suddenly put this minimum demand up to 50+ watts, so the engineers don't waste their time/effort designing a PSU that's stable under that minimum due to cost considerations.

So we now have a minimum draw that has to be met and since most PSUs are designed for maximum efficiency at 75-80 percent load; engineers always include a safety/fudge factor; so based on those two factors and selecting 65% as my desired loading, I determine how much power I need and then select the best match from stock.

If there is nothing very close, then I have to compromise, preferably oversized unless it's more then 30% over, then I go closes match down. If there is no match available within 30% of load either way, then I rethink my hardware selections to push the envelope to get close to a standard size. Now the problem I'm having here is that most standard cases are including 300+ PSU's and for this system that's going to be OVERKILL.

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Sat May 21, 2005 7:38 am

I'm replying publicly here to fastturtles private reply to my post, since there's no reason to keep it private I think.

Quote from mini-box.com's pw-200-m's user manual (except power):

voltage max current peak current max power
5V 6A 10A 30W
5Vsb 2A 10A 10W
3.3V 6A 10A 20W
-12V 0.1A 0.2A 1.2W
12V 12A 13.5A 144W

So most of the power is available only on the 12V line, but if you knew the exact requirements per line for your machine then this might be ok?

Oh and my server is quite similar to your system, and it only draws 40W on idle, 60W max. (measured with a power meter plugged between the wall socket and the psu power cord) Specs:

P3 733MHz slot1
512MB 133MHz sdram
80GB seagate barracuda IV pata
matrox millennium 1 (1997)
intel mobo with integrated sound
2* intel 100Mbps nic
4* papst 80mm fan @ 5V

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

Post by fastturtle » Sat May 21, 2005 9:28 am

Thanks for the specs on the DC/DC psu.
lm wrote: Oh and my server is quite similar to your system, and it only draws 40W on idle, 60W max. (measured with a power meter plugged between the wall socket and the psu power cord) Specs:

P3 733MHz slot1
512MB 133MHz sdram
80GB seagate barracuda IV pata
matrox millennium 1 (1997)
intel mobo with integrated sound
2* intel 100Mbps nic
4* papst 80mm fan @ 5V
Your CPU draws= 20 watts
Hard drive (ballpark) 25 watts
Motherboard (ballpark) 25 watts
Fans (ballpark) 4@5 watts = 20 watts

Total estimated system wattage= 90 watts. Pretty close to what I've designed. What sized PSU are you running?

To All :oops:
Sorry; I got wrapped up in my calculations once again. Very Bad habit of mine. Seriously though, what I'm trying to accomplish two things; build a fanless system for noise control and retain performance comparable to a 950 T-bird. This means I'm not wedded to any specific PSU as yet (Thanks to lm once again for the DC/DC specs) but I would prefer building this system with as much off the shelf hardware as I can because I may end up building several of them for family/friends.

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Sat May 21, 2005 9:55 am

Hmm, that 5W is way too much for these fans, even at startup.
According to papst, my model draws 0.5W when running. And I have undervolted them (actually I can't remember for sure if I went for 7V or 5V).
Startup is probably a bit more but I doubt it could be ten times as much as for running it.

Take any quiet 80mm fan and it's not going to eat 5W I think. To be quiet they need to spin at around 1krpm and don't need much power to do that.

CPU max draw according to intel specs is 19.1W

According to seagate, my drive uses 13W on seek/read/write and 9.8W on idle, with 2.8A required current on 12V line when starting up (ie. about 30W).

If we drop the display card and both nics then I guess we see at least 5W drop in idle, and at least as much for max load, giving us 35W idle and 55W on load.

As I never heard about 45% minimum load stability limit before, I'm using sth like a 250W-350W PSU but I don't really remember and opening the case is too hard after what I've done to it. In any case my system is very stable, and I've achieved uptimes of over half a year, only went down because of power outage or kernel upgrade. So I think it's safe to say my PSU doesn't have such stability limit, or it's way lower than 45%.

Sadly I haven't used the dc/dc psu I mentioned myself, I was going to but then I decided my current modded PSU is quiet enough anyway and it's not worth the money for me. On their website they are claiming that you could use a P4 with it.

