Talk:ATX/Archive 1

BTX
The article claimed that ATX will eventually be replaced by BTX. While this is Intel's intention, their plans don't seem to be going very well so unless you've got a crystal ball, I think it's a little unreasonable to makes this claim

The advance of atx power connectors
The atx form factor was created in 1995 and completely out ran the ordinary at supply.The atx was a solution to all the disadvantages and problems that were caused by the at power supply.The atx power supply has a switch in the back to make sure that no power is flowing to the motherboard and this is also a safety precaution.

wikified
Doesnt this need to be wikified? And isnt it a stub?

more
what about explaining what something like "atx 12V v1.2" means etc. Great start but could do with more


 * I'd certainly love to see a bit more info on power supplies, etc. I see PSUs described as "PSII ATX" or "PSIII", and have no idea what this means :) Steved424 09:51, 4 February 2006 (UTC)


 * And I see things like "ATX 2.2" when picking out a PSU, and I don't know what that means. --BennyD 23:41, 7 July 2006 (UTC)

Baby AT and PS/2
I am positive that some of the Intel motherboards, perhaps OEM-only models, had PS/2 ports. Perhaps even back to 486 times. The majority of PS/2 on Baby AT boards came on a bracket, however. --Swaaye 02:09, 8 July 2006 (UTC)
 * PS/2 mouse ports were seen on quite a few (i think less than half though) of baby AT machines on a bracket but i don't think i've ever seen one that didn't have 5 pin din keyboard port. I have seen some (pre ATX) big brand machines with two PS/2 ports but i don't think they were baby AT form factor. Plugwash 22:05, 8 July 2006 (UTC)

Corrected btx link
Updated link for BTX since the page has been moved. --  RND   T  C   17:33, 27 July 2006 (UTC)

Extended ATX
Anyone up for making an article or section on Extended ATX? Jay Kana 14:40, 6 October 2006 (UTC)

Old Power supply pictured
The I think a picture of a power supply with the new 6 pin PCIE connector and without the 3 pin fan connector (new power supply fans are not controlled by the motherboard) should replace the one currently listed.

Airflow
I believe there are ATX power supplies with different airflows:
 * From top to front (fan)
 * From back to front (fan)
 * From front to top (fan) to CPU

Would this warrant a discussion?
 * In the older ATX specs, the power supply was designed to actually take air *into* the case and exhaust it onto the CPU. When processor power dissapation became so great that they required their own CPU coolers, the issue became moot and power supply airflow direction changed to exhaust the case.  N Yo FACE 00:34, 23 April 2007 (UTC)


 * Around the turn of the century I made it my standard practice to flip over the fans of all ATX systems I worked on so the power supply exhausted air from the case. If the CPU didn't have a fan, I installed one. Internal openings on the power supply, other than for the fan, got blocked with aluminum foil "flue tape". The measured results were around a 10 degree Fahrenheit reduction in temperature inside the case. With higher performance CPUs (that already had a large CPU heatsink and fan) the temperature difference could be felt on the outside of the case. The backwards airflow power supply was a stupid idea from the beginning- perhaps it made some sense with the CPUs at the *start* of the ATX design process, but the concept wasn't practical with the CPUs being used when the ATX standard was finalized.


 * Some OEM systems compensated for the back-asswards power supply fan by mounting a large duct over the CPU heatsink- with a 4" exhaust fan. Flipping both fans around resulted in lower case temperatures. —The preceding unsigned comment was added by Bizzybody (talk • contribs).


 * I have removed the section on airflow *replaced


 * "...drawing air into the case and exhausting it down onto the motherboard" implied that all ATX chassis were towers with PSU mounted top back. I don't believe they were.


 * It may be the case that some early ATX PSU's had reversed fans, but Intel's NLX and ATX design suggestions from a similar era recommend a PSU fan exhausting air from the chassis.


