# How to make two computer power supplies work together

Status
Not open for further replies.

#### blueroomelectronics

##### Well-Known Member
LOL there was a poster looking for a way to glue together a half dozen or so 65W mini-atx supplies for his Warcraft supercomputer to make it somehow more portable...

#### arhi

##### Member
/me under the table

I saw much in the "old" days .. when everything was expensive... (I remember my first amber hercules screen in 85 ... wooooow that was expensive)... but today, everything is there, it is available and it is not expensive.. my phone have more ram then my first hard drive ... but that's another story

"glue together a half dozen or so 65W mini-atx"

#### blueroomelectronics

##### Well-Known Member
I can't make this stuff up, search for "warcraft" in these forums, it's there.

My Commodore PET 2001 cost me $1695 CDn in 1976. I worked all summer for it. Now it sits in a corner of my basement. #### notauser ##### New Member When I cracked open my 8088 beast after I got a 286, I thought it would be funny to relabel (as professionally as possible in 1990) the 30MB hdd to say 30GB.. At the time I thought that number would seem outrageous for centuries.. I miss the days when only cool ppl knew how to use a computer.. #### arhi ##### Member I can't make this stuff up, search for "warcraft" in these forums, it's there. My Commodore PET 2001 cost me$1695 CDn in 1976. I worked all summer for it. Now it sits in a corner of my basement.

Oh I believe you I was just crazy with laughter

as for commodore ... I got C64 soon after it was made; I was young but if my memory serves me well, that was new year 1983 (01.01.1983), I got it for my birthday that's why I remember the date . Few year later (summer '84 or '85 cannot say for sure) I got XT (some german brand, with some plastic card with holes as a key ) with 512K RAM and 20M HDD (and only 5.25" floppy) ... that was the box .. hercules ... IIRC, the C64 was ~8000DEM and XT was (with herc+amber screen and star lc10 color printer) around 10000DEM. As we can agree that 1dem then is pretty much like 1eur today (maybe even more) imagine the computer you can get today for 10KE

#### hazardouzx

##### New Member
the true way to connect two psu's into one computer

Cut the green and grey wires on both psu's.
Make sure to cut half way on main psu. Reconnect the wires on the main psu.
That way the wires on the main psu is in a v shape. Now connect the green and grey wires from the other psu to the main psu. Remember green to green and grey to grey.

Therefore plug cords into power outlet and press the power on your computer. Both psu then will start up running smooth and great.

WATCH THIS VIDEO I FOUND https://www.youtube.com/watch?v=BviJjNYsIag

I AM RUNNING MY COMPUTER LIKE HOW I SAID AND HOW IT SHOWS IN THE VIDEO FOR 2 YEARS NOW NEVER HAD A PROBLEM

Last edited:

##### Well-Known Member
This thread is soon to be two years old. However, in the interest of contributing something of value I have attached a few images that clearly show my solution to the power demands of today's graphics cards! Graphics cards rely on 12 volt power! Simple enough. I should point out this method works just fine.

Ron

#### Attachments

• BATT 1.jpg
100 KB · Views: 218
• BATT 2.jpg
139 KB · Views: 269

#### cng1024

##### New Member
I've investigated the possibility of using multiple power supplies for quite some time now. At one point, I even found a couple products that will do the job. In fact, I found this review https://www.hardocp.com/article/2011/08/24/black_art_dual_psus_in_your_enthusiast_pc/1 of one auxiliary and one redundant power supply and two kits that allow multi-power supply configurations. I've been planning a gaming PC build for a couple months now, and have been unable to find a power supply that provides > 1300 watts from a retailer and manufacturer I trust. (I am quite particular about who I give my money to.) In addition, most of my electrical outlets (120VAC) are connected through a 15 amp circuit breaker. Considering monitors, printer, and allowing room for aging breakers, I probably wouldn't want to power much more than 1300 watts from one 15 amp breaker anyway. The obvious solution, that is the one not requiring house rewiring, is to split the load between a couple different 15 amp circuits.

There is a misconception about differences between power and data connections within a computer system. To me, it seems incredible that even those who are otherwise knowledgeable with regard to computers, are apparently unaware that data lines from component to component in a computer are generally electrical connections with no real isolation. For example, it seems a common misconception that powering a hard drive from a separate power supply means the drive is electrically isolated from the motherboard, even though there is a data cable connecting the two components. Even if this is true of some special components, it is not standard design practice and as such, you must assume any component connected to the computer, either by power or by data lines, is not electrically isolated from the computer. Such an assumption could be both destructive and dangerous.