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sat May 21, 2005 11:30 am

lm wrote:As I never heard about 45% minimum load stability limit before, I'm using sth like a 250W-350W PSU but I don't really remember and opening the case is too hard after what I've done to it. In any case my system is very stable, and I've achieved uptimes of over half a year, only went down because of power outage or kernel upgrade. So I think it's safe to say my PSU doesn't have such stability limit, or it's way lower than 45%.
If such a minimum load requirement did exist for PSU's, then most of the custom built PC's in the world would be unstable (which they are not).

45% may be a good minimum load percentage to achieve maximum efficiency for a given PSU, but the Seasonic S12 is more efficient at 25% load than most other PSU's at any load.

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

Post by fastturtle » Sat May 21, 2005 1:49 pm

LM:

As I stated the wattage I've listed is ball park for some of those components. As to the Thermal Envelope of the CPU, sure Intel does say it's 19.1 watts but to keep things simple I normally round up to whole numbers since I'm not building a satelite and have to account for every erg/watt/joule of energy consumed.

m0002a:

As to the 45% stability limit, how many people could tell if that windows crash was related to the PSU or software? I sure as hell couldn't tell you myself without having a hardware voltage monitor hooked to all rails constantly outputing to dedicated storage and without that, not even a good engineer could tell for sure unless he knew the PSU was oversized.

Fact is that I rarely see anyone else asking about the PSU unless they've indicated a major upgrade recently. Most times it's "is my video card going bad?" or My system is very slow, what can I do? To many people don't understand or even think about how much power they're system needs for stability and they're not aware that it's possible to have too much power and not run the PSU efficiently and that can cause as much problem as to much load for the system simply because unlike an underpowered car climbing a steep hill in the fast lane, you aint aware of it.

In my case, it's simply the desire to build not only a silent system but one that's as efficient as possible. Every little bit helps especially when the #&^%$ powrer goes out and I'm on the UPS once again, so a correctly sized PSU is going to extend the battery life of my UPS due to efficiency.

:idea: I think I've figured it out. LM Thanks for the info on that DC/DC PSU. What I'm going to do, is build a custom setup using a battery charger to keep a pair of 12 volt Gell Cells topped off and run the system using the DC/DC PSU from those 12 volt batteries. Consider it as installing a UPS right in the case.

mathias
Posts: 2057
Joined: Sun Jul 11, 2004 3:58 pm
Location: Toronto
Contact:

Post by mathias » Sat May 21, 2005 2:48 pm

fastturtle wrote: In my case, it's simply the desire to build not only a silent system but one that's as efficient as possible. Every little bit helps especially when the #&^%$ powrer goes out and I'm on the UPS once again, so a correctly sized PSU is going to extend the battery life of my UPS due to efficiency.
I've never even considered getting a UPS, so I don't know much about it, but I think I've heard that active PFC makes a huge difference with UPS's.
fastturtle wrote: :idea: I think I've figured it out. LM Thanks for the info on that DC/DC PSU. What I'm going to do, is build a custom setup using a battery charger to keep a pair of 12 volt Gell Cells topped off and run the system using the DC/DC PSU from those 12 volt batteries. Consider it as installing a UPS right in the case.
It does seem like a much better idea to have the UPS placed after the AC-DC conversion, but are you sure that the batteries will give stable enough power or the DC-DC PSU will regulate it well enough?

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sat May 21, 2005 3:52 pm

fastturtle wrote:As to the 45% stability limit, how many people could tell if that windows crash was related to the PSU or software? I sure as hell couldn't tell you myself without having a hardware voltage monitor hooked to all rails constantly outputing to dedicated storage and without that, not even a good engineer could tell for sure unless he knew the PSU was oversized.

Fact is that I rarely see anyone else asking about the PSU unless they've indicated a major upgrade recently. Most times it's "is my video card going bad?" or My system is very slow, what can I do? To many people don't understand or even think about how much power they're system needs for stability and they're not aware that it's possible to have too much power and not run the PSU efficiently and that can cause as much problem as to much load for the system simply because unlike an underpowered car climbing a steep hill in the fast lane, you aint aware of it.

