DFI sli-dr and 20 pin power supply. How many are using 20pin ??

**EMC2** · 05-07-2005, 12:58 PM

Not going to single out posts... just add my pennies to the pile

1) A 24-pin PS that has seperate wires going all the way back to the PCB in the PS IS a 24-pin PS. It is not a "re-treaded 20-pin supply".

2) On the issue of seperate +12V rails - this is about compliance with certain governmental regulations in various locals and has nothing to do with a PS being a 24-pin PS nor with it meeting the requirements of a ATX 2.0x supply.

3) The extra 4 pins on a 24-pin PS is not just about current carrying capabilities and/or extra pins. It also involves issues of parasitics in the power distribution system for your MB and add-in cards... including IR drop, inductance, frequency response, and noise. It also has to do with compliance to the ATX 2.x specs. And it requires seperate connections all the way back to the PS.

4) Based upon my own design experience, there are valid reasons why DFI is requiring true 24-pin PS's. See comment 3.

5) It is possible with a good 20-pin PS to create the equivalent of a 24-pin PS using a 24 to 20 pin adapater - it's done by cutting the wires between the extra four pins of the adapter at the point they join the standard 20-pin power pins and connecting the now unconnected 4 extra pins' wires all the way back to the PS's PWB with their own seperate wires. One way to do this is to use one of the SATA power connectors, which has all 4 required connections (+5,+12,+3.3,Gnd) and seperate wires all the way back.

6) I initially ran my DFI NF4 with a 20-pin PCPnC 450ATX supply without any problems on my bench (although in a stripped system configuration). But... it did hinder the performance. Whether or not a 20-pin PSU can work depends on the PS, what other components are in the system, and how close to the edge you intend to push your system.

So in regards to can you use a 20-pin PSU... given certain conditions, yes. Will it provide you with the most stable, overclockable system without at least modification as discussed in comment 5....NO. Does DFI know what it's talking about making a 24-pin PS a requirement...YES! For much the same reasons, plugging the extra MB power connectors into your PS is also recommended.

By way of a recent experience (one of many, many, many dealing with power and power distribution issues I've seen), maybe you'll understand DFI's stance on this issue....

On a project I'm working on at work, we had performed exhaustive testing on some new hardware and it had performed flawlessly in our development test environment. When we did a manufacturing run of the item in question, they were "failing" during manufacturing testing. I ended up during a 72 hour period, with only two 3 hour stints of sleep intermixed, in the manufacturing facility tracking down WTH was going on (and then lending a helping hand to testing to meet our deadline). To make a long story shorter, the problem ended up being someone who thought they could disregard the guidelines I had given regarding the power distribution in the test setup and used single wire connections between the test station power supplies and the test adapter for power distribution, along with improper power distribution inside the test adapter. The net result was perfectly good products with excellent margins "failing" test and my aforementioned grueling 3 days isolating, identifying, proving, and fixing the test setup they were using so that it conformed to the initially stated requirements. Afterwards the batch of product passed all testing with flying colors and substantial margins and are now being happily used by the customer

Oh... one last comment... it is *possible* under certain conditions for what the OCZ rep said to occur...

**bachus_anonym** · 05-07-2005, 01:26 PM

I knew you will sooner or later find this thread, Daniel

I literally was just waiting for you to step up as soon as I found that you posted few comments today

As I will not argue on what you say here (not even cloooooose in knowledge on that matter as you are), but you have to admit one thing....

There is just a whole bunch of folks running even SLi systems on DFI with 20-pin PSUs. Even one buddy of mine runs water-cooled, OCed up his &%$ 2x6600GT cards, with Winnie @ 2800MHz, couple HDDs, DVD-ROM and CD-burner... All with 480W 20-pin Antec, but with all connectors plugged-in to the mobo.

Like I said, I would gladly get 24-pin PCP&C like the one you previewed (excellent info there!!!) but as of now I just cannot justify that as of yet. I might swallow what I just said, after I kill sth, but I would not jump the conclusion so quickly attributing it to 20-pin

Great to see you dropping by here with some great input :thumbsup: Thanks!

**EMC2** · 05-07-2005, 02:06 PM

Originally Posted by bachus_anonym

I knew you will sooner or later find this thread, Daniel

I literally was just waiting for you to step up as soon as I found that you posted few comments today

You voyeur you

Not a matter of finding per say... been busy and not around here... have some time today in between other things

As I will not argue on what you say here (not even cloooooose in knowledge on that matter as you are),

...nothing wrong with discussing things

but you have to admit one thing....