For example, consider a situation involving two power supplies powering different components on the same computer. Say PSU A is powered from a 120VAC 10 amp house circuit and PSU B is powered from a different, 120VAC 15 amp circuit, with ground +5V higher than PSU A. (Remember, voltage is simply potential difference from one point to another and ground is relative.) Now lets say we have a motherboard powered by PSU A connected to a hard drive powered by PSU B. At best, the motherboard chipset and hdd electronics won't be able to communicate and your system thinks the drive isn't connected. It is quite possible for the motherboard chipset to be destroyed though. Remember, the hard drive's ground is +5 volts to the motherboard, so when the drive tried to communicate, all of it's high data, the ones, were around +10v relative to the motherboard.

I was wondering if anyone would mention the PWR_OK control line and also wish to address the comment from hazardouzx:
the true way to connect two psu's into one computer

Cut the green and grey wires on both psu's.
Make sure to cut half way on main psu. Reconnect the wires on the main psu.
That way the wires on the main psu is in a v shape. Now connect the green and grey wires from the other psu to the main psu. Remember green to green and grey to grey.

Therefore plug cords into power outlet and press the power on your computer. Both psu then will start up running smooth and great.

As I understand it, this suggests a simple splice, connecting PWR_OK from both power supplies together. In my opinion, a simple splice is unwise and may even cause damage at some point. The PWR_OK line is a TTL signal, 0V indicating the power is not ok and +5V when the power has stabilized. Suppose there are two power supplies connected, PSU A and PSU B. Now, what is PSU B is slower than PSU A, or worse, PSU B has failed and will never signal power good? PSU A PWR_OK is now at +5V but PSU B PWR_OK is at 0V. Hopefully the circuits are current limited or the power supplies turn off if a short develops. If you aren't so lucky, one or both power supplies and possibly other components could be damaged.

I do think multi-power supply configurations can work, but things must be well thought out. The supplies should be made to share a common ground, provided by some fairly substantial wire or circuit traces. The various supply voltages should never be connected and each component should be powered exclusively by one power supply.

The PWR_OK line should not be ignored and should not be spliced together to create a single PWR_OK line. I believe the correct solution for the PWR_OK line is to connect the line from each supply to a TTL AND gate input. The output can then be connected to the motherboard side of the PWR_OK line and, if you want to get fancy, you can add LED indicators for easy diagnosing should a power supply fail to provide power good signal. Such an interface circuit could easily be designed to be powered from one of the PSU standby lines.

Last edited:

##### Well-Known Member
Before I even begin to comment I thought my solution, with images provided, was a viable solution to power needs in a high end system. As can be clearly seen great thought went into my design.

I will confess it has been several years since I was active with computer hardware and it's power demands, especially with a focus on today's graphics cards. For a home computer system, even a high end workstation or gaming system I believe the processing power demands peaked with likely the Intel Pentium 4 as GPU cards continued to demand more and more power.

I watched with some interest as the Great PSU Wars began where I saw just about every marketing trick deployed and even learned a few new ploys used by manufacturers to deploy their products. The concept of dual or redundant power supplies was nothing new then and apparently is somewhat still alive and well today. Along with a selection of tips and tricks to make things work. The only time I have used or had a need for a dual PSU redundant system is in a mission critical industrial application and in that scenario there is only one duty PSU and a backup PSU to assume the power role should one fail.

My little philosophy on such systems is simple. My thinking is that if someone wants to invest a thousand (or more) US Dollars in an extreme gaming machine graphics cards they should be able to invest in a single, reliable, PSU to power the thing. That includes the use of a commercial grade PSU built with good design practices. To invest a few thousand dollars in graphics cards and then cobble two power supplies together to power them seems foolish to me.

As to this quote:
the true way to connect two psu's into one computer

Cut the green and grey wires on both psu's.
Make sure to cut half way on main psu. Reconnect the wires on the main psu.
That way the wires on the main psu is in a v shape. Now connect the green and grey wires from the other psu to the main psu. Remember green to green and grey to grey.

Therefore plug cords into power outlet and press the power on your computer. Both psu then will start up running smooth and great.