 * See airflow diagrams at:


 * NLX Thermal Design Suggestions v1.0 - Intel Corporation, May 1997 nlxthermv1.pdf, page 7


 * NLX Chassis Design Suggestions v2.0 - Intel Corporation, Novemeber 1998 nlxchassisv2.pdf, page 65


 * ATX Thermal Design Suggestions v1.0 - Intel Corporation, March 1999 atxthermv1.pdf, page 28


 * and they refer to negative pressure being prefereable:


 * "...using a fan to depressurize the chassis produces significantly greater cooling than does using the same fan (or two fans in series) to pressurize the chassis."


 * NLX Thermal Design Suggestions v1.0 - page 5




 * "Key considerations when selecting/designing a power supply:


 * Evacuate the chassis (rather than pressurize it) with the power supply fan."


 * ATX Thermal Design Suggestions v1.0 - page 19






 * http://www.keepandshare.com/doc/view.php?id=1499003&da=y


 * http://www.keepandshare.com/doc/view.php?id=1499002&da=y


 * http://www.keepandshare.com/doc/view.php?id=1499001&da=y




 * Can someone provide a PDF of the original ATX specfications before adding that back in? We need a verifiable citation.


 * Laptcd (talk) 06:20, 31 October 2009 (UTC)


 * I have replaced the section on airflow and added a new citation.


 * The excerpts from the old content:


 * "Early ATX systems simply didn't have processors or components with thermal output that required special cooling considerations."


 * and


 * "Later ATX systems with significantly greater heat output would not be aided in cooling by a power supply..."


 * and


 * "...modern ATX power supplies universally exhaust air from the case."


 * would have required citations I don't think we'd find. (For example, passive power supplies are not unheard of).


 * Further, a quote from the previously cited page:


 * "At around the time ATX was introduced, cooling fans were becoming the standard for the newer, faster CPUs on the market."


 * conflicted with the first of the excerpts I've referred to above.


 * Hopefully the new wording is cleaner and removes the question of whether all early ATX chassis were towers.


 * Laptcd (talk) 09:48, 5 November 2009 (UTC)

New Formfactor: DTX
http://www.theinquirer.net/default.aspx?article=36833 Released by AMD. --Satsuki 06:43, 12 January 2007 (UTC)

Shorting a power supply
FTA: ''Because the ATX PSU uses the motherboard's power switch, turning on the power in situations that do not utilize an ATX motherboard is possible by shorting the green wire from the ATX connector to any black wire on the connector (or ground). This trick allows easy re-use of an old PC power supply.''

I've heard that this can be very dangerous (as well as fatal to your hardware). Can anyone clarify? Nemilar 01:19, 2 February 2007 (UTC)


 * No shorting a psu in this manner is one of the main ways to add a second psu for the purpose of running the highest end graphics cards or water cooling units

[[User:Mdogma]10:40, 8, Mar 07]
 * I'd think it highly unlikely this would be dangerous but you have to be carefull about underloading if you don't want to damage stuff. Some PSUs start to lose regulation if certain rails (mainly the 5V rail afaict) are not sufficiantly loaded. Plugwash 21:12, 30 March 2007 (UTC)

If you mess around and don't understand electricity and circuits, you could hurt the equipment or yourself. To correctly trick an ATX PS to turn on, you are shorting a pin that is *not* a power voltage. It does not matter what color the wire is. What matters is that it is the right pin (pin 16 on a standard ATX). If you trick it on, and you don't have it loaded right, anything could happen. Most likely, it will quickly shut down, but you could get wrong output voltages. Don't electrocute yourself. Don't burn yourself. Have fun... -71.174.189.6 (talk) 04:03, 14 May 2009 (UTC)

Display bug
When i [hide] the menu 'Contents', the first picture and the [edit] link don't move, but the text and the second image do. So there is a bad mix (Firefox 1.5 Win XP SP2)

Five power supply designs, or seven?
The article says "The ATX form factor has had five main power supply designs throughout its lifetime" but the list that follows has seven entries, because of the three variants of the ATX12V. This is a bit confusing. I'm moving the EPS12V up one position, so that the three ATX12Vs are in a row, and indenting the list before the ATX12V 2.0 and 2.2.