There is just a whole bunch of folks running even SLi systems on DFI with 20-pin PSUs. Even one buddy of mine runs water-cooled, OCed up his &%$ 2x6600GT cards, with Winnie @ 2800MHz, couple HDDs, DVD-ROM and CD-burner... All with 480W 20-pin Antec, but with all connectors plugged-in to the mobo.

See my comment 6, along with 1st & last sentence of the paragraph which followed

Like I said, I would gladly get 24-pin PCP&C like the one you previewed (excellent info there!!!)

Now, now, quit brown-nosing...

<--- Bachus

You could take comment 5 under consideration

but as of now I just cannot justify that as of yet. I might swallow what I just said, after I kill sth, but I would not jump the conclusion so quickly attributing it to 20-pin

Don't worry bro, I never jump to conclusions... notice I simply commented it was possible

Not enough info of the specific case to even hazard an estimate of probability.

Great to see you dropping by here with some great input :thumbsup: Thanks!

You're welcome

Had to have my XS fix... was having withdrawal symptoms

**bachus_anonym** · 05-07-2005, 02:12 PM

I see that you have all our emoticons figured out here

You see what you did now... I just grabbed some tools and started an autopsy on that 24-pin PSU I got...
I need to get better in that field, electronics I mean

**Krohling** · 05-07-2005, 09:47 PM

Ok, I dont understand much about electronics

But in the DFI LP UT NF4 Ultra-D motherboard, the connections between the +12V, +5V, +3.3V and Ground are phisically connected to the other regular 12V input, 5V input, etc

So, if there is a short circuit between this lines in the motherboard itself, how would a 24 pin PSU act different, you would cause a short circuit between the first 12V line, and the other 12V line?

If you understand what I asked, please try to clarify, because I didn't understand. I expected the extra 4 pins were unconnected from the other same lines in the board layout.

**Pjotr** · 05-08-2005, 01:47 AM

Let me try a simple answer: No, they are absolutely NOT short circuited.

They are kept separate because they feed different parts of all the electronics on the motherboard. However, the PSU itself can often draw the power from the same source for the different feeds. A budget PSU will draw all lines form the same power source, meaning if you load the +12V rail, you have less available for +3.3 and +5. On a high end PSU, they are all from different sources. This means multiple power sources in the PSU and hence a higher cost.

This also means there is a very big difference from a budget 600 W PSU and a high end 600 W PSU. Often a high end 300 W PSU will be able to power more equipment stable than a budget 450 W PSU.

**Krohling** · 05-08-2005, 09:45 AM

Well man, so my multimeter may be defective because if I disconnect all power connectors and try only the motherboard, there is contact between the 12V connector of the "regular" 20 pins and the 12V connector of the "extra" 4 pins.

**EMC2** · 05-08-2005, 06:25 PM

Originally Posted by Krohling

But in the DFI LP UT NF4 Ultra-D motherboard, the connections between the +12V, +5V, +3.3V and Ground are phisically connected to the other regular 12V input, 5V input, etc

Correct

Originally Posted by Krohling

how would a 24 pin PSU act different, you would cause a short circuit between the first 12V line, and the other 12V line?

No, I am aware of no PS's which have seperate output rails for the extra 4 connections of a 24-pin connector.

First an observation - just because different connections are electrically the same (read basically short across a conventional DMM), does not mean that they are all tied into the same power planes, and in fact could be tied to seperate planes that are only referenced together through a single via or trace (don't know if this is the case in the DFI Nf4 or not for any of the rails).

But... the difference is as a result of what I previously commented on - it deals with IR drops, trace inductance, frequency response (yes, power distribution systems have frequency response components), and noise issues - also known as the real world of electronics and parasitic reactances.

To give an example of how it likely could make a difference in the specific case of the DFI motherboard....

Present day CPUs and memory and GPUs all have very high transient current requirements (i.e. the current they require can change quite rapidly by a considerable amount in very short time frames measured in milli-seconds and even micro and nano-seconds). Up to a limit, determined by the amount and quality of the decoupling of the power locally (i.e. on the MB, DIMM, or Gfx card), this can cause a drop in the supply rails under certain conditions.