In my case, it's simply the desire to build not only a silent system but one that's as efficient as possible. Every little bit helps especially when the #&^%$ powrer goes out and I'm on the UPS once again, so a correctly sized PSU is going to extend the battery life of my UPS due to efficiency.
Your claim that PSU's are unstable unless using 45% of the load is ridiculous and there is no documentation of such a phenomenon, and not even any serious anecdotal reports. If you Windows OS crashes frequently, you have other problems. My Windows 2000 or XP has only blue-screened two or three times in the last 5 years, and both times caused by an application problem.

Apparently you don't understand or don't agree with what I said earlier about PSU power draw. A given PSU may be more efficient at 45% load than at 25% load, but the Seasonic S12 has much better efficiency than the average PSU at any load factor. The S12-330 is the lowest power high-efficiency PSU that I know about, so it will draw less power than an average 200 watt PSU. I guess that I don't know any other way to explain it.

paapaa
Posts: 198
Joined: Wed May 04, 2005 1:24 am
Location: Finland

Post by paapaa » Sat May 21, 2005 4:26 pm

fastturtle wrote:Most ATX PSU's are designed with a minimum demand load for stability. This seems to be between 42-48 percent of designed maximum output.
How did you come to the above conclusion? Where is the source for this claim?

And what do you mean by stability: do you mean that the PSU starts to oscillate or do you mean that the voltage regulation fails below that level?

Can you post a link to a test that shows this behaviour?

madman2003
Posts: 182
Joined: Sat Jan 29, 2005 9:02 am
Location: Planet earth

Post by madman2003 » Sun May 22, 2005 1:31 am

Check some psu specs.

Phantom 350 has very low minimum loads: http://www.antec.com/specs/Phantom350_spe.html

I've seen higher requirements, but usually not higher than 1 A or so on each rail.

StarfishChris
Posts: 968
Joined: Fri Jan 07, 2005 7:13 pm
Location: Bristol, UK
Contact:

Post by StarfishChris » Sun May 22, 2005 2:24 am

paapaa wrote:And what do you mean by stability: do you mean that the PSU starts to oscillate...?
Aren't the screws supposed to prevent that?

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

Post by fastturtle » Sun May 22, 2005 6:05 am

To paapaa:

And what does a 1A load equate to on each rail? 3w/5w/12w; the simple method is to multiply the amps times volts and you get the watts. Now according to your post, Antec states a Minimum of 20w load for this PSU.

Now who can come up with figures for their chosen PSU? Lets put em in a list and see what all we come up with to verify/debunk the 45% minimum load rule of thumb.

Remember that a Rule of thumb is just that. It's not carved in stone. Instead it's a generally accepted design practice. Good example is Doorways. Why don't we build our houses with Round Doors as hobits do?Rule of Thumb.

Everyone seems to disbelieve the 45% stablity loading of a PSU:

Here's the explanation as I understand it - Note that I am not an engineer so some of it could be wrong.

An AC/DC transformer which is what a PSU is although it has several voltage taps, is designed for a specific minimum load. This is based upon the CPU, a common requirement of the Motherboard and a Minimum amount of system memory.

Originally, these figures where very low with a maximum output of 30-60 watts being considered high. Ck an old PC (pre 286)
General rule of thumb for PSU's was to design a minimum load requirement based upon these factors and they still do. This factor is generally 45% of the maximum design load for cost efficiency.

Using these generally accepted base values, reduces the design cost and complexity of a PSU for mass production. If you really examine the insides of most PSU's what do you see. Very little variation in the actual components. You may see some in the layout, heatsinks and such but you'll see very little difference in components between vendors because of mass production. The big difference you'll see though is quality of the transformer. It's this component that affects the entire quality of the rest of the unit. Does it have good windings for the conversion, are they using good qaulity connections to the wheatstone bridge for converting AC to DC?

Originaly PSU's didn't include active/passive PFC, That's only come around in the last 3 years. So originally PSU's had stability issues due to the simple fact that they weren't designed to filter their output. I will have to check the ATX specs for allowable voltage fluctuation but when you have a quality PSU that according to the Link provided by paapaa:

These are the acceptable maximum overvoltages before recycle of AC to reset:
+5V trip point < +6.2V
+3.3V trip point < +4.1V
+12V trip point < +14.3V

Here's the voltage regulation per rail from the same link:
±5% ±5% ±5% ±5% ±5% ±5%

With such a allowable swing in voltage regulation, I'd have to wonder just how stable this thing is going to be at minimum loads but these figures are actually pretty standard of most PSU's. What makes the difference is the listing of the minimum load current on each rail. I'm assuming that means the least it can provide.