I agree that is flat out wrong for the very reason(s) you covered and the correct method would be the use of an AND gate to insure each PSU had a PWR_OK status. Anything less defeats the logic behind the PWR_OK signal.

As to residential wiring in the US. Should there be a 5 volt RMS difference between grounds from outlet to outlet a problem exist. The same would be true for a 5 volt RMS difference between neutrals in an older wiring scheme sans the ground. I just do not see a potential difference that great happening in a residential US system that complies with NEC code. A few mV yes but nothing like 5 volts. However, if we assume such a condition did exist it would be cancelled within the case chassis. Taking the case of a HDD for example. If I measure the resistance between a HDD power connector ground and the case of the HDD, with just the HDD lying on the bench, I will find they are electrically the same point. Therefore if I were to power the HDD from a PSU connected to an outlet with a 5 volt RMS voltage difference than the PSU powering the rest of the system the grounds would be common to each other and no potential difference could exist. So I fail to see where the data stream common would be offset?

As to PSU power demand and running a PSU on a 15 amp circuit. If we assume (and this is a big assumption) the PSU was in fact a 1300 watt PSU capable of delivering a full 1300 watts across all it's rated outputs (keep cross loading in mind) it would not draw 1300 watts at the mains power connector. It would draw considerably more. In reality if it was 80% efficient it would draw about 13 amps at the mains which as continuous duty would really be pushing a 15 amp service line in a residence. OK, so a decision is made to split the power demand between two different service lines and run let's say 750 watt PSUs. To my way of thinking it becomes easier for me to run a new 20 amp AWG 12 service branch with ground than to screw around using different branch outlets. Just my thinking on that note.

Looking at a PSU by design specifications. The 12 volt rail for example should be 12 volts +/- 5% so anything between 11.4 and 12.6 volts is acceptable. So what happens to regulation if one PSU outputs 11.5 volts (in spec) and the other outputs 12.5 volts (in spec). How will each PSU handle regulation? Even if I use one PSU strictly for the PCIE Aux connectors on the GPU card. Does the GPU also draw 12 volt power from the PCIE motherboard connector powered by another PSU? Should at any point those two 12 volt lines share a bond what will happen?

That being just my take on some of this involving the use of parallel PSUs to acheive more power.

Ron

#### cng1024

##### New Member
To invest a few thousand dollars in graphics cards and then cobble two power supplies together to power them seems foolish to me.
As a means of reducing cost, I would definitely agree that this is not a good way to do it. In fact, when I build a new computer, the power supply is one of the most thought out component in the build.

Should there be a 5 volt RMS difference between grounds from outlet to outlet a problem exist. The same would be true for a 5 volt RMS difference between neutrals in an older wiring scheme sans the ground.
I also agree with you here. I was simply making a point and picked a number at random.

However, if we assume such a condition did exist it would be cancelled within the case chassis.
I see this argument quite often in computers. The problem here is, you assume there is a proper chassis ground connection. Considering the plethora of aftermarket computer products, things like painted mounting screws and silicon grommets, a sound electrical connection between component and computer chassis cannot be guaranteed. I own an Antec computer case that uses non-conductive silicon grommets (pre-installed from factory). The mounting screws never touch the computer case. For some drives there is electrical contact, but for others there is nothing. Then you must also consider the case itself. There are many plexiglass cases on the market and even a few wood cases. These obviously are electrical insulators. The point is, you cannot rely on the presence or absence of a chassis ground.

If we assume (and this is a big assumption) the PSU was in fact a 1300 watt PSU capable of delivering a full 1300 watts across all it's rated outputs (keep cross loading in mind) it would not draw 1300 watts at the mains power connector. It would draw considerably more.
Agreed. If you do the calculations, even the specs given for most power supplies indicates that they deliver nominal power that is less than rated maximum.

To my way of thinking it becomes easier for me to run a new 20 amp AWG 12 service branch with ground than to screw around using different branch outlets. Just my thinking on that note.
I suppose that depends on your situation. I know how to put in my own electrical lines, but am not in the best of health, so the job becomes much more difficult. There are also many who don't have the authority to do these upgrades. For example, apartment dwellers.