Normally the local decoupling (provided by capacitors, which store charge and resist changes to voltage caused by changing current demands) handles these temporary requirements.... BUT... there is a limit after which they "run out of charge" and can no longer supply the extra power being demanded. At that point, the supply rail starts dropping rather rapidly - not a good thing at all. To prevent this occurence, there must exist the capability to supply more current from the power supply before the capacitors stop compensating for the increased current demand. This is where that 24-pin connector (and the type of wiring used between it and the PS) can make a big difference.

Each wire between the power supply and the MB has inherent parasitic values - it isn't a 0 ohm, 0 inductance connection. The first of these is resistance. If the resistance is too high, the increased temporary current demand can cause a drop in the input voltage being supplied - causing the situation to worsen. The resistance in the wires also affects the amount of time it takes to replenish the charge of the onboard capacitors - possibly leading to their inability to handle the next transient current demand if it comes soon enough.... or maybe the 3rd one, etc.. or maybe just in those instances when everybody is depending increased current all at once - leading to those "pretty" windows BSODs.

The other aspect of the wire is that it has inductance. Inductance is a material's (the wires in this case) "resistance" to a change in current (versus the capacitors "resistance" to a change in voltage). Note that in the case of both capacitance and inductance this "resistance" (properly called impedance) is dependent on frequency. Now inductance can be either beneficial or detrimental to a circuit, depending upon the circuit (as can capacitance). In the case of power distribution (not power conditioning or conversion) inductance has a pronounced negative affect. When that sudden large change in current requirement occurs, the inductance of the wire resists the change and causes a lag in time before the increase in current demand can be met by the power supply.

In regards to how it affects noise - the overall impedance (the frequency dependant resistance of a component or system) determines the amount of noise coupling between two seperate, but adjacent, circuits. The lower the impedance of a connection, the smaller the magnitude of noise that will be coupled to it by an adjacent circuit. Increased inductance and resistance (in the case of no real net change in capacitance) causes the impedance to be higher - causing more noise to be coupled, in this case to the connections between the power supply and the MB.

And the last piece of the puzzle as it were... in the case of both resistance and inductance... when you parallel components (in this case the wires) with these characteristics, the result is a decrease in the net (total) value. Put two resistors or inductors of equal value in parallel and the value is halved. Put 3 in parallel and the value is reduced to 1/3rd. etc., etc.

And of course... the actual connector pins themselves also have these same circuit properties as well.

So, by adding the extra wires and connector pins between the power supply and the MB, the transient response characteristics of the power distribution system is improved, as is the noise characteristics, leading to cleaner, more stable power for all those OC'd goodies on your MB

Hope that answers your Q

Peace

**Krohling** · 05-08-2005, 06:53 PM

Very complex, but answer

Ty

**EMC2** · 05-09-2005, 05:30 PM

Made it as clear and short as possible while still trying to answer your Q completely... left out some of the more complicated issues. The short-short answer was my comment 3 in post #76

YW

**thesaucier** · 05-09-2005, 07:03 PM

Originally Posted by EMC2

Correct

No, I am aware of no PS's which have seperate output rails for the extra 4 connections of a 24-pin connector.

First an observation - just because different connections are electrically the same (read basically short across a conventional DMM), does not mean that they are all tied into the same power planes, and in fact could be tied to seperate planes that are only referenced together through a single via or trace (don't know if this is the case in the DFI Nf4 or not for any of the rails).

But... the difference is as a result of what I previously commented on - it deals with IR drops, trace inductance, frequency response (yes, power distribution systems have frequency response components), and noise issues - also known as the real world of electronics and parasitic reactances.

To give an example of how it likely could make a difference in the specific case of the DFI motherboard....

Present day CPUs and memory and GPUs all have very high transient current requirements (i.e. the current they require can change quite rapidly by a considerable amount in very short time frames measured in milli-seconds and even micro and nano-seconds). Up to a limit, determined by the amount and quality of the decoupling of the power locally (i.e. on the MB, DIMM, or Gfx card), this can cause a drop in the supply rails under certain conditions.

Normally the local decoupling (provided by capacitors, which store charge and resist changes to voltage caused by changing current demands) handles these temporary requirements.... BUT... there is a limit after which they "run out of charge" and can no longer supply the extra power being demanded. At that point, the supply rail starts dropping rather rapidly - not a good thing at all. To prevent this occurence, there must exist the capability to supply more current from the power supply before the capacitors stop compensating for the increased current demand. This is where that 24-pin connector (and the type of wiring used between it and the PS) can make a big difference.