Again from their Chart:
0.3A 0.4A 0.4A 0.3A 0A 0A

So if the PSU will not run unless it's putting at at least .9w on the 3.3 and 2w on the 5 and 4.8w on the 12 volt rails, I'd almost have to say that I've now been proven. What you will not find is the actuall specification for a 45% minimum load. This figure may and does vary and is never published. Once again, it's a rule of thumb and as such it's unstated but as those who use windows know, it's a rule of thumb to at least have antivirus on their computers.

StarfishChris
Posts: 968
Joined: Fri Jan 07, 2005 7:13 pm
Location: Bristol, UK
Contact:

Post by StarfishChris » Sun May 22, 2005 6:22 am

fastturtle wrote:Here's the voltage regulation per rail from the same link:
±5% ±5% ±5% ±5% ±5% ±5%

With such a allowable swing in voltage regulation, I'd have to wonder just how stable this thing is going to be at minimum loads but these figures are actually pretty standard of most PSU's. What makes the difference is the listing of the minimum load current on each rail. I'm assuming that means the least it can provide.

Again from their Chart:
0.3A 0.4A 0.4A 0.3A 0A 0A

So if the PSU will not run unless it's putting at at least .9w on the 3.3 and 2w on the 5 and 4.8w on the 12 volt rails, I'd almost have to say that I've now been proven. What you will not find is the actuall specification for a 45% minimum load. This figure may and does vary and is never published.
That's 7.7W, which is less than 3% of a 300W PSU.
Have you read the 80 Plus specification (there was an item about that on SPCR a while ago)? The PSU tested (a Seasonic) was over 80% efficient at 100%, 50% and 20%. That's 80W of the 400W capacity, a typical idle load, and you're trying to say it isn't stable? How many computers do you know that reboot or BSOD when you leave it doing nothing for a few hours? Everything you've said confirms a minimum stability requirement, but it's nowhere near your inflated figures.

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

Post by fastturtle » Sun May 22, 2005 6:43 am

For those who still doubt a 45% rule of thumb for PSU sizing:

Qestion: What happens when you upgrade a system to higher performance components but keep a weak PSU?
Answer: You end up with an unstable system.

Question: What happens when you o'clock a system w/o thought?
Answer: You end up with an unstable rig and that's what happens with a PSU that's oversized. It basically o'clocks the rig w/o you even being aware of it due to the voltages being high.

To many people think it's good to put the biggest/baddest PSU into their system and then wonder why it doesn't run stably.

That's why I use the 45% rule of thumb. I want a PSU that's large enough to handle the load and yet small enough that I'm not buying more then I need since I don't see any reason to go out and spend $300 on a PSU when a $50 dollar unit will suit my needs.

Devonavar
SPCR Reviewer
Posts: 1850
Joined: Sun Sep 21, 2003 11:23 am
Location: Vancouver, BC, Canada

Post by Devonavar » Sun May 22, 2005 9:37 am

fastturtle wrote:Qestion: What happens when you upgrade a system to higher performance components but keep a weak PSU?
Answer: You end up with an unstable system.
Yes, sometimes this happens. Often it doesn't. I don't think it happens enough to say that a power supply isn't stable throughout most of its range.
fastturtle wrote:Question: What happens when you o'clock a system w/o thought?
Answer: You end up with an unstable rig and that's what happens with a PSU that's oversized. It basically o'clocks the rig w/o you even being aware of it due to the voltages being high.
This is just wrong. If anything it would overvolt the motherboard, not overclock the CPU. But, the reality is that it will do neither. The extra voltage will simply be absorbed by the PSU circuitry on the motherboard, which steps down the voltage even further before they're fed to the CPU chip. Yes, out of spec PSU voltages can lead to variances in the voltage fed to the CPU, but keep in mind that most CPUs can tolerate a voltage range of almost half a volt. That's about +/-30%, assuming a nominal voltage of 1.5V. People have had systems stable down to below 1V (after unclocking as well), and, although I'm less knowledgable about overvolting, people do frequently add 0.1V in order to improve stability at higher speeds.