Looking at a PSU by design specifications. The 12 volt rail for example should be 12 volts +/- 5% so anything between 11.4 and 12.6 volts is acceptable. So what happens to regulation if one PSU outputs 11.5 volts (in spec) and the other outputs 12.5 volts (in spec). How will each PSU handle regulation? Even if I use one PSU strictly for the PCIE Aux connectors on the GPU card. Does the GPU also draw 12 volt power from the PCIE motherboard connector powered by another PSU? Should at any point those two 12 volt lines share a bond what will happen?
This is probably the main reason why I haven't attempted installing a second PSU. What happens if the regulators should 'cross paths' is likely going to be nasty. If one regulator is content to run at 11v4 and the other is content to run at 12v6, you basically end up with a tug of war. One vreg compensates upward in an attempt to keep the voltage at 12v6 while the other compensates downward in an attempt to keep the voltage at 11v4. This could theoretically cause massive current flow through whichever component(s) connect the two vregs.

I must admit that two power supplies probably aren't the greatest idea. At least, not without a lot of research to prove it will work. The fact that there are auxiliary power supplies out there, designed for just this purpose lends some credence to this idea. If one is willing to try it, they should know that there is significant risk involved. Personally, I'd be interested to build a proper circuit to try it out, but I wouldn't be doing those tests on a new machine.

Last edited:

##### Well-Known Member
You make a very good point as to the Antec case using the screws through the grommets. I have built my share using that configuration and the thought never crossed my mind.

It has been several years since I was really involved with hardware. I was active with a few hardware evaluation sites several years ago. Mostly involving PSU evaluations and setting up test benches for PSU testing. There were a few lessons learned in that line. Just as a side note something I always found interesting was the advertising. Most of the reviews were canned in that they always began the same way: Today we will be reviewing the Antec 750 True Power followed by blah, blah, blah.

Now if you see banner advertising paid for by the manufacturer or distributor splashed on the page do you really figure you will get a 100% independent review as if they were an independent lab? That advertising pays the bills and site cost. The cool part was you got to keep plenty of high end hardware. However, you don't write bad reviews for companies that are paying to advertise on your site. Overall, I just got tired of all the ********.

Over those years I did see the dual PSU issue surface over and over again. I kicked around a few circuits to do it but generally with one issue solved another issue would surface. The concept being to make something the average home user or enthusiast could easily use short of hacking and redesigning two power supplies on the kitchen table to get them to work together as one. All of that discounting how to make them fit the average case.

As to the testing? My view was that if you want to test a PSU you should test in accordance with a PSU design guide similar to this one with well defined test methods and procedures. My design guide link may be a rev or two behind. Anything less will not give a true indication as to the products worth. That being just my opinion and not gospel. It was fun designing load banks and automated test systems though.

The more I got into running dual power supplies in parallel the more little problems surfaced. Oh hell, they would work OK but I could not say for how long. Some things, exactly along the lines you mentioned. You can't simply tie two PWR_OK lines together and market the concept. Yeah, works fine unless one PSU fails. The list goes on. While I never overly pursued the regulation aspect it always bothered me.

Then there is the concept of only adding additional 12 volt power. Looking at today's systems the large power demand is on the 12 volt supply. So why screw around with 3.3 or 5 volt power? There were a few small units marketed that were dedicated to 12 volt power only. I am not sure how those fared out. The concept looked good.

If I may ask, what are you building? What graphics cards will you be using? While I don't game I generally use high quality graphics cards when I do build myself a new system. However, they are frequently along the lines of commercial workstation cards, which actually sort of suck for gaming.

While I don't have an earth shattering solution for paralleling two PSUs it does make for interesting conversation.

Ron

#### cng1024

##### New Member
I always wondered about hardware reviews. Who paid the bills and what strings are attached for those fortunate enough to land an interesting component a couple months before release. I admit, the free high end hardware would be tempting, but I also have a predilection for brutal honesty so I doubt I'd have much hardware to review. lol

We seem to have similar views with regard to design and testing philosophy. When I started researching multi-psu configurations, the first thing I did was to download the atx 2.2 development guidelines and specifications. Unfortunately, these specs don't seem well suited to handle multiple power supplies working in parallel.

A 12 volt only aux supply would be quite useful for most 'enthusiast' computer users, at least for video cards. Not sure I'd try powering the cpu with an aux power supply.