Each wire between the power supply and the MB has inherent parasitic values - it isn't a 0 ohm, 0 inductance connection. The first of these is resistance. If the resistance is too high, the increased temporary current demand can cause a drop in the input voltage being supplied - causing the situation to worsen. The resistance in the wires also affects the amount of time it takes to replenish the charge of the onboard capacitors - possibly leading to their inability to handle the next transient current demand if it comes soon enough.... or maybe the 3rd one, etc.. or maybe just in those instances when everybody is depending increased current all at once - leading to those "pretty" windows BSODs.

The other aspect of the wire is that it has inductance. Inductance is a material's (the wires in this case) "resistance" to a change in current (versus the capacitors "resistance" to a change in voltage). Note that in the case of both capacitance and inductance this "resistance" (properly called impedance) is dependent on frequency. Now inductance can be either beneficial or detrimental to a circuit, depending upon the circuit (as can capacitance). In the case of power distribution (not power conditioning or conversion) inductance has a pronounced negative affect. When that sudden large change in current requirement occurs, the inductance of the wire resists the change and causes a lag in time before the increase in current demand can be met by the power supply.

In regards to how it affects noise - the overall impedance (the frequency dependant resistance of a component or system) determines the amount of noise coupling between two seperate, but adjacent, circuits. The lower the impedance of a connection, the smaller the magnitude of noise that will be coupled to it by an adjacent circuit. Increased inductance and resistance (in the case of no real net change in capacitance) causes the impedance to be higher - causing more noise to be coupled, in this case to the connections between the power supply and the MB.

And the last piece of the puzzle as it were... in the case of both resistance and inductance... when you parallel components (in this case the wires) with these characteristics, the result is a decrease in the net (total) value. Put two resistors or inductors of equal value in parallel and the value is halved. Put 3 in parallel and the value is reduced to 1/3rd. etc., etc.

And of course... the actual connector pins themselves also have these same circuit properties as well.

So, by adding the extra wires and connector pins between the power supply and the MB, the transient response characteristics of the power distribution system is improved, as is the noise characteristics, leading to cleaner, more stable power for all those OC'd goodies on your MB

Hope that answers your Q

Peace

Thank you for the very informative responce to my question!!
This still leads me to believe the only thing that can happen is instability. If wires were soldierd strait to the bridge rectifirer or the point in the power supply were the +12 volt lines ect. come from would you not over come many of the things you speak of. Ie line resistance and noice ect.? If the power supply had the amp and watt cappability of cource?

**EMC2** · 05-09-2005, 07:35 PM

YW2

BTW...read last sentence of post #76 above

The things I spoke of all deal with the connections between the PS and the MB... the wires and the connector... ( didn't even go into PWB/trace parasitics and reactances... nor wire types, skin effects,,,).... so short answer is no.

Peace

**thesaucier** · 05-09-2005, 08:24 PM

Originally Posted by EMC2

YW2

BTW...read last sentence of post #76 above

The things I spoke of all deal with the connections between the PS and the MB... the wires and the connector... ( didn't even go into PWB/trace parasitics and reactances... nor wire types, skin effects,,,).... so short answer is no.

Peace

I would like to ask you this seeing as you seem to be knowledgable.
What abou running two power suppies?
One for the board and one for all the drives and fans and periferals?

**shiznit93** · 05-10-2005, 04:12 PM

i got all the parts i need for my new rig, but i can't wait anymore for bfg to send me my replacement psu. i'm gonna try it with my old antec 430 20pin. i don't have an adaptor so how do i plug in the 20 pin connector? which side should the unconnected 4 pins be on?

**EMC2** · 05-10-2005, 04:51 PM

--- thesaucier ---

Yes, it can be done... 2 important things - you must have a good common ground connection between the two supplies (would suggest a HDD molex<->molex adapter cable with only the ground connection wired in it)... you want to make sure that you don't have any of the power rails directly connected together. Oh, and try to adjust power rails within 0.2V of each other.

--- shiznit ---

The connector is "keyed" such that it should only go in one way (properly)... but the extra pins are 21->24... look for the Pin 1 identifier on your MB and the cable plugs into that "side" of the connector.

Peace

**shiznit93** · 05-10-2005, 05:32 PM

thanks, its just gonna be for a couple of days til bfg get their act together. will i run the risk of damaging anything by using the 20pin psu. i'm not gonna overclock the venice until my rma bfg 550sli psu gets here.