Having tested many power supplies personally, I can say that the voltage rails are almost invariably closest to their nominal output at lower current draws. Variance in voltage is almost always a function of sudden changes in load, not the size of it. With a low-power rig, the distance between idle and full load is likely to be much smaller than with a high power rig (say, 30W instead of 100W), so it should be less likely to cause instability than a higher power rig. Since most PSUs are designed for mid-to-high current draws, they should have no problem dealing with the smaller swings in load that your low power system will have.

wsc
Posts: 161
Joined: Sat Jun 21, 2003 9:25 am
Location: Oklahoma

Post by wsc » Sun May 22, 2005 10:09 am

fastturtle I would be very interested to read whatever material it is that has given you this impression about the 45% requirement. Perhaps several years ago the cheapest of the cheap-o PSUs needed a 45% load for stability, that would be much more believable. Modern high quality PSUs employ more advanced switching controllers and can easily handle light loads. There has been alot of progress in this area and unless your source of information is pretty new, it might best be taken with a grain of salt.

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sun May 22, 2005 10:19 am

fastturtle wrote:Remember that a Rule of thumb is just that. It's not carved in stone. Instead it's a generally accepted design practice. Good example is Doorways. Why don't we build our houses with Round Doors as hobits do?Rule of Thumb.

Everyone seems to disbelieve the 45% stablity loading of a PSU:

Here's the explanation as I understand it - Note that I am not an engineer so some of it could be wrong.
What we have here is what Plato called the difference between a true opinion and knowledge.

The 45% “rule of thumb” is probably reasonable (true opinion), but not for the reasons you stated. It is usually reasonable because most PSU’s reach maximum efficiency at about 45%, it provides sufficient reserve capacity, and specking a PSU higher than that is a waste of money. So with the 45% rule, you usually get less heat and use less electricity (because of higher efficiency) and save money.

But there is problem with this rule of thumb in this particular circumstance. The Seasonic S12 is more efficient at any load than the vast majority of PSU's at their most efficient load. So the Seasonic S12-330 (with active PFC) will use less electricity and generate less heat for your system than any lower power ATX PSU that I know about.

The lowest power version of the highly efficient S12 is the 330, so if you want to have an efficient, quiet, and cost effective PSU, then the S12-330 is your best choice.

Your statements trying to relate stability problems caused by overclocking, and overloading a system, with supposed stability problems (undocumented) using a PSU that draws less than 45% of available power, is intuitively incorrect to even the most technologically challenged posters on this forum.

Your statement that you are "not an engineer so some of it could be wrong" is the understatement of the year, even though we still have over 7 months to go.

paapaa
Posts: 198
Joined: Wed May 04, 2005 1:24 am
Location: Finland

Post by paapaa » Sun May 22, 2005 1:49 pm

fastturtle wrote:Now according to your post, Antec states a Minimum of 20w load for this PSU.
First of all, I didn't post that link. Where did you get the 20w figure? I get about 12w from the Antech specs (0.3A*5V+2*0.4A*12V+0.3A*3.3V). 12w is 3.5% of 350w. Not very close to your 45%...
fastturtle wrote:This factor is generally 45% of the maximum design load for cost efficiency.
Again, you just throw the figure with no data to show where you derived it from.
fastturtle wrote:Originaly PSU's didn't include active/passive PFC, That's only come around in the last 3 years. So originally PSU's had stability issues due to the simple fact that they weren't designed to filter their output.
PFC has nothing to do with output filtering. PFC tries to keep input AC voltage and current at the same phase thus maximizing the power factor (real power/apparent power).
fastturtle wrote:I will have to check the ATX specs for allowable voltage fluctuation but when you have a quality PSU that according to the Link provided by paapaa:

Here's the voltage regulation per rail from the same link:
±5% ±5% ±5% ±5% ±5% ±5%
Actually, ATX standard does specify basically (expect -12V rail) the above margins for voltage regulation. In addition, +12V rail can go from 10.8V to 13.2V during peaks. Yep, that is a 2.4V range. Check http://www.formfactors.org
fastturtle wrote: Again from their Chart:
0.3A 0.4A 0.4A 0.3A 0A 0A

So if the PSU will not run unless it's putting at at least .9w on the 3.3 and 2w on the 5 and 4.8w on the 12 volt rails, I'd almost have to say that I've now been proven.
Actually it just tells that Antec guarantees the PSU to give the performace described if the minimum conditions are met. It does NOT mean that the PSU "will not run" if less load is used. And no, you have not shown anything that proves the 45% figure. Again, the above equals to 12w load.