This is probably not the best solution, maybe someone has a better idea. — Graf Bobby 09:45, 4 April 2007 (UTC)


 * I guess the five main designs are ATX, ATX12V 1.x, ATX12V 2.x, GES and EPS12V. The minor revisions within each of the major versions of ATX12V don't really have much practical significance. Plugwash (talk) 14:05, 19 November 2009 (UTC)

Extended ATX?
Can't find any info for that, the Form Factor used by AMD Quad FX platform as in ASUS L1N64-WS Pro, where is the infor for that, besides a redirection from Extended ATX to ATX?

What a crappy specification
every revision has a different PSU connectors, not every case manufacturer states which ATX revision they use, and even when they claim the case should come with ATX 2.0 PSU (24+4), it comes with another, old one (20+4). And Asus states in the manual that the motherboard won't even boot without ATX 2.0 supply. of course all these cases are "just ATX" :-( QuestPC 19:45, 18 May 2007 (UTC)


 * ATX 2.0 isnt the same as ATX12V 2.0. The latter is described in this Wikipedia page. ATX 2.0 is older, and probably has just 20 pins. http://www.pcguide.com/ref/power/sup/partsMotherboard-c.html That page describes old ATX, and says that "their" revision is ATX 2.03, which means ATX 2.0 is even older. I agree, this is confusing, especially if the people in the business trying to sell things dont know what theyre talking about >:( 81.232.32.135 15:06, 20 August 2007 (UTC)

Another form factor
My old motherboard MSI MS-6309, a part of my old Fujitsu Siemens computer, has another form factor, not documented here. It is 304 mm high, just like ATX, but 192 mm wide. Quite confusing! The power supply connector is an ordinary "old ATX" connector, with a single 20-pin connector. What do you think it is? ;-) 81.232.32.135 14:58, 20 August 2007 (UTC)

Tabs, the unforeseen connector issue
PSUs with 24-pin connectors typically have an enlarged tab that is centered across the 24-pins (and is therefore off-center in a 20-pin socket). This introduces an unexpected large bulk on that side of the power connector. This can make it impossible to plug into some older motherboards because of the placement of capacitors or other objects near the socket. One example is the Abit NF7-S2. If you are buying a PSU for an older system you need to be aware of this. Unfortunately the most cost-effective solution (old, good PSUs aren't any cheaper than new ones, if you can even find them) is probably to buy a 24-20 pin adapter. Even though an adapter is technically unnecessary, the 20-pin plug on the adapter will usually have the original skinny tab. You can also cut the tab off but doing this will void warranties. Ham Pastrami 10:24, 7 November 2007 (UTC)

Fact check needed
The reference I've been staring at say ATX12V 1.3 has a 6 pin tertiary connector. Article lists it as 8 pin for all ATX12V. Think details like this really have to be checked and re-checked. —Preceding unsigned comment added by Clearthought1000 (talk • contribs) 06:02, 10 November 2008 (UTC)

PCIe power connector
Does anyone know what if any spec defines this connector, it certainly doesn't seem to be mentioned in the intel specs linked from this article.


 * It isn't part of the ATX specification. It is not surprisingly covered by the PCI express specification.  See http://www.pcisig.com/specifications/pciexpress/pcie_cabling1.0/ for the official standard.  Unofficail guides are also up online - I suggest taking a look at the external links section of  PCI Express in the first instance. CrispMuncher (talk) 16:07, 5 December 2008 (UTC).