I agree, parallel psu config is quite interesting. But then, I am a nerd and as such, just about anything electronic or scientific is fascinating to me! lol

Here is a preliminary list of components for the new rig:

Silverstone Strider 1500W Modular Power Supply - **broken link removed**
Intel Core i7-3930K Socket 2011 Processor - **broken link removed**
X79A-GD45 Gen 3 w/ DDR3 2400, 7.1 Audio, Gigabit Lan, PCI-E, Triple CrossFireX / SLI - **broken link removed**
GSKILL Ripjaws X Series 8GB PC3-17000 Dual Channel Kit (2 x 4GB) (x2) - **broken link removed**
eVGA GeForce GTX 580 3GB (possibly x2 SLI config) - **broken link removed**
LG BH12 Super Multi Blue 12x Internal SATA Blu-ray Disc Writer, Black w/ Lightscribe - **broken link removed**
Western Digital VelociRaptor 600GB 10,000rpm SATA III w/ 32MB Cache - **broken link removed**

I'll want to do further investigation for most of the parts on this list before I'm satisfied of their quality and performance. I'll be waiting a couple months before I start seriously considering a purchase. One thing I'm interested in is solid state drives. I just can't quite convince myself of their reliability though. I'm not sure if I'm just old school, or don't like that there is a limited number of writes for ssd devices. Any input provided on the parts list, or ssd devices would be appreciated.

##### Well-Known Member
You are going to have yourself one hell of a gaming machine build looking at the parts selection. As to the solid state drives, I have yet to use them. However, I was quite active in a computer forum and a member named "Chumly" did extensive work with them. The guy was a wealth of information. Overall we had several members who were very knowledgeable on SSDs. I would suggest you visit a forum dedicated to home computing like Maximum PC (based on the magazine) and get more info. As I mentioned, I have not been active with computers (newest latest and greatest) for a few years and hardware changes as often as I change socks.

I very much agree with doing extensive homework and research before building anything. I always lived by that rule and it is good practice. This was especially true when building higher end workstation systems.

Something you may want to invest in is one of those little Kill-A-Watt devices. They come in handy and are relatively inexpensive. It's nice to know the line power draw and they fit the ticket.

Looking at your choice of GPU and reading a little I personally don't see a need for a 1.5 KW PSU but obviously it won't hurt. The card is touted as a 250 watt card under full load.

Ron

#### cng1024

##### New Member
Definitely some good advice. I figured I might be a bit over the top with my psu choice, but this was a preliminary list with many tweaks yet to come. Still, when it comes to power supplies, I always consider the actual output to be around 20% lower than advertised. Considering this, and the possibility of future additions such as another video card or possibly RAID, I figure I'll want at least 1200-1250 watts.

On the topic of multi-psu configurations, I did a little deeper investigation with the products I referenced earlier. It seems that even these products don't properly address the PWR_OK line and they do rely on chassis grounding to make the grounds common. In my view, these solutions introduce an unacceptable risk for reasons stated above. It does make for good debate, and perhaps even a neat project, but it would be useless for me because I wouldn't power that awesome system listed above with multiple power supplies. I figure, if I want to make useless electronics, I could do something more fun like a POV display or handheld pong video game.

##### Well-Known Member
You pretty much called it. Most of the stuff out there does not address everything that should be addressed to configure a pair of PSUs to run in parallel. All that glitters is not gold, even when a hardware review site seems to figure it is.

RAID is something I include in everything I build. Generally RAID 1 consisting of two HDDs for the OS and installed programs. Then RAID 5 using 4 HDDs for mass storage and backup. Unlike years ago, HDDs are relatively inexpensive. That works for me as I really don't have the need for speed. The real beauty of building is you can build around your own personal wants and needs.

Enjoy your new system when you get it together. Maybe this spring I'll build myself one after we invest a few bucks in the house. My workstation could also use an upgrade.

Ron

#### CharlyV

##### New Member
People posting negative comments about this should understand DC theory and how switching power supplies work. As long as the commons are tied together, you are not defeating any of the power regulation in each supply. This is a great solution to having one power supply service the motherboard and fans, and the other can be used for GPU's, Disk Drives, CD Drives. The pullup transistor and resistor are the perfect solution for ensuring both supplies start at the same time when the power switch is toggled. You basically need very little work or skill to implement this, and I think the OP did a great job of presenting it. It is without a doubt, very practical, economical as well as safe.

Status
Not open for further replies.

Replies
4
Views
1K
Replies
2
Views
269
Replies
16
Views
2K
Replies
3
Views
3K
Replies
13
Views
2K