And your claims that stability issues are because people are using PSUs with too high capacity is just plain ridiculous. I think most PSUs will only show voltage drop at excessively high loads. Not voltage rise at very small loads. I've not tested many PSUs so I can't make generalizations but I've not seen any tests that shows the opposite.

Efficiency is a totally different matter. One should definitely try to get a PSU that has the best efficiency at load level that is close to actual usage. Some PSUs has the best efficiency at 90%l load, some at 50% load. Can't make any generalizations here either.

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

Post by fastturtle » Sun May 22, 2005 5:49 pm

There's been some good points raised by others, WSC you do raise a valid point about quaility improvements. LM your input has been appreciated.

The biggest problem is that I'm not making myself clear about the 45% rule of thumb. In over 20 years of computer use(1st pc in 1982), I've learned that any PSU that has a load of less then 45% it's stated rating could cause system instability and I haven't seen a good reason to violate a rule of thumb with 20+ years experience behind it.

I just used the PSU calculator from here and discovered that my projected power requirements has risen from the meager 120 watts I had been at (changed the mobo and needed a graphics card) to 185 watts. I have also dropped the case on advice that it's noisy and the PSU is junk so I can now justify the purchase of the Seasonic S12-330 as the 45% rule of thumb is 148 watts. I'm over that by simply using any PCIe card The only problem I now have is how accurate the calculator is and I'm taking the information with a Very Big grain of salt.

I'll try to answer everyone's negative points now.

paapaa:
In double-checking the webite, it looks as though you are correct. I misread the page.

Devonavar:
As I have never o'clocked a system, I can only say the evidence is heresay but I have had a system that was unstable due to a PSU that was oversized. Took the techs 3 months to determine it. What I saw happen was data corruption in memory that caused the system to crash. This was back in 1985 so it may not be as likely due to design improvement of the motherboard but why risk it when it's easily avoidable.

StarFish Chris:
Your question/comment about oscilation is correct. All PSU's oscilate around the core rail voltage. This is shown in the allowable variation/tollerence from standard of plus/minus 5%. What happens though is that the PSU's now have tighter voltage control so the oscilation is at a higher frequency, which could explain why the caps on some motherboards begin to whine/squeel.

On very early PSU's the voltage swings could reach levels high enough to actually damage a board. Some were recorded as high as 36 volts on the 5 volt rail. Certainly not good for the board. This is one of the reasons I am unwilling to change the 45% rule of thumb. Even though the new PSU's are higher quality, I don't want one to start oscilating that badly and although it's not likely to happen, it's a risk I'm not willing to take when a bit of planning can prevent it.

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sun May 22, 2005 6:42 pm

fastturtle wrote:In over 20 years of computer use (1st pc in 1982)...

<snip>

...On very early PSU's the voltage swings could reach levels high enough to actually damage a board.
We look at the present through a rear-view mirror. We march backwards into the future.
--Marshall McLuhan

paapaa
Posts: 198
Joined: Wed May 04, 2005 1:24 am
Location: Finland

Post by paapaa » Mon May 23, 2005 12:37 am

fastturtle wrote: What I saw happen was data corruption in memory that caused the system to crash. This was back in 1985 so it may not be as likely due to design improvement of the motherboard but why risk it when it's easily avoidable.
All I can say is that your information is totally out-dated. Your are talking about AT or even pre-AT days!!!! There is no point in comparing 1985 AT supply to a modern ATX/BTX PSU. Please read this page:

http://www.pcguide.com/ref/power/sup/ou ... ing-c.html

"Modern power supplies have drastically reduced the degree to which loading is an issue."
fastturtle wrote:All PSU's oscilate around the core rail voltage. This is shown in the allowable variation/tollerence from standard of plus/minus 5%. What happens though is that the PSU's now have tighter voltage control so the oscilation is at a higher frequency, which could explain why the caps on some motherboards begin to whine/squeel.
The regulation tolerances shouldn't have anything to do with oscillation. Oscillation happens, as you said, in higher frequencys. These stability requirements (stability has nothing to do with voltage regulation) is also covered in ATX design guide at Formfactors.