Wasted energy / Critic?
This article seems to glorify the "wake on lan" and the like, while in reality they are rarely used. That means that the ATX switch simply replaces the "true" switch of AT PSUs, so it adds exactly zero benefit while, as stated in the reverted version, that it simply wastes power for (alomost) everyone (I doubt that even 10% of all ATX PSUs get truly switched off at all, if they even have a proper switch, and if the owners even know the difference). Given the increasing energy shortage, this, and the lack of easily accessible true switch, is a major design flaw which should be mentioned. —Preceding unsigned comment added by 217.230.54.246 (talk) 22:20, 13 January 2009 (UTC)
 * I presume that you are the same user who added text to this effect a couple of weeks ago and which I reverted. I did that because your edit was fundamentally POV and ill-informed, not because this is non-notable.  If you think this should be covered then add it, but ensure that your material is accurate and neutral.  Wake on LAN is used extensively.  Maybe not in the home environment so much (although I certainly use it here), but let's face it, most home networks are not particularly well developed and only exist to share Internet access and possibly files and printers.  Where it comes into its own is in the large corporates with thousands of machines that need to be remotely manageable overnight.  Without wake on LAN users in such environments were told simply to leave their machines running at full power, and this is still the practice at some places.  Wake on LAN allows the power consumption to be dramatically cut in that instance.  Which uses more power?  A few home machines using a maximum of 10W (which is all +5VSB is good for) when not in use or thousands of business machines using (say) 300W for the 16 hours a day when they are not in use?  There is a case to made that ATX "wastes" power in the manner you describe but you have to be careful how you frame your argument. CrispMuncher (talk) 22:58, 13 January 2009 (UTC)
 * Thank you for the detailed reply. Yes, your presumption is accurate, I am that user; after posting the comment I have considered to amend it to state that fact clearly, but then I figured that'd be OK to leave to this reply and not clutter the edit history with a minor edit (given this is not an article, which option is preferrable?). I am however not certain about the level of ill-informedness you declare. I was (and still am, mostly) under the impression that WOL has even less use in companies than wake on modem, which would allow (in theory) to have dial-in servers for rarely used services turn on only on demand. Regardless of such nit-picking, I also was under the impression that in companies the machines either run 24/7 (servers for making back-ups or because of outsourcing to overseas countries or subsidaries), are located in datacenters, likely on virtual servers (and therefore uncontrollable), or are serviced during working hours by on-site staff (employee workstations). Therefore I do not see WOL as something that would find much use outside the hobbyist sector (webcam or other private-only servers), so I am not convinced that the amount of power saved in corporations even matches the amount of power wasted in private homes.


 * Anyway, as said above, I doubt that many people even know that their computer is still using power when they believe it to be completely off. Of the ones who do know, most probably do not care or deem it significant - after all, the VCR also runs 24/7. I am guessing that, on average, every person in america owns one computer (not counting the office computers that likely also never get powered off completely, or are these also being remotely administered overnight?). If they are using 10W (given that by default almost everything is powered for wake-on-anything), that makes for 306,134,000 computers, using 3,061,340,000W (wow, 3 GIGA Watts! That surprises even me) for the (estimated) 16 hours they are not used daily, ignoring vacation. I have no data to guess how much energy might be saved by corporations using WOL.


 * Assuming that a corporate WOL computer actually draws 300W (not sure they do, they don't use high-powered GFX cards but use high-powered disks), then to save the 3 GW it has to be 10,204,466 computers (not accounting for the remaining 10W).


 * Thinking about it this way now, one could even believe that this is another standard that gets corporations a little savings while burdening everybody else with cost, but this os, of course, pure speculation.


 * Lastly, I pull the plug on all my computers and servers (even the switches) when they are not going to be used for a while. 217.230.63.78 (talk) 04:16, 15 January 2009 (UTC)

EPS12V - Opteron/Xeon-specific?! I doubt it...
Most higher-end PSUs sold today are EPS12V, and almost all modern motherboards have EPS12V connectors. Nevertheless, the article lists EPS12V as a rarity used only be Opterons and Xeons, which I highly doubt (given that my Core 2 system is in fact EPS12V, and so is every single Core 2 system I've seen so far). --NetRolller 3D  21:50, 6 February 2009 (UTC)


 * Article is out of date. WP:Be bold. — Aluvus  t/c 01:49, 7 February 2009 (UTC)