**thesaucier** · 05-11-2005, 11:40 AM

Originally Posted by EMC2

--- thesaucier ---

Yes, it can be done... 2 important things - you must have a good common ground connection between the two supplies (would suggest a HDD molex<->molex adapter cable with only the ground connection wired in it)... you want to make sure that you don't have any of the power rails directly connected together. Oh, and try to adjust power rails within 0.2V of each other.

--- shiznit ---

The connector is "keyed" such that it should only go in one way (properly)... but the extra pins are 21->24... look for the Pin 1 identifier on your MB and the cable plugs into that "side" of the connector.

Peace

If I connected the common lines togetther from each power supply and connect the power on line from each pwer supply will that work for good common ground and also start both power supplies at the same time when the computer power on button is hit?

**EMC2** · 05-11-2005, 04:54 PM

--- shiznit ---

Should be no issue with damaging anything at stock settings... but I would still suggest getting a 20->24 adapter and modding it as I describe earlier in this thread.

--- Saucier ---

With a good connection between grounds on the two supplies (for example via a HDD pwr cnx) your good there.

Yes, you can connect the "Power_On" signals together for the two supplies... it's a TTL input signal (to the PS's) that goes low to turn the PS's on...

Peace

**shiznit93** · 05-11-2005, 08:18 PM

I have to say that bfg has THE BEST customer service I have EVER dealt with. All the reps I talked to were very polite and did whatever they could to help me. I got my replacement psu within a week of them receiving the broken one, they shipped it monday and I received iy wednesday afternoon. Looks like i won't have to bother with the antec after all, tomorrow is gonna be one fun day.

**Formann** · 05-11-2005, 10:30 PM

Been running a Enhance 460w 20-pin ever since I got my DFI.

I have done a 20 -> 24pin mod on it, so im not if that counts

Running Clawhammer 3500+ @ 3.1+ with 1.85vcore and 3.6+ vdimm.. nice and stable.

**Cyrilix** · 05-14-2005, 07:51 PM

Originally Posted by mrlobber

Many many great 24pin PSU's are just 20pin + extra 4 pins added from the same lines, just as bachus_anonym and some others already said. This applies to the PCP&C 510W SLI as well (notice that PCP&C will rewire your existing 20pin PSU (vanilla 510W for example) to a 510W SLI model for a $50 charge), which indicates the PSU doesn't actually conform with the ATX2.0 standard (dfi-street.com had a hefty discussion on that "conforms" or "is compliant" stuff as well). Yet it is even in the nVidia SLI certified list.

Or take the OCZ Powerstream 520W which is not an ATX2.0 PSU as well but is considered to be a fine PSU to run with DFI. This leads me to believe that the "absolute must of a 24pin" is a bit too harsh stressed because I've not seen any documented explanation whether such a 24pin PSU (which is essentially a 20pin one) is worse than a 24pin ATX2.0 PSU with split +12v rails when used with the DFI motherboard. The dfi-street.com guys for the most part have avoided this discussion by simply using their main weapon "24pin PSU is a must" to bring down any questions on this subject (but I understand the behaviour if a user with a crappy 350W 20pin asks "why do I have trouble"?) However, if it is the case that many "24pin" PSU's are considered to be fine, this means a quality 20pin PSU should work without any problems except maybe the most extreme cases where the PSU complete power itself would be challenged (and I don't see 300htt or 2.6 Ghz CPU as an extreme overclock especially if you're not running SLI).

So this comes down to trying and isolating the cause of instabilities, and if they persist (by swapping memory, cpu's, maybe even motherboards), it's worth looking into your quality 20pin PSU direction, just as mdzcpa said.

You said that the OCZ Powerstream 520W is not ATX2.0 compliant. On the specifications for OCZ Powerstream, however, it tells a different story.

http://www.ocztechnology.com/products/PSUSpec.pdf

Only the 600W one is ATX2.01 compliant, but I don't think that's a requirement for the DFI board, is it?

**longsiew** · 05-17-2005, 09:48 PM

I am running my DFI NF4-ultra D on 20 pin PSU. It is icute 420watt. Max power is actually 395.5watt. But, there is flactuate on the voltage supplied. It is always over volt by 0.1 volt. So far, my comp is stable but died on me due to flashing of Bios.

**Ubermann** · 05-17-2005, 10:32 PM

Can someone show a pic of what volt the 4 extra pin should be feed with ?
Been reading several threads about this but still havent seen what cable go where.
But i would guess i am now an electric engineer after all these deep into electric conversations, maybe i build my own 24 pin psu =)