Regulation tolerances just specify the limits where the voltage should be with various loads.

StarfishChris: no, screws won't help, not even big ones ;-)

Fastturtle, of course you can use any "rule of thumb" you want. But you are limiting yourself seriously by basing your ATX PSU buying desicions on 20+ years old AT PSUs! Just try to forget the 45% figure as it is now more like 5% figure depending on PSU. And if you are even close to using only 5% of maximum output you are likely using the PSU at very low efficiency (very commonly efficiency is quite bad at very low loads). That just means the PSU is way too oversized. Most PSU manufacturers don't even specify the minimum load as that info has no practical meaning nowadays.

Just remember: "The world is changing..." :)

fastturtle
Posts: 198
Joined: Thu May 19, 2005 12:48 pm
Location: Shi-Khan: Vulcan or MosEisley Tattonnie

Post by fastturtle » Mon May 23, 2005 10:24 am

My entire purpose in starting this thread was to get feedback from others who have them for a unit that met my stated needs of X watts based upon my calculations that was also fanless or at least very quiet. All of the input has been apreciated, even the attempts to prove me wrong on some of my statements but No One ever proved that my 45% rule of thumb was invalid for the usage I made of it. Admittedly some of the information it's based upon my no longer be valid but that still doesn't destroy it's usefulness in planning/designing a system.

The only thing that I have been shown is several options as to manufacturers and designs such as the DC/DC psu, that could meet my originally stated needs and for that I thank everyone who's taken time to post. With the recomended changes on compenent choices that I have made in the spec thread, (opening post) I was able to follow my 45% rule of thumb without issue and select the Seasonic S12-330 PSU.

paapaa:
I appreciate the link you provided as it did update me on the changes in PSU design.

Prior to my current system (self built) all of my systems have been off the shelf. The only thing I bothered to keep up on was things like video cards along with scanners and printers and keeping windows stable. The only reason I'm even looking at building a new system is a growing need for increased storage and a growing desire for both less noise and energy consumption.

paapaa
Posts: 198
Joined: Wed May 04, 2005 1:24 am
Location: Finland

Post by paapaa » Mon May 23, 2005 3:24 pm

fastturtle wrote: All of the input has been apreciated, even the attempts to prove me wrong on some of my statements but No One ever proved that my 45% rule of thumb was invalid for the usage I made of it.
You made the next claim which you can't prove:
fastturtle wrote:Most ATX PSU's are designed with a minimum demand load for stability. This seems to be between 42-48 percent of designed maximum output.
Umm, the problem here is that the 45% figure (or 42%-48%) is nonsense. I'm totally sure that all the ATX PSUs you'll ever use will function perfectly at 45% load. But they will also function perfectly at 20% load. Many if not all will function perfectly even at 5% (and in that case the PSU is oversized because of bad efficiency at low load levels) as Phantom 350w does, for example. The 45% figure has no practical meaning at all. A few examples from specs:

The minimum required load of Antec True Power 550 or Neopower 480 is about 2.5% of max output.

Nexus NX-4090 has the next spec:
"The power supply will be capable of being operated with no load on any or all outputs without damage." That is 0% of max output.

ZM400B-APS:
Minimum needed load is 0.4% or 1.6w from 400w.

We are not even close to 45% or even 10%. So you really don't have to worry about using your PSU below 45% load.

<a bad analogy>
Someone could say: "Don't let your car engine exceed 1200 rpm. Otherwise it will get damaged."

The rule works but has no meaning nowadays as modern engines can probably handle more rpms as the engines at the beginning of 20th century (with no limiters).
</a bad analogy>

Use a rule you wish, but keep it a secret :lol: (I think this issue is now cleared)

Locked