 * I think the EPS12V's 8 pin secondary connector is a rarity, used only on dual CPU systems rather than the 4 pin CPU connector. I had a EPS12V spec PSU in a server fail on Friday and no other machine (including Core 2 systems) with the exception of another dual CPU (2 Core 2 chips) had that 8 pin connector.  It may be that most power supplies come with that, or at least with two 4 pin connectors that can be combined to make the 8 (speculation based on hunting for a replacement today).  I'm quite prepared to be wrong (and unlucky in my searching) but certainly nothing within screwdriver reach had that 8 pin connector which seems to be the one connector that is unique to EPS12V sjwk (talk) 14:54, 9 February 2009 (UTC)
 * Looking at a couple of core 2 compatible boards one high end and one low end (most popular 775 board on dabs and cheapest 775 board on dabs respectively) the high end one seemed to use the 8 pin connector. The low end one seems to use a 4 pin connector. Many PSU makers do indeed provide a pair of 4 pin connectors rather than one 8 pin for wider compatibilty.
 * P.S. If you get really desperate for a PSU with the correct connectors for your board there is always wire cutters and terminal block or solder and heatshrink ;)
 * P.P.S check which connectors are which on a PSU carefully. The 6 and 8 pin PCIe power connectors have the 12V and ground the opposite way round from the motherboard 12V connectors. While keying is supposed to make them non-interchangable I wouldn't rely on it. Plugwash (talk) 21:57, 9 February 2009 (UTC)


 * By far most Core 2 boards are still ATX. Only a few of the higher-end models have the 8-pin connector, which is required to be considered EPS.  Most EPS boards are still server boards.  It does seem to be true that most higher-end PC power supplies support EPS, but it's still necessary to specify EPS as a requirement when buying cases for a server board, lest you get an incompatible PSU. Thanny (talk) 20:52, 17 February 2009 (UTC)

My motherboard has an 8-pin connector, and is considered ATX. But a 4-pin *should* work fine in these types of motherboards. —Preceding unsigned comment added by 69.81.154.218 (talk) 23:32, 10 April 2009 (UTC)

The "width" should rather be called "hight"
I think, that ATX is confusing, because most cases used to day are towers. So it is more usefull, to call the 305mm "hight" instead of "width" and the 244mm "depth" instead of "lenght". --MrBurns (talk) 16:13, 19 April 2009 (UTC)


 * The usage in this article is even stranger when you consider that the ATX spec refers to "width" and "depth" (not "length"). The spec uses "height" to refer to how far things are above the plane of the motherboard (the I/O area is 1.75 inches tall, for instance).  In any event, the article should probably provide some sort of explanation for whatever terms are used. — Aluvus  t/c 18:07, 19 April 2009 (UTC)


 * This is a convention that goes well beyond computer motherboards. Look through any electronics components catalogue and you will find countless reference to height meaning "above the board" or even overall height taking into account below the board mountings.  It is a convention so well established that we should not think of changing it, particularly as it is one that the ATX specifications themselves follow. CrispMuncher (talk) 15:20, 14 May 2009 (UTC)

+12V Rails
The article says that current ATX power supplies are supposed to have "two independent 12V rails (12V2 for the 4 pin connector and 12V1 for everything else) with independent overcurrent protection". Current standard units for sale (newegg) actually offer 1, 2, 3, or 4, 12V rails. The article should explain this detail, and how it is implemented. Does this just mean extra 4 pin connectors for each rail?

The article says "it was determined that it would be much easier (both from economical and engineering perspectives) to power most PC components from 12V rails". This should be explained, and or linked to details. I assume it means that on-board regulators are used to somehow produce local lower voltages at higher currents? Even for the cpu core voltage? Otherwise, it would be very inefficient! How is this done? -71.174.189.6 (talk) 23:19, 13 May 2009 (UTC)


 * For safety reasons, a single rail is not supposed to carry more than 240 VA (which for a 12 V DC supply, means 20 A of current). The ATX12V Power Supply Design Guide recommends putting the 4-pin connector for the processor on its own rail, with its own over-current protection.  I don't have a copy of ATX 2.0, but ATX 2.2 does not describe any of this in the standard itself (as stated, the Power Supply Design Guide discusses the topic).  In practice, manufacturers have implemented a variety of solutions, but ultimately most have ignored (to some degree or another) the notion of multiple rails with indpendent OCP.  It is more convenient to carry 12 V around the system (higher voltage means lower current for a given DC load, and lower current means less heat and thinner wires).  Voltage is then regulated down by the receiving device. — Aluvus  t/c 00:58, 14 May 2009 (UTC)

If 12V is "regulated down" in a normal (resistor and series-pass transistor) way, to under 6V, over half the power would be wasted! That would be stupid, and not modern. Something fancier must be happening here. What, exactly? (I think it has to do with inductors, buck/boost circuits, etc -- if someone who understands these matters stumbles on this sentence, please educate us with good links etc.) -71.174.189.6 (talk) 04:11, 14 May 2009 (UTC)


 * See Voltage regulator and Voltage regulator module. — Aluvus t/c 04:55, 15 May 2009 (UTC)

20+4 Pin Main Connector
Current standard units for sale (newegg) sometimes offer "20+4Pin" main connectors. Does this mean the standard 24 pin connector is split into two, for use with both old 20-pin and new 24-pin mainboards? Please add info about "20+4Pin" to the article.

And, would such units be likely to offer higher currents on the +3.3V and +5V rails, so they would actually be usable with older mainboards? -71.174.189.6 (talk) 23:24, 13 May 2009 (UTC)


 * Yes, part of the connector can be detached for easier connection to older motherboards. There is no consistent difference in the power that the supply is able to deliver.  Nearly all modern ATX power supplies are built on the assumption that the greatest load will be on the 12 V rail. — Aluvus  t/c 01:00, 14 May 2009 (UTC)

Test ATX or build power supply with old ATX PSU
If you want to test an ATX PSU or power something else with one, you'll need to trick it into turning on by grounding pin 16. But you also need to supply proper minimum load.

This source says "The ATX spec says that if minimum loads of 0.3A for 3.3V and 1.0A for +5.0V exist, all ratings must be met." In order to do that, use a 10 ohm resistor from 3.3V to ground. Use two 10 ohm hi-power resistors in parallel on the 5V to ground; each will dissipate 2.5 watts, they will get hot! That source also has good pictures and instructions for testing.

These sources have instructions for using an old ATX PSU to build a general-purpose power supply:
 * http://www.instructables.com/id/Yet-Another-ATX--Lab-Bench-Power-Supply-Conversion/
 * http://letsmakerobots.com/node/1355

-71.174.189.6 (talk) 04:46, 14 May 2009 (UTC)

"A modern ATX supply has minimum loads of about 1A on the +12 supples, 0.3A on the +5V, and 0.5A on the +3.3V. The -12V and the +5V standby ouput have no minimum current. These figures can be used as a general guide if the exact specifications are unknown. I used 12Ω, 3W sand power resistors in parallel as a load for +5V and +3.3V, which gives currents of 0.83A and 0.55A... For +12V, I used a 20Ω, 10W sand power resistor, which gives a current of 0.60A. These resistors were soldered to 7-terminal tie point strips... The power dissipation is less than the ratings, but the heat will be present, so the resistors should be supported in the air and not touching anything." (J. B. Calvert, Created 8 September 2006)  -96.237.2.113 (talk) 14:30, 14 May 2009 (UTC)

Most computer power supplies seem to be labeled with how much current each voltage can supply. The article should include a picture of such a label. But the labels don't usually fully specify the ways the outputs interact: it may be possible to draw more current from some outputs when drawing less from other outputs; the article should at least mention this as part of a thorough discussion of the typical Total Wattage rating, which is actually the most common way computer PSUs are rated.

It would be good to include a full set of output specs for a typical PSU, comparing Original ATX, ATX12V 1.x, and ATX12V 2.x. And common Total Wattage ratings -- from under 200W to over 500W output? What were and are common efficiences, with light-typical-heavy loads?

Also, if someone knows who the biggest actual makers of ATX PSUs were and are, and can back that up with references, would be good addition to the article. The are probably all in Asia? -96.237.2.113 (talk) 12:56, 14 May 2009 (UTC)

20-pin vs. 24-pin numbering change
Note that pins 11 to 20 on the ATX 20-pin connector are THE SAME PINS as pins numbered 13 to 22 on the ATX 24-pin connector. So, to turn on this PSU ground pin 14 on 20-pin or pin 16 on 24-pin, which are actually the same pin. The article should include a table for the 20-pin connector to clarify this. -96.237.2.113 (talk) 14:58, 14 May 2009 (UTC)

Efficiency
For reliable operation, it is important that the power supply be "big enough" for the application, not overloaded. It is tempting to think "the bigger the better", but this is not entirely true. Bigger will have a better margin of safety, should run cooler, and might run quieter. Bigger costs more to buy, might be physically bigger, might not fit. There may be a trade-off with efficiency. Usually a power supply will be more efficient at the upper end of it's design range. So an oversized PSU will be more lightly loaded, compared to its rating, and may be less efficient than a more appropriate small-but-adequate supply. -96.237.2.113 (talk) 14:59, 14 May 2009 (UTC)


 * Peak efficiency for a PC switching power supply is typically at around 50% of rated output. — Aluvus t/c 04:51, 15 May 2009 (UTC)

PSU Fan Noise
There can be considerable variation in the amount of noise from an ATX PSU, due to the fan. This is something to consider when buying. Some fans may be variable speed, only running faster (and noisier) when higher power and increased cooling are need. -96.237.2.113 (talk) 15:00, 14 May 2009 (UTC)

All rotating cooling fans in computers will fail eventually -- depending partly on quality (ball-bearings or not) and the amount of dust in the air. As the bearings fail, the fans often become noisy loud etc. This is a sign of impending doom. If the fan seizes up and stops turning, the noise will stop, the cooling will stop, and something will overheat and fail. It might just break, or there could be smell, smoke, or even fire.

Most of the fans in computers use either 5VDC or 12VDC, and are somewhat standard and interchangeable. Some are two-wire, and some are three-wire, with an extra connection to monitor the fan speed RPM. Even the fan inside a PSU may be somewhat standard and replaceable, with care. -96.237.14.189 (talk) 19:09, 14 May 2009 (UTC)

higher quality connectors
From the article: "Another minor revision, the main change was a call for higher-quality connectors on the motherboard power connectors." Can anyone give any more details on this and/or a citation to back it up? I didn't see anything obvious when I looked. Also are the "power supply design guidelines" the official standard or is there another standard somewhere? I note that the original edit that introduced this comment was made by an anon (http://en.wikipedia.org/w/index.php?title=ATX&action=historysubmit&diff=267513221&oldid=267511212) Plugwash (talk) 02:10, 22 November 2009 (UTC)


 * ATX 2.2 introduced the 24-pin main connector to replace the 20-pin main connector. This is one of the "key changes" noted on FormFactors.org and also appears is the revision history of the standards document itself.  This is probably what the previous editor was getting at.  The primary document of the standard is the ATX Specification v2.2; the "Power Supply Design Guide" is just a series of suggestions on implementing a power supply that complies with the standard. — Aluvus  t/c 05:42, 22 November 2009 (UTC)
 * Thanks for the link, however it appears from that link that ATX version numbers and ATX12V version numbers are different. "include Main Power Connector changed from 20 pin to 24 pin ( 2 x12) to support PCI-Express* requirements and removed Aux Power Connector Recommendation if using a power supply designed using ATX12V Power Supply Design Guide Rev 2.0 or greater. ". still doesn't explain where the higher quality connectors thing came from though. Since it seems the spec is freely availible i'm going to take a look to see if that reveals anything. Plugwash (talk) 00:27, 23 November 2009 (UTC)
 * I've taken a look at the ATX spec and it is clear that ATX12V and ATX use seperate version numbering, still not too clear if the "power supply design guides" are the only specs for ATX12V or not. I'm removing the unsubstanciated comment on higher quality connectors from the article. Plugwash (talk) 00:34, 23 November 2009 (UTC)

Original fan direction suggestion, source
From ftp://download.intel.com/design/motherbd/atx_201.pdf [note: dead link, but file readily available elsewhere]:

spec dated Feb. 1997

"The intended location and fan direction in an ATX system is for the power supply fan to draw in cool air from outside the chassis and exhaust it directly onto the processor. In this configuration, cooling of the processor without the need of an active fan heatsink (heatsink with small fan mounted on top) is achievable in many cases (see section 4.5 for a complete discussion of power supply airflow)...

The preferred airflow solution is to pull air through the power supply from outside the chassis and direct it onto the processor."

Some old ATX power supplies did follow this suggestion and suck air into the case. This suggestion was reversed in later versions of the spec. Mirror Vax (talk) 19:22, 31 May 2010 (UTC)